(Trends in Linguistics. State-Of-The-Art Report - 22.) Jacek Fisiak-Contrastive Linguistics - Prospects and Problems-Mouton Publishers

Contrastive Linguistics
Trends in Linguistics
Studies and Monographs 22
Editor
Werner Winter
Mouton Publishers
Berlin New York Amsterdam
Contrastive Linguistics
Prospects and Problems
edited by
Jacek Fisiak
Mouton Publishers
Berlin New York Amsterdam
Professor Jacek Fisiak
Institute of English
Adam Mickiewicz University
Poznan, Poland
Papers prepared for the

4th International Conference of Contrastive Projects
held at Charzykowy, Poland, 3 - 6 December 1980.
Library of Congress Cataloging in Publication Data
Contrastive linguistics.
(Trends in linguistics. Studies and monographs; 22)

Based on papers presented at the 4th International
Conference of Contrastive Projects held in Charzykowy,
Poland, 3-6 Dec., 1980.
Includes index.
1. Contrastive linguistics Congresses. I. Fisiak,
Jacek. II. International Conference of Contrastive
Projects (4th. 1980 : Charzykowy, Poland) III. Series.
134.063 1984 410 8319430
ISBN 90-279-3260-3
Copyright 1984 by Walter de Gruyter & Co., Berlin. All rights reserved, including
those of translation into foreign languages. No part of this book may be reproduced in
any form - by photoprint, microfilm, or any other means - nor transmitted nor trans-
lated into a machine language without written permission from Mouton Publishers,
Division of Walter de Gruyter & Co., Berlin.
Typesetting: Mercedes-Druck, Berlin. - Printing: Druckerei Hildebrand, Berlin. -
Binding: Lderitz & Bauer Buchgewerbe GmbH, Berlin. Printed in Germany
Preface
The present volume contains a selection of papers prepared for the 4th Inter-
national Conference of Contrastive Projects held at Charzykowy (Poland)
from 3 - 6 December 1980. (Two papers have been accepted for publication
(i.e. Filipovic and Gussmann) although they were not presented at the con-
ference).
Unlike the previous conferences of contrastive projects in Zagreb (1970),
Bucharest (1975) and Trier (1978), the one at Charzykowy had no papers
summarizing the progress of particular contrastive projects. The organizers
assumed that the ideas and trends developed within different projects can be
most profitably presented by papers addressing themselves to substantive
issues in the field of contrastive linguistics. Therefore it was considered that
most interesting results would be achieved if some scholars were invited to
speak on specific topics and others to discuss problems of special interest to
them. It seemed to us that this would help to identify additional problems
for further research and would point to further uses of contrastive research
apart from the ones widely discussed over the last few years.
Those who expect to find definite answers to numerous, often contro-
versial questions posed by contrastive research will be disappointed. What the
conference has demonstrated is firstly the value of contrastive research for
finding solutions to theoretical linguistic issues, e.g. papers by Borsley, Grze-
gorek, Horn, Gussmann, Jaworska or Zabrocki. Secondly, the conference wit-
nessed a very productive broadening of the scope of contrastive research by
going beyond the more traditional sentence-bound orientation into the fields
of contrastive pragmatics, text linguistics and discourse analysis (e.g. Enkvist,
Fillmore, Faerch and Kasper, Oleksy and Szwedek). Thirdly, the conference
has proved the necessity for further refinement of some fundamental concepts
of contrastive linguistics (e.g. Krzeszowski and Filipovic). Fourthly, the con-
ference has made it clear that in spite of scepticism voiced by some scholars as
regards the application of contrastive research (cf. Khlwein), the present
research in the area of second language acquisition and psycholinguistics (e.g.
Sajavaara and Sharwood Smith), reinvestigation of error analysis (e.g. Mukat-
vi Preface
tash) and research in the field of translation leave no doubt that the uses of
contrastive linguistics are quite numerous and of a different calibre than ex-
pected at the outset of pedagogically oriented contrastive work.
In general it can be easily noticed that the volume contains more theoretic-
ally oriented linguistic works (for my distinction between theoretical and
applied contrastive analysis made earlier in 1968, 1970, 1973, 1975 etc. see
several references in this volume, e.g. papers by Khlwein and Sajavaara) al-
though the applied side has not been totally neglected. It seems that this
reflects the situation in contrastive linguistics in general if one looks both at
the history of the discipline and the amount of theoretical works produced
in both hemispheres nowadays.
The editor of this volume hopes, therefore, that the essays presented here
will be of interest both to theoretical linguists as well as those readers who
will look for some insight into the field of application of contrastive analysis.
The 4th International Conference on Contrastive Projects was sponsored
by Adam Mickiewicz University in Poznan and organized by its Institute of
English. On this occassion words of thanks are due here to Professor Stefan
Kozarski, Pro-vice-Chancellor of Adam Mickiewicz University, for substantial
financial support in the days of budgetary cuts and economic stringency.
Our thanks also go to Ms. Barbara Piocinska, M.A., for a skillful and
devoted handling of the administrative arrangements of the conference in all
stages of its organization.
Finally our gratitude is due to all the conference members whose active
participation has greatly contributed to the final shape of the present volume.
Poznaii, May 1981 Jacek Fisiak

Contents
Preface
List of conference participants ix
ROBERT D. BORSLEY
Free relatives in Polish and English 1
STEFAN DYtA
A note on inversion/conjunct ascension constructions in Polish and
English 19
NILS ERIK ENKVIST
Contrastive linguistics and text linguistics 45
CLAUS F/ERCH and GABRIELE KASPER
Ja und? - og hva s? a contrastive discourse analysis of gambits in
German and Danish 69
RUDOLF FILIPOVIC
What are the primary data for contrastive analysis? 107
CHARLES J. FILLMORE
Remarks on contrastive pragmatics 119
MARIA GRZEGOREK
English sentences with introductory there and their Polish counter-
parts 143
EDMUND GUSSMANN
Abstract phonology and contrastive analysis 171
RAYMOND HICKEY
Towards a contrastive syntax of Irish and English 187
GEORGE M. HORN
Constraints on transformations: evidence from contrastive analysis . 205
viii Contents
EWA JAWORSKA
On the structure of adverbial subordinate constructions in English
and Polish 251
ANDRZEJ KOPCZYNSKI
Problems of quality in conference interpreting 283
TOMASZ P. KRZESZOWSKI
Tertium comparationis 301
WOLFGANG KHLWEIN
Pedagogical limitations of contrastive linguistics 313
LEWIS MUKATTASH
Contrastive analysis, error analysis and learning difficulty 333
WIESLAW OLEKSY
Towards pragmatic contrastive analysis 349
JERZY RUBACH
Rule ordering in phonological interference 365
KARI SAJAVAARA
Psycholinguistic models, second language acquisition, and con-
trastive analysis 379
MICHAEL SHARWOOD SMITH
Learnability and second language acquisition 409
ALEKSANDER SZWEDEK
Some problems of contrastive analysis and text linguistics 419
TADEUSZ ZABROCKI
On the nature of movement rules in English and Polish 431
Index of names 445
List of conference participants
at the fourth International Conference of Contrastive Projects held at Charzy-

kowy, Poland, December 3 - 6 , 1 9 8 0
Director
Professor Jacek Fisiak Adam Mickiewicz University, Poznaii
Participants
Docent Wieslaw Awedyk Adam Mickiewicz University, Poznan
Professor John L. Behling University of Sofia
Dr. Robert Borsley University College, London
Dr. Paul van Buren University of Utrecht
Dr. Donna Christian Adam Mickiewicz University, Poznan
Docent Liudvika Drazdauskiene Vilnius State University
Dr. Stefan Dyta Adam Mickiewicz University, Poznan
Dr. Emilia Dykczak Adam Mickiewicz University, Poznan
Professor Nils Erik Enkvist bo Akademi
Dr. Claus Faerch University of Copenhagen
Professor Charles J. Fillmore University of California, Berkely
Dr. Jadwiga Fisiakowa Adam Mickiewicz University, Poznan
Professor Rosemarie Glser University of Leipzig
Professor Claus Gnutzmann University of Hannover
Ms. Elzbieta Gorska Adam Mickiewicz University, Poznan
Dr. Maria Grzegorek Adam Mickiewicz University, Poznari
Dr. Raymond Hickey University of Bonn
Dr. Gabriele Hohenwart University of Salzburg
Dr. Helga Hhlein Technical University, Aachen
Professor George Horn University of Newcastle, Australia
Dr. Aleksandra Horn University of Newcastle, Australia
Mr. Gerhard Jakob University of Mannheim
X List of conference participants
Docent Karol Janicki Adam Mickiewicz University, Poznad

Ms. Ewa Jaworska Adam Mickiewicz University Poznan
Docent Roman Kalisz University of Gdadsk
Dr. Gabriele Kasper University of Aarhus
Ms. Maria Kawinska Adam Mickiewicz University, Poznari
Docent Andrzej Kopczydski University of Warsaw
Dr. Barbara Kryk Adam Mickiewicz University, Poznan
Professor Tomasz P. Krzeszowski University of Gdadsk
Professor Wolfgang Khlwein University of Trier
Dr. Bogdan Lawendowski University of Warsaw
Dr. William R. Lee IATEFL, London
Dr. Barbara Lewandowska University of Lodz
Professor Grace Mancill American University, Washington, D. C.
Dr. Bogus!aw Marek M. Curie-Sktodowska University, Lublin
Ms. Brbel Miemietz University of Bonn
Professor Lewis Mukattash University of Jordan, Amman
Professor Ruta Nagucka Jagiellonian University, Cracow
Dr. Nina Nowakowska Adam Mickiewicz University, Poznari
Dr. Mirostaw Nowakowski Adam Mickiewicz University, Poznad
Dr. Wieslaw Oleksy Pedagogical University, Bydgoszcz
Ms. Janina Ozga Jagiellonian University, Cracow
Docent Olaf Paulsson University of Gteborg
Dr. Alicja Pisarska Adam Mickiewicz University, Poznan
Dr. Andrzej Pisarski Adam Mickiewicz University, Poznan
Ms. Barbara Ptocmska Adam Mickiewicz University, Poznan
Dr. Wilfried Rathay University of Rostock
Docent Jerzy Rubach University of Warsaw
Professor Kari Sajavaara University of Jyvskyl
Dr. Ewa Siarkiewicz-Bivand Adam Mickiewicz University, Poznad
Dr. Laura Sergo University of Salzburg
Dr. Michael Sharwood Smith University of Utrecht
Dr. Marek Szatek Adam Mickiewicz University, Poznari
Docent Aleksander Szwedek Pedagogical University, Bydgoszcz
Dr. Tomasz Wachtel University of Warsaw
Ms. Elizabeth White US Embassy, Warsaw
Ms. Ewa Willim Jagiellonian University, Cracow
Professor Werner Winter University of Kiel
Docent Tadeusz Zabrocki Adam Mickiewicz University, Poznan
Professor Luis Zuck . Curie-Ski-odowska University, Lublin
ROBERT D. BORSLEY
Free relatives in Polish and English
Bresnan and Grimshaw (1978) reject the traditional assumption that English
free relatives are headless. Instead, they argue that the wh-phrases are heads,
and that they involve not w/i-movement but a rule of controlled-pro-deletion.
They assume, for example, that the free relative in (1) has the underlying
structure represented in (2).
(1) I'll buy what you 're selling.
what you are selling PRO
They also discuss free relatives in certain other languages. They appear to be
suggesting that their analysis is universally applicable. They suggest, for
example, that the form of free relatives in Tok Pisin, an English-based pidgin
spoken in Papua New Guinea, "strikingly confirms our hypothesis" (1978:
377). It is hard to see how this could be relevant unless the analysis is assumed
to be universal. (Woolford 1978, which Bresnan and Grimshaw draw on, ex-
plicitly assumes that free relatives are essentially similar in all languages.) They
go on, however, to deny that the analysis is universally applicable by suggest-
ing that free relatives are headless in Moroccan Arabic. In this paper, I will
look in some detail at Polish free relatives. I will show that there is strong
evidence that they are headless relatives involving w/z-movement and thus that
Bresnan and Grimshaw's analysis is certainly not universally applicable. Then,
I will argue that it is not at all obvious that their analysis is the correct one
for English.
An important weakness of Bresnan and Grimshaw's discussion is their failure
to recognize that there are at least two distinct headless analyses that might
be proposed for free relatives. (This point is clearly recognized in Woolford
1978.) On the one hand, there is an analysis in which they are isomorphic
2 Robert D. Borsley
with embedded questions. On the other, there is an analysis on which they

are only superficially headless, involving a head that is deleted or that is
phonologically empty. On the former analysis, the free relative in (1) will
have the underlying structure in (3). On the latter, it will have that in (4).
(3)
COMP
(that) COMP
you are selling what
As I will illustrate below, Bresnan and Grimshaw's arguments are directed

primarily against an analysis of the former kind. It is an analysis of the latter
kind that I will suggest is appropriate for Polish and perhaps for English. It is
worth noting here that there is a second headed analysis that might be ad-
vanced. This is an analysis involving w/i-movement and a subsequent rule
moving wft-phrases from COMP to head position. Such an analysis is pro-
posed for French free relatives in Hirschbhler 1976. Some of the Polish facts
that argue against a Bresnan-Grimshaw analysis are compatible with this ana-
lysis. There are facts, however, that show that it too is untenable. 2
Before I proceed, I must make two further points. Firstly, I want to stress
that I will not be providing anything like a complete description of either
Polish or English free relatives. Both Polish and English free relatives involve
all sorts of complexities whose precise nature I have not investigated. Second-
ly, I must note that there is considerable variation in the judgements Polish
speakers give in this area. I am sure some speakers will disagree with some of
the judgements I give here. The most important judgements, however, seem
to be fairly generally accepted. I assume, then, that the arguments I develop
here are valid for most speakers.
We can now survey the main types of free relatives that occur in Polish.
Like English, Polish has free relatives introduced by a variety of w/i-words.
The following illustrate this.
(5) Kto umyt rgce, moze zaczqc jesc.
who washed hands can begin eat
'The one who has washed his hands can begin eating.'
(6) Co powiesz, b$dzie uzyte przeciwko tobie.

what (you) will say will be used against you
'What you say will be used against you.'
(7) Gdzie Jan mieszka, jest bardzo ciepto.
where John lives is very warm
'Where John lives is very warm.'
(8) Kiedy Piotr jest nami, jest nam bardzo przyjemnie.
when Peter is with us is us very pleasant
'It's very pleasant to us when Peter is with us.'
(9) Jak to zrobisz, na pewno b$dzie dobrze.
how it (you) will do certainly will be good
'The way you do it will certainly be good.'
There are also free relatives introduced by a w/i-word with the suffix -kolwiek.
These correspond to English free relatives introduced by a -word with the
suffix -ever.3
(10) Ktokolwiek zaspiewa, dostanie nagrod.
whoever will sing will get prize
'Whoever sings will get a prize,'
(11) Cokolwiek Jan chciat, by to nieosiqgalne.
whatever John wanted was unobtainable
'Whatever John wanted was unobtainable.'
(12) Gdziekolwiek Jerzy pracowal, na pewno duio zarabiat.
wherever George worked certainly much (he) earned
'Wherever George worked, he certainly earned a lot.'
(13) Kiedykolwiek spotkasz , pozdrow jq.
whenever (you) will meet Mary (you) greet her
'Whenever you meet Mary, say "hallo" to her.'
(14) Jakkolwiek to zrobisz, bgdzie dobrze,
however it (you) will do will be good
'However you do it, it will be all right.'
Finally, there are free relatives introduced by a wh-word and a following

nominal.
(15) Jakq prac$ Jan znalazi, zaraz musiat niej zrezygnowac.
which work John found immediately (he) had to from it
resign
4 Robert D. Borsley
'Whichever work John found, he had to give it up imme-

diately.'
(16) Jakie pieniqdze miatem, dalem Marii.
what money (I) had (I) gave Mary
'What money I had, I gave to Mary.'
(17) Jakqkolwiek ksiqzke przeczytasz, b$dziesz musial mi jq
opowiedziec.
whichever book (you) will read (you) will be have to (to)
me it tell
'Whichever book you read, you will have to tell me the
story.'
(18) Jakiekolwiek buty kupisz, rozlecq si.
whatever shoes (you) will buy (they) will fly themselves
'Whatever shoes you buy, they will fall apart.'
The above data suggest that Polish free relatives are quite like English free
relatives. In some respects, this is the case. There is strong evidence, however,
against a Bresnan-Grimshaw analysis. I will consider this shortly. First, how-
ever, I want to argue that Polish free relatives are not isomorphic with em-
bedded questions.
The simplest argument against such an analysis involves number agree-
ment. Bresnan and Grimshaw point out that when a free relative appears in
subject position, the form of the verb depends on the number of the wh-
phrase. If the w/z-phrase is plural, the plural form of the verb is necessary.
(19) illustrates this.
(19) Whatever books she has marked up with her notes.
This is not the case, however, with an embedded question in subject position,
as (20) illustrates.
(20)J What books she has f!sn t , 1 certain.

[*arentI
The situation is just the same in Polish. We have data like the following:
(21) Jakiekolwiek szeroko

czytane.
whatever books Mary will write will be widely read
'Whatever books Mary writes will be widely read.'
f/esf (sg.)| , .
(22) Jakie ksiqzki ma Maria j ^
I *sq (pi.) J Jemnicel'
what books has Mary is/are secret
'What books Mary has a secret.'
These contrasts are quite inexplicable if free relatives are isomorphic with
embedded questions. They are quite natural on a Bresnan-Grimshaw analysis.
They are also quite natural, however, on a superficially headless analysis. On
such an analysis, the wh -phrase will agree in number with the deleted or phono-
logically empty head, and the latter will determine the form of the verb.
We can turn now to the evidence against a Bresnan-Grimshaw analysis.
Some initial evidence comes from the fact that for every free relative intro-
duced by a simple -word, there is an identical demonstrative-headed rela-
tive clause. Thus, parallel to (5) - (9), we have (23) - (27).
(23) Ten, kto umyi r$ce, moze zaczqc jesc.

he who washed hands (he) can begin eat
'The one who has washed his hands can begin eating.'
(24) To, co powiesz, b^dzie uzyte przeciwko tobie.
it what (you) will say will be used against you
'What you say will be used against you.'
(25) Tam, gdzie Jan mieszka, jest bardzo cieplo.

there where John lives is very warm
'Where John lives is very warm.'
(26) Wtedy, kiedy Piotr jest nami, jest nam bardzo przyjemnie.
then when Peter is with us is (to) us very pleasant
'It's very pleasant to us when Peter is with us.'
(27) Tak, jak to zrobisz, na pewno bdzie dobrze.
so how it (you) will do certainly will be good
'The way you do it will certainly be good.'
If the w/z-words in Polish free relatives are heads, this fact will be quite
accidental. If we assume a superficially headless analysis, however, it will be
quite natural. Such an analysis will either involve a rule deleting a demon-
strative head under certain circumstances or will allow a phonologically empty
head under certain circumstances. In either case, it is only to be expected that
there will be many cases where free relatives are identical to demonstrative-
headed relative clauses.
6 Robert D. Borsley
Some further evidence against Bresnan and Grimshaw comes from the fact
that there is no independent evidence for a general rule of controlled-pro-
deletion. In English, relative clauses introduced by that provide independent
motivation for such a rule. There is good evidence that that is a complemen-
tizer when it introduces a relative clause. (See, for example, Morgan 1972.) It
is natural, therefore, to assume that relative clauses introduced by that in-
volve deletion.4 In Polish, the situation is rather different. Relative clauses
normally involve w/i-movement. There are, however, relative clauses that
appear to involve deletion. Consider, for example, (28).
(28) Jan widziai zlodzieja, co obrabowal bank.

John saw thief what robbed bank
'John saw the thief that robbed the bank.'
I have glossed co here as Vhat' because it occurs elsewhere as a neuter inter-
rogative pronoun. There is evidence, however, that it is a complementizer
here and hence that we have deletion of a subject pronoun. The evidence are
data like the following.
(29) Maria kocha mezczyzn, co go podziwiasz.
Mary loves man what him (you) admire
'Mary loves the man that you admire.'
(30) Jan widziai tegofaceta, co nim rozmawiales.
John saw the guy what with him (you) talked
'John saw the guy that you talked to.'
Like (28), these examples involve relative clauses introduced by co. In each
case, however, the relative clause contains a separate pronoun of some kind.
It seems clear, then, that co is not a -word but a complementizer in these
examples. It is natural, then, to assume that the same is true of examples
like (28) and thus that they involve deletion. The fact that (31) is ungram-
matical shows that this is an obligatory controlled deletion and not the ordi-
nary context-conditioned deletion of a subject pronoun that is a characteristic
feature of Polish:
(31) *Jan widziai zlodzieja, co on obrabowal bank.
The fact that the following are ungrammatical shows that this deletion is
restricted to subjects.
(32) *Maria kocha m$zczyzn$, co podziwiasz.
(33) *Jan zna tego faceta, co rozmawiales.
-movement is not restricted in this way, however, as the following illustrate:
(34) Maria kocha m$iczyzn$, ktorego ty podziwiasz.

Mary loves man which you admire
'Mary loves the man who you admire.'
(35) Jan widziai tego faceta, ktorym rozmawiates.
John saw the guy with which (you) talked
'John saw the guy who you talked to.'
It is natural, then, to assume that free relatives involve wh-movement and not
deletion]
Further evidence that we have wft-movement and not deletion here comes
from the ungrammaticality of the following examples.
(36) *Kto Jan powiedziai, ze kocha , fest gtupcem.
who John said that loves Mary is fool
'The one who John said loves Mary is a fool.'
(37) *Ktokolwiek powiesz, ze zna francuski, dostanie pracf.
whoever (you) will say that knows French (he) will get work
'Whoever you say knows French will get a job.'
(38) *Jakie ksiqzki myslalei ze dostalem na urodziny, sprze-
dalem sqsiadowi.
what books (you) thought that (I) got on birthday (I) sold
neighbour
'Whatever books you thought I got on my birthday I sold
to the neighbour.'
If Polish free relatives involve -movement, the ungrammaticality of these
examples is not surprising. All will involve the movement of a w/i-word into
a higher clause. This is generally impossible in Polish. The following illustrate.
(39) *Kto Jan powiedziai, ze moze przyjsc na wyktad?
who John said that may come on lecture
'Who did John say may come to the lecture?'
(40) *Ktorq ksiqzk? Marek myslal, ze mial przeczytac?
which book Mark thought that (he) had to read
'Which book did Mark think that he had to read?'
Thus, if Polish free relatives involve vv/j-movement, it is only natural that

(36) - (38) should be ungrammatical. If they involve deletion, however, the
ungrammaticality of these examples is rather problematic. This is because
Polish does have unbounded deletion. This is illustrated by sentences like the
following:
8 Robert D. Borsley
(41) Jan jest inteligentniejszy, niz myslelismy, te jest.

John is more intelligent than (we) thought that (he) is
'John is more intelligent than we thought he was.'
(42) Maria jest tak pigkna, jak Jan mowil, ze jest.5
Mary is so beautiful how John said that (she) is
'Maiy is as beautiful as John said that she was.'
Thus, there is no obvious reason why (36) (38) should be ungrammatical
if Polish free relatives involve deletion. Again, then, it is natural to assume
that we have -movement and not deletion in Polish free relatives.
Some more complex evidence against a Bresnan-Grimshaw analysis comes
from free relatives introduced by a w/j-word and a nominal. An important fact
about such free relatives is that the nominal need not directly follow the wh-
word. Thus parallel to (15) - (18), we have (43) - (46).
(43) Jakq Jan znalazi prac$, zaraz musial niej zrezygnowac.
which John found work immediately (he) had to from it
resign
'Whichever work John found, he had to give it up imme-
diately.'
(44) Jakie miatem pieniqdze, datem Marii.
what (I) had money (I) gave Mary
'What money I had, I gave to Mary.'
(45) Jakqkolwiek przeczytasz ksiqik, bqdziesz musiat mi jq
opowiedziec.
whichever (you) will read book (you) will be have to (to)
me it tell
'Whichever book you read, you will have to tell me the
story.'
(46) Jakiekolwiek kupisz buty, rozlecq si$.
whatever (you) will buy shoes (they) will fly themselves
'Whatever shoes you buy, they will fall apart.'
If we assume a Bresnan-Grimshaw analysis, the free relatives in (15) (18)
will have a wh-word and a nominal in head position, and the free relatives in
(43) (46) will involve the postposing of part of the head. It is generally im-
possible, however, to postpose part of the head of a relative clause. The follow-
ing illustrate.
(47) a. Te pieniqdze, ktore Jan znalazi wczoraj, datem Marii.
the money which John found yesterday (I) gave Mary.
gave the money which John found yesterday to Mary.'
b. *Te, ktore Jan znalazi wczoraj, pieniqdze dalem Marii.
(48) a. Ten student, ktory zdat wszystkie egzaminy, wyjezdza

do Australii.
the student which passed all exams is going to Australia
'The student who has passed all the exams is going to
Australia.'
b. *Ten, ktory zdat wszystkie egzaminy, student wyjezdza

do Australii.
Thus, the free relatives in (43) - (46) pose serious problems for a Bresnan-
Grimshaw analysis. Such free relatives pose no problems, however, for a head-
less analysis. A central fact about wh-movement in Polish is that it can violate
Ross's (1967) left branch condition. The following illustrate.
(49) Ktore Jan namalowat obrazy?

which John painted paintings
'Which paintings did John paint?'
(50) Czyj chciatbys przeczytac artykul?
whose (you) would like read article
'Whose article would you like to read?'
Thus, if Polish free relatives are headless relatives involving w/z-movement, the
free relatives in (43) - (46) are only to be expected.
Although the facts surveyed so far provide evidence against a Bresnan-
Grimshaw analysis, they are compatible with an analysis of the kind that
Hirschbiihler proposes. There are facts, however, that argue against both kinds
of analysis.
Some important evidence comes from the fact that there are no cases else-
where in Polish of embedded finite clauses not introduced by some element
in COMP. Ordinary relative clauses are always introduced by some element in
COMP (usually a wh-word). Thus, only the a. examples in the following are
grammatical.
(51) a. Ten cztowiek, ktorego Maria kocha, jest gtupi.
the man who mary loves is foolish
'The man who Mary loves is a fool.'
b. *Ten cztowiek, Maria kocha, jest gtupi.
(52) a. Ksiqzki, ktore mi dates, sq bardzo ciekawe.
books which me (you) gave are very intersting
'The books which you gave me are very interesting.'
b. *Ksiqzki, mi dates, sq bardzo ciekawe.
10 Robert D. Borsley
Similarly, finite complements are always introduced by a complementizer.

Again, only the a. examples in the following are grammatical.
(53) a. Jan mowit, ze Maria wyjechata do Londynu.
John said that Mary went to London
'John said that Mary had gone to London.'
b. *Jan mowit, Maria wyjechata do Londynu.
(54) a. Jan chcial, zeby Maria przyszla.
John wanted that Mary came
'John wanted Mary to come.'
b. *Jan chcial, Maria przyszla.
It is natural in this situation to assume that all embedded finite clauses in
Polish must be introduced by some element in COMP. Clearly, however, this
is impossible if the w/z-words in free relatives are heads. It is natural, then, to
assume that they are ordinary w/i-words in COMP and thus that free relatives
are headless.
Some further evidence for this position comes from what can be termed
the mobile inflection phenomenon. This is the ability of the person and
number inflections of certain verbs to appear separated from the verb stems
with which they are normally associated and attached to certain other lexical
items. The following illustrate this phenomenon.
(55) a. Ty czytales t ksiqikf.
you read this book
'You read this book.'
b. Tys czytal t ksiqzk.
(56) a. Wykladu wysluchales.
lecture (you) listened
'You listened to the lecture.'
b. Wykiadus wysluchai.
(57) a. Kiedy widziales ten film?
when (you) saw this film
'When did you see this film?'
b. Kiedys widzial ten film?
(57) b. shows that mobile inflections can be attached to items in COMP. It
seems, however, that they cannot appear outside the S in which the verb stem
with which they are normally associated appears. In particular, when a verb
stem with which a mobile inflection is normally associated appears in a rela-
tive clause, the inflection cannot be attached to the head of the clause. The
following illustrate.
(58) a. M^zczyzna, ktorego widziaies, jest bratem Manx.

man which (you) saw is brother Mary
'The man who you saw was Mary's brother.'
b. *Mzczyznas, ktorego widziat, jest bratem Marii.
Notice now that mobile inflections can be attached to the wh-words in free
relatives. Consider, for example, the following.
(59) a. Co powiedziates, zostato nagrane.
what (you) said became recorded
'What you said was recorded.'
(60) a. Cokolwiek tu slyszales, musisz tym zapomniec.

whatever here (you) heard (you) must about it forget
'Whateveryou have heard here,you must forget about it.'
b. Coskolwiek tu siyszai, musisz tym zapomniec.
(61) a. Jakie piwo zamowiles, takie przyrtiosq.
what beer (you) ordered such (they) will bring
'Whatever beer you ordered, they will bring.'
b. Jakies piwo zamowil, takie przyniosq.
Here, then, we have quite strong evidence that the -words in free relatives
are in COMP and are not heads.
Some further evidence comes from the distribution of the possessive re-
flexive swoj. Like ordinary reflexives, swoj must be a clause mate of its ante-
cedent. This means, among other things, that swoj cannot appear in COMP
with an antecedent in a higher clause. (62) illustrates.
(62) *Jarii zapytai, ktorq ze swoichj piosenek lubisz.
John asked which from his songs (you) like
'John asked which of his songs you liked.'
To make this sentence grammatical, it is necessary to replace swoich with the
non-reflexive form jego 'his'. Swoj can appear, however, inside the head of a
relative clause with an antecedent outside. (63) illustrates this.
(63) J an zaspiewa kazdq ze swoichj piosenek, jakq wybierzesz.
John will sing each from his songs which (you) will choose
'John will sing each of his songs that you choose.'
Notice now that (64) is ungrammatical.

(64) *Janj zaspiewa ktorqkolwiek ze swoichj piosenek wy-
bierzesz.
John will sing whichever from his songs (you) will choose
'John will sing whichever of his songs you choose.'
Here, we have a free relative with swoj within the w/i-phrase and the main
clause subject as the antecedent. The fact that such sentences are ungram-
matical suggests strongly that the wh-phrases in free relatives are in COMP and
are not heads.
The distribution of swoj in fact provides another argument against a Bresnan-
Grimshaw analysis. Both (62) and (64) are grammatical if the deleted com-
plement subject is the antecedent of swoich. On either a Bresnan-Grimshaw
analysis or a Hirschbhler analysis, (64) will on this interpretation involve a
form of swoj with an antecedent in a lower clause. In general, this is impos-
sible. (65) illustrates.
(65) *Jan zaspiewa kazdq ze swoichf piosenek, jakq tyi wy-
bierzesz.
John will sing each from your songs which you will choose
'John will sing each of your songs that you choose.'
This is no problem for a Hirschbhler analysis since on such an analysis swoich

will be a clause mate of the deleted subject prior to the raising of the wh-
phrase into head position. It provides an important argument,however, against
a Bresnan-Grimshaw analysis.
A final argument against both a Bresnan-Grimshaw analysis and a Hirsch-
bhler analysis involves the phenomenon of pied piping. On either analysis,
one would not expect pied piped prepositions before the w/z-words in free
relatives. Free relatives with a pied piped preposition before the -word are
often ungrammatical. There are grammatical examples, however. The follow-
ing illustrate.
(66) kimkolwiek Jan rozmawia, Maria go lubi.

with whoever John talks Mary him likes
'Whoever John talks to, Mary likes him.'
(67) kimkolwiek porozmawiasz, zrozumie ciq.
with whoever (you) will talk (he) will understand you
'Whoever you talk to will understand you.'
Clearly, such examples provide evidence against both analyses. On a super-
ficially headless analysis, however, they are only to be expected.
I have now presented nine arguments against a Bresnan-Grimshaw analysis
of Polish free relatives. With five of the arguments, the key facts are com-
patible with a headed analysis of the kind that Hirschbhler (1976) proposes.
This is not the case, however, with the other four arguments. With all the
arguments, the key facts are exactly what a superficially headless analysis
leads one to expect. They are also compatible with an analysis in which free
relatives are isomorphic with embedded questions. We saw earlier, however,

that there is evidence against such an analysis. It seems clear, then, that a
superficially headless analysis is appropriate for Polish free relatives.
We can return now to English free relatives. I noted at the outset that
Bresnan and Grimshaw's arguments are directed primarily against an analysis
of free relatives in which they are isomorphic with embedded questions. I
subsequently pointed out that the number agreement, facts to which they
draw attention argue against such an analysis but are quite compatible with a
superficially headless analysis. I want now to show that the same is true of
certain other facts that Bresnan and Grimshaw discuss.
An important feature of free relatives which Bresnan and Grimshaw point
out is what they term the "matching effect", the fact that the wh-phrase in a
free relative is of the same category as the whole free relative. In this, free
relatives are quite unlike embedded questions where the category of the wh-
phrase is completely independent of the category of the whole construction.
Clearly, then, the matching effect argues against an analysis of free relatives
in which they are isomorphic with embedded questions. For a Bresnan-Grim-
shaw analysis, the matching effect is no problem. As Bresnan and Grimshaw
point out, it is an automatic consequence of the following, well motivated
constraint.
(68) A phrase and its head have the same categorial specification.
There is no problem either, however, for a superficially headless analysis. In
such an analysis, the matching effect follows from (68) and the following,
rather natural constraint.
(69) The head of a relative clause and the wA-phrase have the
same categorial specification.
Thus, the matching effect argues against an analysis of free relatives in which
they are isomorphic with embedded questions but is quite compatible with a
superficially headless analysis.
The situation is similar with certain facts relating to what Bresnan and
Grimshaw term the Internal NP-over-S Constraint. This constraint, which has
its origins in the work of Ross (1967) and Kuno (1973), stipulates that sen-
tences exhaustively dominated by NP cannot appear internal to phrases of
which they are immediate constituents. It accounts for the dubious character
of sentences like the following.
(70) ?Can whether you are right or not matter?

(71) ?I found what the consequences were unclear.
These sentences contain embedded questions. Similar sentences containing

free relatives are perfectly acceptable. The following illustrate.
(72) Can what you want be on the table?

(73) I found what she cooked delicious.
Clearly, this contrast argues that free relatives are not isomorphic with em-
bedded questions. It poses no real problems, however, for a superficially
headless analysis. If the constraint is a surface filter, it will provide evidence
against a superficially headless analysis involving a deleted head. It will pose
no problems, however, for a superficially headless analysis involving a phono-
logically empty head. Again, then, the facts Bresnan and Grimshaw cite pro-
vide evidence against an analysis of free relatives in which they are isomorphic
with embedded questions, but they are quite compatible with a superficially
headless analysis.
It is clear, then, that three of Bresnan and Grimshaw's arguments only pro-
vide evidence against an analysis of free relatives in which they are isomorphic
with embedded questions. There are three other arguments that might seem
to provide some evidence against a superficially headless analysis. I will show,
however, that doubts can be raised about all of them.
The first argument involves certain wh -constructions where it is fairly clear
that the -words are not ordinary wh-words in COMP. The following illu-
strate.
(74) She wrote whenever possible.
(75) Eat what you please.
Bresnan and Grimshaw show that it is not plausible to derive the w/i-construc-
tions here from ordinary free relatives and hence that the wh -words cannot be
ordinary wh-words in COMP. They seem then to suggest that considerations
of simplicity support the view that the wh -words in free relatives are not
ordinary wh -words in COMP either.
Polish provides an objection to this argument. In Polish, -words with the
suffix -kolwiek occur quite widely in contexts where they are clearly not
ordinary -words in COMP. Roughly, they occur in all the contexts in which
English any forms occur except negative environments, where negated forms
appear. The following illustrate.
(76) Czy widziales tam kogokolwiek?

Q (you) saw there anyone
'Did you see anyone there?'
(77) Ktokolwiek moze tarn wejsc.

anyone can there enter
'Anyone can go in there.'
(78) Daj to komukolwiek.
give it anyone
'Give it to anyone.'
(79) Jezeli kiedykolwiek bgdziesz w Poznaniu, zadzwon do mnie.

if any time (you) will be in Poznaii, ring to me
'If you are in Poznari any time, give me a ring.'
It is clear, then, that the appearance of w/j-words in contexts where they can-
not be ordinary w/i-words in COMP does not in any way necessitate the con-
clusion that they are not ordinary wft-words in COMP in free relatives.
The second argument involves pied piping. Unlike Polish free relatives,
English free relatives never involve pied piped prepositions. A superficially
headless analysis leads one to expect pied piped prepositions in free relatives.
It looks, then, as if we have a real argument against such an analysis here. It
is not clear that this is the case, however. As we have seen, free relatives with
pied piped prepositions are sometimes acceptable in Polish. Often, however,
they are ungrammatical.None of the following,for example, are grammatical. 6
(80) *Z kimkolwiek Jan rozmawia, Maria lubi.

with whoever John talks Mary likes
'Whoever John talks to, Mary likes.'
(81) *Jan spotkat kim rozmawiales.
John met with whom (you) talked
'John met the person you talked to.'
(82) *Maria rozmawiata kim Jan ciqgle mowi.
Mary talked with about whom John continuously talks.
'Mary talked to the person John talks about all the time.'
It is clear, then, that a superficially headless analysis is quite compatible with

the ungrammaticality of many free relatives with pied piped prepositions. It
is not at all obvious, then, that such an analysis is incompatible with the un-
grammaticality of all such free relatives.
The final argument that might seem to provide some evidence against a
superficially headless analysis is not fully spelled out. Its form, however, is
reasonably clear. It involves pairs of sentences like the following:
(83) I'll live in whatever town you live in.

(84) I'll live in whatever town you live.
For Bresnan and Grimshaw, (83) involves an NP-headed relative clause in a PP
and deletion of a pro-NP, while (84) involves a PP-headed relative clause and
deletion of a pro-PP. If one assumes a superficially headless analysis, examples
like (83) are quite straightforward. Examples like (84), however, will presum-
ably require a rule deleting a pied piped preposition under identity with a
higher preposition. Such a rule seems rather ad hoc. Thus, the existence of
sentences like (84) might seem to provide evidence against a superficially
headless analysis. Again, however, Polish provides an objection. Sentences like
(84) are quite common in Polish. The following illustrate.
(85) Zamieszkam w jakimkolwiek kraju ty zamieszkasz.
(I) will live in whatever country you will live
will live in whatever country you will live.'
(86) Pojad do jakiegokolwiek miasta ty pojedziesz.
(I) will go to whatever town you will go
will go to whatever town you go.'
It is clear, then, that a superficially headless analysis is quite compatible with
the existence of sentences like (84).
I have now outlined the main arguments advanced by Bresnan and Grim-
shaw for their analysis of English free relatives. I have shown that three of
their arguments only provide evidence against an analysis of free relatives in
which they are isomorphic with embedded questions and that Polish facts
cast doubt on their other three arguments which might seem to provide some
evidence against a superficially headless analysis. I have not shown that their
analysis is untenable. I think I have shown, however, that their arguments are
much less compelling than they initially appear and thus that it is not clear
that their analysis is preferable to a superficially headless analysis.
We have seen, then, that there is strong evidence that Polish free relatives
are headless relatives and thus that Bresnan and Grimshaw's analysis is not
universally applicable. We have also seen that their arguments about English
free relatives are rather weak and hence that it is not at all obvious that their
analysis is the correct one for English.
Notes
1. I am grateful to Ewa Jaworska for placing her Polish intuitions at my disposal and
providing much other assistance, and to Frank Heny, George Horn, Dick Hudson and
Neil Smith for a number of helpful comments.
2. A further headed analysis is advanced in Jackendoff 1977 (chapter 9). This involves
a rule moving a wA-word from S directly to head position. The facts that argue against
a Hirschbhler analysis also argue against an analysis of this kind.
3. Another interesting class of free relatives involve an idiomatic use of the negative
particle nie. Free relatives of this kind seem to be similar to free relatives involving
kolwiek. The following illustrate.
(i) Co nie zrobi$, zawsze jest tie.
what not (I) do always is bad
'Whatever I do is always wrong.'
(ii) Kogo nie zapytasz, wskaze ci drogq.
who not (you) will ask (he) will show you way
'Whoever you ask will show you the way.'
4. Chomsky, of course, would assume that such relative clauses involve wA-movement
and deletion in COMP. I would reject this view. For criticisms of Chomsky's
views on wA-movement and deletion, see Bresnan (1976), Grimshaw (1974), Allen
(1980) and Borsley (1981).
5. I have glossed jak here as 'how' because it appears elsewhere as an adverbial wA-word.
There is strong evidence, however, that it is not a wA-word here and that such exam-
ples involve deletion. See Borsley (1981).
6. A plausible suggestion, I think, is that pied piped prepositions are only possible in
free relatives that have an adverbial function. It seems reasonable to say that the free
relative in (66) has an adverbial function. The free relative in (67) appears to be a
subject. It is possible, however, that the sentence involves a deleted subject pronoun
and thus that the free relative has the same role as in (66). It is interesting to note in
this context that Old English had a distinctive class of free relatives used in an ad'
verbial role. See Allen (1980).
References
Allen, C.
1980 "Movement and deletion in Old English", Linguistic inquiry 11:261-323.
Borsley, R. D.
1981 "WA-movement and unbounded deletion in Polish equatives", Journal of
linguistics 17:270-288.
Bresnan, J. W.
1976 "Evidence for a theory of unbounded transformations", Linguistic analysis
2:353-393.
Bresnan, J. W. - J. Grimshaw
1978 "The syntax of free relatives in English", Linguistic inquiry 9:331-391.
Farkas, D. - W. M. Jacobsen - K. W. Todrys (eds)
1978 Papers from the Fourteenth Regional Meeting of the Chicago Linguistic
Society (Chicago, Illinois: University of Chicago).
Ford, A. - J. Reighard - R. Singh (eds)
1976 Papers from the Sixth Annual Meeting of the North Eastern Linguistic Soci-
ety (Montreal: McGill University).
Grimshaw, J.
1974 "Evidence for relativization by deletion in Chaucerian Middle English", in:
Kaisse - Hankamer (eds) 1974:216-224.
Hirschbiihler, P.
1976 "Two analyses of free relatives in French", in: Ford et al. (eds) 1976.
Jackendoff, R. S.
1977 X syntax: A study of phrase structure (= Linguistic inquiry monograph, 2)
Cambridge, Mass.: M.I.T.Press).
Kaisse, E. - J. Hankamer (eds)
1974 Papers from the Fifth Annual Meeting of the North Eastern Linguistic Soci-
ety (Cambridge, Mass.: Harvard University).
Kuno, S.
1973 "Constraints on internal clauses and sentential subjects", Linguistic inquiry
4:363-385.
Morgan, J. L.
1972 "Some aspects of relative clauses in English and Albanian", in: Peranteau
et al. (eds) 1 9 7 2 : 6 3 - 7 2 .
Peranteau, P. M. - J. Levi - G. Phares (eds)
1972 The Chicago which hunt (Chicago, Illinois: Chicago Linguistic Society).
Ross, J. R.
1967 Constraints on variables in syntax (Ph. D. dissertation, M.I.T.).
Woolford, E.
1978 "Free relatives and other base generated WH constructions", in: Farkas et al.
(eds) 1978:482-490.
STEFAN DYLA
A note on inversion/conjunct ascension

constructions in Polish and English1
1 Introduction
In this paper, I will attempt to give a tentative account of the syntax of a

number of complex inversion constructions in Polish and English within the
theory of arc pair grammar (henceforth: APG) as advanced by Johnson and
Postal (1980). More specifically, I will deal here with those types of inversion
constructions which are also conjunct ascension (henceforth: CA) construc-
tions. While doing so, I will assume the reader's familiarity with basic claims
and hypotheses concerning the nature of language made in APG, like e.g.
the claim that sentences of natural languages are reconstructible in terms of
formal objects called pair networks (PNs).
1.1 Inversion and conjunct ascension in arc pair grammar

Prior to any discussion, let me say a few words about inversion and conjunct
ascension themselves. In relational grammar, these two terms have been used
to refer to two relation-changing rules of universal grammar: Inversion was a
demotion rule sanctioning the 3-hood of a former 1, while conjunct ascension
was a raising-type rule sanctioning the termhood of a former conjunct. 2
In the theory of arc pair grammar, there is nothing like the notion of a
universal rule. Rather, there are universal constructions, inversion construc-
tions and CA constructions among them, whose well-formedness at the level
of universal grammar is determined by the set of APG PN laws, which are
interpreted as material implications (in the standard logical sense) checking
for the co-occurrence or the non-co-occurrence of specified properties of PNs
in all languages. A construction is well-formed at the level of a grammar of
some particular language if the grammar does not contain a rule which would
block such a construction. 3
Put into APG terms, inversion is a construction involving a cjth stratum 1
arc locally sponsoring a cj+ j th stratum 3 arc, whereas conjunct ascension is a
20 Stefan Dyia
construction involving an immigrant arc whose foreign predecessor is a Con

arc and whose local sponsor is a Term arc which is a support of the Con arc.
Diagrammatically,
(1) Figure 1
(2) Figure 2
Inversion/conjunct ascension in Polish and English 21
1.2 Evidence for inversion and conjunct ascension in Polish and English
1.2.1 Evidence for inversion in Polish

To begin with, consider evidence for the well-formedness of inversion con-
structions in Polish. Polish exhibits a great number of constructions where
nominals marked with the dative case, a typical flagging used to mark heads
of final 3 arcs in Polish, behave like heads of 1 arcs with respect to some tests
for 1-hood, viz. those tests which apply not only to heads of final 1 arcs but
also to heads of non-final 1 arcs. Consider the following doublet: 4
(3) Przypomnialo mi sif, ze . . .
(remembered/IMPERS me/DAT REFL that)
remembered (willy-nilly) t h a t . . . '
(4) Po przyjsciu do domu przypomniaio mi si?, ze . . .

(after having-come to home/GEN remembered/IMPERS
me/DAT REFL that)
'Having returned home, I remembered (willy-nilly) that..
The sentence in (4) involves what might be called equi into action nomi-
nalization functioning as a temporal adverbial. In APG, it is reconstructed as
a PN of the form shown in Figure 3.
In other words, (4) involves foreign erasure of a final 1 arc whose support
is a Temporal arc by a 3 arc which is local successor of a 1 arc and a neighbor
of the Temporal arc in question.
Also note that final 1 arcs and Cho arc local successors of 1 arcs can un-
hook final 1 arcs in such constructions.
(6) Po przyjsciu do domu przypomnialem sobie, ze . . .

(after having-come to home/GEN I-remembered REFL
that)
'Having returned home, I remembered t h a t . . . '
(7) Te klucze zostaiy znalezione przez Olg? podczas sprz^tania
twojego pokojui
(these keys/NOM got found by ACC while cleaning your
room/GEN)
'These keys have been found by Olga when she was cleaning
your room.'
Surprisingly, 3 arcs which are not local successors of 1 arcs cannot do the
unhooking, as is evidenced by the ill-formedness of the following sentence
under a reading where it was me who returned home.
22 Stefan Dyta
(8) *Po przyjsciu do domu mama data mi kolacjf.

(after having-come to home/GEN mother/NOM gave
me/DAT supper/ACC)
'Having returned home, Mother fed me supper.'
Summing up, the ability of the dative-marked nominals in question to con-
trol this variety of equi can be accounted for only if these nominals are heads
of final 3 arcs which are local successors of 1 arcs.
1.2.2 Evidence for inversion in English

Consider the following sentence:
(9) That book is too difficult for John.
When embedded in a super-equi context, it may take the form of either
(10) or (11).
(10) John, admitted that that book turned out to be too diffi-
cult for hinij.
(11) John admitted that that book turned out to be too difficult.
Of the two forms, the one which will concern us here is that in (11). What
seems to happen in (11) is that a final 3 arc is foreign erased by a final 1 arc
which is its remote support. Interestingly, final 3 arcs which are not local pre-
decessors of 1 arcs cannot be erased, as is shown in (12).
(12) a. John { said that it was quite clear that the letter had been
sent to hinii by Fritz.
b* John said that it was quite clear that the letter had been
sent by Fritz.
Note that this type of foreign erasure can be encountered in sentences like
the following one:
(13) Making a fool of herself in public amused Rita.

Again, the eraser is a remote support of the erased arc, which is a final
1 arc.
Given the ability of the for-phrase in (9) to be a super-equi victim and the
fact that heads of final 3 arcs which are not local successors of 1 arcs cannot
be victims of super-equi, one has to conclude that the/or-phrase in (9) is the
head of a final 3 arc which has a 1 arc local predecessor.
1.2.3 Evidence for conjunct ascension in Polish

Consider the following pair of sentences:
(14) Marek i Jola poszli do kina.
(NOM and NOM went/Pi to cinema/GEN)
'Marek and Jola went to the movies.'
(15) Marek poszedl J0I3 do kina.
(NOM went/Sg. with INSTR to cinema/GEN)
'Marek went to the movies with Jola.'
24 Stefan Dyla
One might claim that (15) involves conjunct ascension because it is truth-
conditionally equivalent to (14), in which the conjunction Marek iJola is the
head of a 1 arc at both the initial and final stratum. This, however, is a very
weak argument. A much more telling one is provided by the fact that the
following sentence, which is also truth-conditionally equivalent to (14), is
perfectly well-formed.
(16) Marek J0I3 poszli do kina.

(NOM with INSTR went/Pi. to cinema/GEN)
'Marek and Jola went to the movies.'
Despite the fact that it is Marek and not Marek Jolq which is the head of
a final 1 arc, the verb appears in plural. This means that what we have to do
here is an instance of agreement triggered by an initial 1 arc. If Marek Jolq
were not a conjunction at any stratum, then the plural agreement on the verb
in (16) would be unaccountable. The conclusion follows that sentences like
(15) and (16) involve an immigrant 1 arc whose foreign predecessor is a Con
arc and whose 1 arc local sponsor is overrun.
1.2.4 Evidence for conjunct ascension in English

Evidence for the well-formedness of CA constructions in English is harder to
come by. The existence of contrasting doublets like those in (17-18) by it-
self is a weak argument for conjunct ascension in English.
(17) John and Mary went to the movies.

(18) John went to the movies with Mary.
Nonetheless, there seems to exist a slightly stronger piece of evidence to

this effect. In relational grammar, it has been assumed that central gram-
matical relations are predicatable from semantic relations which nominals
bear in their clauses. This assumption has been taken over, though inexplicitly,
by arc pair grammar. And so, for instance, Agents are assumed to be always
heads of initial 1 arcs.
Note that the phrase with Mary in (18) is in the same semantic relation to
the clause as the nominal John Given the above assumption, both with Mary
and John should be heads of initial 1 arcs. Since the Stratal Uniqueness
Theorem does not allow, inter alia, two distinct 1 arcs at the same stratum, it
follows then that with Mary and John are heads of Con arcs which are branches
of the initial 1 arc of the clause.
2.0 Other constructions involved

In what follows, the following additional types of constructions will also be
referenced: 21 advancement, 32 advancement, beheading, genitive ascen-
sion, and antipassive. All of these types of constructions are well-formed in
Polish, whereas English exhibits all but antipassive. Subnetworks correspond-
ing to each of these constructions are shown in figures 4 through 8.
(19) Figure 4
(20) Figure 5
Stefan Dyia
(22) Figure 7
(23) Figure 8
Below, I will attempt to provide evidence for the well-formedness of each

of the above types of constructions in Polish and English. The existence or
nonexistence of antipassive constructions in English need not concern us
here as no type of English inversion-conjunct ascension construction involves
antipassivization.
2.1 Evidence for 21 advancement

Evidence for the well-formedness of this type of construction hinges upon the
assumption, made in relational grammar and adopted in arc pair grammar,
that intransitive strata can "contain" either a 1 arc not contrasting with a
2 arc or a 2 arc not contrasting with a 1 arc. The former type of stratum is
referred to as an unergative stratum, the latter as an unaccusative stratum.
The Final 1 Arc Law predicts that only non-final strata can be unaccusative.
Moreover, there cannot be two successive unaccusative strata, which is in-
sured by the Unaccusative Instability Law and the Unaccusative Law.
Underlying the distinction between unergative and unaccusative strata is
the fact that certain intransitive clauses are ambiguous with respect to the
volitional vs. non-volitional character of the action expressed. Consider the fol-
lowing pairs of sentences:
(24) a. John slid (on ice), (volitional)
b. John slid (while going downstairs and broke a leg), (non-
volitional)
28 Stefan Dyta
(25) a. Janek slizgal sif (na zamarzniftym stawie).

(NOM was-sliding REFL on frozen pond/LOC)
'Janek was sliding on a frozen pond.' (volitional)
b. Janek slizgai sif (id^c po oblodzonej drodze).
(NOM was-sliding REFL going on ice-covered road/
LOC)
'Janek was sliding while walking an ice-covered road.'
(non-volitional)
In APG, the (a)-forms of (2425) are analyzed as involving unergative ini-
tial strata, while the (b)-forms as involving unaccusative initial strata. This
analysis correlates nicely with the fact that John in(24.a.) and Janek in(25.a.)
are both agents, while John in (24.b.) and Janek in (25.b.) are both patients.
Given that patients are invariably heads of initial 2 arcs, the fact that John in
(24.b.) and Janek in (25.b.) are both heads of final 1 arcs can be accounted
for only if the initial 2 arcs in (24.b.) and (25.b.) local sponsor 1 arcs, which
happen to be final arcs.
2.2 Evidence for 3 - 2 advancement

Consider the following sentences:
(26) Jack gave Mary a bracelet.
(27) Jack showed Mary a picture of Rembrandt.
(28) Ten facet okradi Tomka pienifdzy i zegarka.
(this guy/NOM stole ACC of money/GEN PI.
and watch/GEN)
'This guy stole Tomek his money and wrist watch.'
(29) Ten facet obrabowat Tomka pienifdzy.
(this guy/NOM robbed ACC of money/GEN PI.)
'This guy robbed Tomek of his money.'
Note that the nominale Mary in (26-27) and Tomka in (28-29) have the
same semantic roles as to Mary and Tomkowi in (3031) and (3233), re-
spectively.
(30) Jack gave a bracelet to Mary.
(31) Jack has shown this picture to Mary.
(32) Ten facet skradi Tomkowi pieni^dze.
(this guy/NOM stole DAT money/ACC PI.)
'This guy has stolen Tomek his money.'
(33) Ten facet zrabowat Tomkowi pieni^dze.

(this guy/NOM robbed DAT money/ACC PI.)
"This guy has robbed Tomek of his money.'
More specifically, Mary in ( 2 6 - 2 7 ) and to Mary in (3031) are both reci-
pients, while Tomka in ( 2 8 - 2 9 ) and Tomkowi in ( 3 2 - 3 3 ) are both sources.
If one assumes that recipients and sources are always heads of initial 3 arcs,
then the 2-hood of the final arcs headed by Mary in ( 2 6 - 2 7 ) and Tomka in
(2829) follows from the fact that these arcs have 3 arc local predecessors.
2.3 Evidence for beheading 5
Beheading constructions are characterized as involving foreign erasure of an

arc whose neighbor is a Gen arc by an arc whose neighbor is another Gen arc,
where the first two neighbors are branches of an arc which is a neighbor of
an arc whose branches are the other two neighbors. This type of foreign era-
sure is to be found in, e.g., comparatives.
(34) John's apartment is bigger than Tom's.
(35) Samochd Janka jest szybszy niz Tomka.
(car/NOM GEN is faster than GEN)
'Janek's car is faster than Tomek's.'
That the above examples are instances of beheading constructions can be

seen in the fact that the standard of comparison in the above sentences is not
a full-fledged nominal but a genitive modifier, which is the head of a Gen arc
at both the initial and final stratum. Compare (3435) with the following
sentences where the standard of comparison takes the form of a full-fledged
nominal.
(36) John's apartment is bigger than Tom's apartment.

(37) Samochd Janka jest szybszy niz samochd Tomka.
(car/NOM GEN is faster than car/NOM Gen)
'Janek's car is faster than Tomek's car.'
Note also that the head of a Gen arc in (3435) is semantically possessor.
Since possessors are semantically paired with possessions, it will not be
counterintuitive to claim that every self-sponsoring Gen arc must have a
neighbor at the initial stratum which is not a Gen arc. Thus, one may con-
clude that the Gen arcs headed by Tom's and Tomka in (34) and (35), re-
spectively, have neighbors at the initial stratum which are foreign erased by
final arcs headed by apartment and samochod, respectively.
30 Stefan Dyia
2.4 Evidence for genitive ascension 6

Consider the following pairs of sentences:
(38) a. John's arrogance irritates me.
b. John irritates me with his arrogance.
(39) a. Zachowanie Janka bawi mnie.

(behavior/NOM GEN amuses me/ACC)
'Janek's behavior amuses me.'
b. Janek bawi mnie swym zachowaniem.
(NOM amuses me/ACC REFL behavior/INSTR)
'Janek amuses me with his behavior.'
Note that the relation between John and arrogance in (38.b.) and Janek
and zachowaniem in (39.b.) is exactly the same as that between John's and
arrogance in (38.a.) and Janka and zachowanie in (39.a.). This is a possessor-
possessed relation. Support for this claim comes from the ill-formedness of
(4041).
(40)* John irritates me with Sally's arrogance.
(41)* Janek bawi mnie zachowaniem Tomka.
The ill-formedness of (4041) follows from the fact that John in (38.b.)
and Janek in (39.b.) are not heads of 1 arcs at the initial stratum but rather
heads of Gen arcs. In order to account for the 1-hood of the final arcs headed
by these nominals, one must assume that both sentences involve 1 arc immi-
grants whose foreign predecessors are Gen arcs and whose local sponsors are
1 arcs which are supports of these Gen arcs. Thus, genitive ascension con-
structions are well-formed in both English and Polish.
2.5 Evidence for antipassive in Polish

Evidence for the well-formedness of antipassive constructions in Polish is very
similar to that of Postal 1977 for French. 7 To begin with, let us consider the
following examples of action nominalizations in Polish:
(42) bieganie Janka
(running GEN)
'Janek's running'
(43) czytanie ksi^zki

(reading book/GEN)
'the reading of a book'
(44) istnienie Boga

(existence God/GEN)
'the existence of God'
It follows from the above data that the genitive case marking is assigned to
heads of final absolutive arcs in action nominalizations. Surprisingly, what
seems to be the head of a final ergative arc in the examples below is likewise
marked with the genitive case.
(45) ogolenie sif Tomka
(having-shaved REFL GEN)
'Tomek's having shaved himself
(46) palenie Marka

(smoking GEN)
'Marek's smoking'
(47) poslizgni?cie si? Ewy
(having-slid REFL GEN)
'Ewa's having slid'
The fact that the final 1 arcs of (4547) have their heads marked with the
genitive strongly suggests that these arcs are absolutive rather than ergative.
To be able to give any systematic account of this fact, one must assume that
(4547) are instances of antipassive constructions and that therefore they all
involve a 1 arc locally sponsoring a 2 arc which overruns an initial 2 arc and
locally sponsors a 1 arc, which turns out to be a final arc.
3.0 Review of inversion/CA constructions in Polish and English

This section is devoted to a presentation of a number of inversion construc-
tions which are also conjunct ascension constructions. Each of these con-
structions will be illustrated with an example and accompanied with a partial
list of predicates which can occur in a given construction.
3.1 Type I
Consider the following sentences:
(48) Sue is similar to Pat.
(49) Rysiek jest podobny do Leszka.
(NOM is similar to GEN)
'Rysiek is similar to Leszek.'
32 Stefan Dyia
This type of construction is a combination of 21 advancement, inversion

and conjunct ascension. What happens in (4849) is that the initial 2 arc
locally sponsors a 1 arc which, in turn, is a local sponsor of the final 3 arc and
the final 1 arc, which is an immigrant arc whose foreign predecessor is a Con
arc. Diagrammatically,
(50) Figure 9
Other predicates which can occur in this type of construction:

a. English: equal, equivalent, identical, related, square, perpendicular, parallel,
distinct, different, to differ, etc.,
b. Polish: rowny 'equal', prostopadty 'square, perpendicular', rownolegly
'parallel', zbiezny 'converging', skosny 'oblique', zblizony 'ap-
proximate', etc.
3.2 Type II
Consider the following English sentence:
(51) The street parallels the railway.
It appears that this type of construction is almost identical to Type I except
that it additionally involves a 2 arc local successor of a 3 arc, which defines it
as a 32 advancement construction. Thus, it is reconstructive in terms of
APG PNs as a PN of the form shown in Figure 10.
(52) Figure 10.

a. English: to equal
b. Polish: the construction is unattested.
3.3 Type III

Consider the following sentences:8
(53) Harry's weight is equal to Sam's.
(54) Waga Janka jest rowna Tomkowej.
(weight/NOM GEN is equal DAT)
'Janek's weight is equal to Tomek's.'
Comparison of the above examples with (4849) shows that this type of con-
struction differs from Type I only by virtue of involving foreign erasure of a
neighbor of a Gen arc by a neighbor of another Gen arc, i. e. by virtue of be-
ing a beheading construction. In terms of APG PNs, the structure of such
sentences looks like that shown in Figure 11.
a. English: the same as in Type I.
b. Polish: the same as in Type I.
34 Stefan Dyta
3.4 Type IV
Another type of the inversion/CA construction can be best illustrated by the

following example:
(56) Harry's weight equals Sam's.
The fact that Sam's immediately follows the verb indicates that the nominal
it modifies is the head of a final 2 arc in (56). Therefore, it is quite safe to
claim that this type of construction differs from the previous one only in
that it is also a 3 - 2 advancement construction. In other words, sentences
like (56) involve 2 arc local successors of 3 arcs. In APG, such sentences are
analyzed as involving PNs of the form presented in Figure 12.
a. English: to parallel
3.5 Type V
A more complex type of construction is shown by the following examples:
(58) Harry is equal to Sam in weight.

(57) Figurel2.
(59) Janek jest rowny Tomkowi wag^.

(NOM is equal DAT weight/INSTR)
'Janek is equal to Tomek in weight.'
(60) Ten towar jest podobny do tamtego towaru pod wzgl? dem
jakosci.
(this merchandise/NOM is similar to that merchandise/GEN
with regard quality/GEN)
'This merchandise is similar to that merchandise as far as
quality is concerned.'
(61) Ten towar jest jakosciowo podobny do tamtego towaru.

(this merchandise is qualitywise similar to that merchandise/
GEN)
"This merchandise is similar to that merchandise quality-
wise.'
It appears that this type of construction differs from Type I only in that
it also involves a pair of immigrant Con arcs whose foreign predecessors are
Gen arcs and whose local sponsors are Con arcs which are supports of their
foreign predecessors. Diagrammatically,
36 Stefan Dyta
(62) Figure 13.

a. English: similar, identical, different, to differ, etc.,
b. Polish: podobny 'similar', zblizony 'approximate', rozny 'different', etc.
3.6 Type VI
Consider the following example:
(63) He equals you in strength.
Like the previous type, this type of construction involves a pair of immigrant
Con arcs whose foreign predecessors are Gen arcs and whose local sponsors
are Con arcs which are supports of their foreign predecessors. But it also in-
volves a 2 arc locally sponsored by a 3 arc. Thus, in terms of PNs, the struc-
ture of sentences like (63) is reconstructed as a PN of the form shown in
Figure 14.
a. English: parallel
3.7 Type VII

Consider the following example:
(65) Waga Janka rowna si? wadze Tomka.
(weight/NOM GEN equals REFL weight/DAT GEN)
'Janek's weight is equal to Tomek's weight.'
This type of construction is similar to Type I except that it additionally
involves a dummy pronoun si. This pronoun is the head of a copy arc which
is cosponsored by an initial 2 arc and its 1 arc local successor, its R-sign be-
ing 2. However, since s/'f is totally immune to any test for final 2-hood, which
is not the case with its full-fledged counterpart siebie, it seems reasonable to
assume that any type of construction involving is an antipassive construc-
tion. In the light of the evidence for the well-formedness of antipassive con-
structions in Polish, this solution has some initial plausibility. The problem is
whether the copy 2 arc is overrun by a 2 arc which is a remote local predeces-
sor of the final 3 arc or by a 2 arc which is a local predecessor of the final
1 arc. Thus, sentences like (65) can be reconstructed either in the way shown
in Figure 15 or in the way shown in Figure 16.
38 Stefan D y i a
(66) Figure 15.
1/c5 iL L/c 3 3lc3 1\C2 2\C, 2VZcHrtc3c4c5
o f
(67) Figure 16.


a. English: the construction is unattested
b, Polish: unknown.
3.8 Type VIII

Sentences like (68) constitute another type of the inversion/CA construction,
which differs from the previous type only by virtue of being a beheading con-
struction.
(68) Waga Janka rowna sif Tomkowej.
(weight/NOM GEN equals REFL DAT)
Diagrammatically,
(69) Figure 17.

b. Polish: unknown.
Of course, sentences like (68) can also be analyzed in the way shown in
Figure 18.
40 Stefan Dyta
C3C4C5
(70) Figure 18.
3.9 Type IX
Consider the following sentence from Polish:
(71) Maria rozni si? od Ewy zachowaniem.
(NOM differs REFL from GEN behavior/INSTR)
'Maria differs from Ewa in the way she behaves.'
Sentences like the above have something in common with Type V, as they
are genitive ascension constructions too, and with Type VII, as they involve
antipassivization. Like Types VII and VIII, they can be analyzed in two ways:
either in the way shown in Figure 19 or in the way shown in Figure 20.
b. Polish: unknown.
3.10 TypeX
Hitherto, I have been dealing with inversion/CA constructions in which an im-
migrant, 1 arc whose foreign predecessor is a Con arc and a 3 arc have the
CH(AC 3 C 4 C5
\t
Slf
(72) Figure 1 9 .
C 2 C3C4C5
Yes
(73) Figure 2 0 .
42 Stefan Dyta
same 1 arc local sponsor. The number of predicates which can occur in such
constructions is rather limited, in some cases, I could list only one such pre-
dicate. In this section, I will present a construction in which a 3 arc is locally
sponsored by an immigrant 1 arc foreign sponsored by a Con arc. Moreover,
as will become clearer later, this type of construction is fully productive, as
any predicate which can take a conjunction as a head of the final 1 arc can
occur in this type of construction. This type of construction is illustrated by
the following pair of examples:
(74) Markowi dobrze sif Krzysztofem pracuje.
(DAT well REFL with INSTR work/IMPERS)
'Marek feels that his joint work with Krzysztof goes well.'
(75) Marii nie mieszka sif EWQ zbyt wygodnie.
(DAT NEG live/IMPERS REFL with INSTR too comfort-
ably)
'Maria does not feel comfortable sharing a room with Ewa.'
Characteristically, sentences like the above obligatorily take Manner arcs.
Their structure can be reconstructed as a PN of the form shown in Figure 21.
(76) Figure 21.

The reason why I argue here that in sentences like ( 7 4 - 7 5 ) the final 3 arc
is locally sponsored by the immigrant 1 arc and not vice versa is that neither
Polish nor English CA constructions allow immigrants whose foreign pre-
decessors are Con arcs to be locally sponsored by 3 arcs. That this is true can
be seen in the fact that sentences like ( 7 7 ) and ( 7 8 ) can only be interpreted in
the way suggested in ( 7 9 ) and (80), respectively.
(77) ? I gave those books to John with Mary.
( 7 8 ) ? Datem te ksi^zki Jankowi Mari^.

(I-gave these books/ACC DAT with INSTR)
( 7 9 ) Mary and I gave these books t o John.
(80) Maria i ja dalismy te ksi^zki Jankowi.

(NOM and I/NOM gave these books/ACC DAT)
'Maria and I gave these books to Janek.'

b. Polish: wspotpracowac 'to cooperate', rozmawiac 'to talk', spiewac 'to
s i n g \ d y s k u t o w a c 'to debate', podrozowac 'to travel', etc.,
Notes
1. I would like to express here my gratitude to Dr. Robert Borsley for insightful com-
ments.
2. For a discussion of Inversion in as diverse languages as Russian and Kannada, cf. Perl-
mutter (to appear). For a discussion of conjunct ascension, cf. Pullum 1975, 1977.
3. This approach rules out the distinction between a syntactic rule and a lexical redun-
dancy rule, as a grammar of a particular language does not contain rules like inversion
or conjunct ascension. Of course, restrictions on particular lexical items which can or
cannot occur in a given type of construction are idiosyncratic and should be handled
in the lexicon.
4. In the glosses, the following abbreviations will be used: NOM = nominative, ACC =
accusative, GEN = genitive, INSTR = instrumental, LOC = locative, NEG = negation,
IMPERS = the impersonal form of the verb, PI. = plural, REFL = reflexive, Sg. =
singular.
5. Beheading should not be confused with uncontrolled beheading discussed by Borkin
(1972).
6. In TG, the relation between sentences like the (a)-forms and the (b)-forms of ( 3 8 -
39) was handled by a rule called Promotion (Ross 1973). The following predicates
can occur in this type of the genitive ascension construction:
a. English: to annoy, to amuse, to astonish, to amaze, to bother, to horrify, to ir-
ritate, to scare, to frighten, to terrify, to worry, etc.
b. Polish: denerwowac 'to make nervous', dziwic 'to amaze', rmrtwic 'to worry',
bawic 'to amuse', irytowac 'to irritate', przerazic 'to scare', rozimieszac
'to make laugh', etc.
44 Stefan Dyta
7. Postal (1977) noted that Reflexive, Reciprocal and Unspecified Object (UO) Clauses
behave as if they were finally intransitive with respect to Clause Union, and to ac-
count for this strange fact he postulated that French has a rule of antipassive.
8. Note that the following Polish sentences is ill-formed:
i. *Waga Janka jest rowna Tomka.
weight/NOM GEN is equal GEN
One might claim that the genitive case marking is not sufficient to code the 3-hood of
the final arc headed by the beheaded nominal, but how to explain then the well-
formedness of sentences like (35) above?
References
Borkin, A.
1972 "Conference and beheaded NP's", PIL 5:28-45.
Johnson, D. . - P. M. Postal
1980 Arc pair grammar (Princeton: Princeton University Press).
Perlmutter, D. M.
to appear "Evidence for inversion in Russian and Kannada", Studies in relational
grammar, edited by D. M. Perlmutter.
Postal, P. M.
1977 "Antipassive in French",NELS 7:273-313.
Pullum, G. K.
1975 "Squibs on relational grammar", unpublished papers nos. 1 - 6 , University
College, London.
1977 "Word order universals and grammatical relations", SynS 8:107-111.
Ross, J. R.
1973 "Nouniness", Three dimensions of linguistic theory, edited by Osamu
Fujimura (Tokyo: TEC corp.), 137-257.
NILS ERIK ENKVIST
Contrastive linguistics and text linguistics
0 The overall title of the volume harbouring my paper is Contrastive linguis-

tics: prospects and problems. The broad range of this title is a hint for me to
contribute a general overview of a very comprehensive set of problems, rather
than to bore deep into some specific detail. I shall try to map some areas in
which contrastive and text linguistics overlap, and discuss a few of the con-
crete questions that have come to the fore in the past several years. My paper
will fall into four parts. In the first section I shall attempt a nutshell survey of
the state of the art in text linguistics, in terms of four types of text models
that are in current use. In the second section I shall very briefly list some
types of contrastive problems that can be investigated with the aid of these
different models. In the third part I shall once again raise the complex ques-
tion of models and parameters in contrastive word-order studies. And in the
fourth section I shall, again briefly, make some programmatic remarks about
views of language as structure and language as process, insofar as this distinc-
tion seems to bear on contrastive linguistics.
1 It is by now a truism that all contrastive linguistics must be based on some

theory and model of linguistic description. Without such a foundation we
simply cannot compare and contrast languages in a systematic way. It is an-
other matter that in applied contrastive work the approach may well be eclec-
tic; picking and choosing among different theories and models and perhaps, if
so desired, even using different theories for different areas of work. Those to
whom applied linguistics is an eclectic discipline will hardly be shocked when
linguists change models, like horses, in mid-stream.
To get started, from a considerable height let me give a bird's-eye view of
some major types of text models that have been offered in text and discourse
linguistics1 over the past decades. One could certainly write a history of text
and discourse linguistics in terms of the four model types I shall mention.
Text linguistics began with efforts at explaining intersentential cohesion with
the aid of traditional grammatical concepts, and then enlarged its sphere to
46 Nils Erik Enkvist
encompass various aspects of cognition and interaction. But here I shall only
describe current models, without filling in the full bibliographical data or
viewing the material in a historical perspective. Nor shall I try to relate my
descriptive classification to any linguistic or syntactic theory.
Allowing for some overlap, then, text models can be descriptively grouped
into four major categories:
(a) Sentence-based text models. These models regard texts as strings of
sentences which are given as an input for analysis and description. Sentence-
based models add textual features to the traditional concepts of sentence
grammars. They can explain many phenomena of intersentential co-reference,
for instance in terms of pronominal reference across sentence borders, and
cross-reference (a good example is Halliday Hasan 1976). They can also
show how sentences are given their textual fit with the aid of lexical and syn-
tactic devices, theme-rheme-focus (or topic-comment-focus) structures, and
overall theme-dynamic patterns of the kind described in DaneS 1974, Enkvist
1974, and a number of other papers. Sentence-based models are thus capable
of describing intersentential cohesion such as it appears in the given sentences
of a given text. But sentence-based models must accept the original sentence
division of the text, which itself is part of the input and cannot be manipu-
lated. If we want to show how such an input might be related to other texts
involving a different sentence division of a similar text, we must go on to
devise
(b) Predication-based text models. These models start out, not from a text
consisting of ready-made sentences but from a set of predications and inter-
predicational semantic relations (temporal, causal, etc.). These predications
must be given in some canonical form based on logic or semantics, or directly
in some natural language (cf. Chomsky's kernel sentences of 1957). The model
then shows how these predications can be textualized with the aid of grouping
which involves conjunction and embedding. Such operations are governed
by a text strategy. Thus different strategies will result in different textualiza-
tions of the same input predications, as illustrated for instance in Sigurd 1977
and Kllgren 1979. One might compare a predication to the physicist's atom
and call it a text atom. Just as the same set of atoms can combine into differ-
ent molecules, the same set of text atoms (predications) can be combined in-
to different texts, each of which reflects a text strategy of its own.
Predication-based models can thus explain relations between different tex-
tualizations of the same input predications. The model can of course also be
used analytically: we then take a text and split it up into its underlying pre-
dications. If we want to compare the given text with other potential textuali-
zations of the same predications we can go on to retextualize them with the
aid of different strategies. This makes possible a comparison of a given text
with the other potential texts that can be built out of the same predications
- an operation that is of particular interest to students of style. A pioneering
experiment along these lines was Ohmann 1964.
What a predication-based model cannot do, however, is to explain the
origin of predications, or the reasons why a given speaker or writer has, in a
given situation, chosen to textualize a certain set of predications instead of
another such set. Meeting this challenge takes us on to
(c) Cognitive text models. In a cognitive text model we do not start out
from ready-made predications. We start out from a body of experience and
knowledge out of which predications can be drawn. The usual way of model-
ling this predication-producing process is based on associative networks (Find-
ler 1979). The concepts themselves are usually placed in nodes of the net-
work, their relations appearing as paths between the nodes. A text strategy
then becomes definable in terms of a set of points of entry into, and paths
chosen through, the network.
It may be easy enough to model certain limited areas of knowledge, and
perhaps cognition, in terms of networks. Trying to capture the full range of
generic and episodic experience and knowledge of a normal adult is a daunt-
ing and probably hopeless task. Today different investigators discuss cognitive
models with various degrees of optimism and pessimism; one way of resolving
such debates is to view a given network in terms of a definite task: what can
the network do, and where does it fail?
Even if a cognitive model can show how predications arise and how they
are connected, it cannot explain a speaker's or writer's motives and intentions.
It cannot explain why a given speaker in a given situation has chosen to ex-
tract a particular set of predications out of storage. To do this we shall need
text models of a fourth type, namely
(d) Interactional text models. Interactional models are concerned with the
interactional behaviour patterns of the people communicating. They reckon
with a speaker/writer's intentions, for instance to distinguish between direct
and indirect speech acts or between literal meanings and irony. They are also
concerned with felicity conditions, conversational maxims, face, and the ways
in which turn-taking affects discourses and texts. Under various labels, includ-
ing 'pragmatics', these questions have been discussed in a flood of publications
over the past several years. Samples of interactional approaches, often linked
with cognitive ones, can be consulted for instance in Clark Clark 1977,
Duncan - Fiske 1977, Freedle 1977-80, Givon 1979, Goody 1978, Just -
Carpenter 1977, Osgood - Bock 1977, Rosenberg 1977, Schank - Abelson
1977, Schenkein 1978, and a host of other recent publications in psycholin-
guistics, sociolinguistics and cognitive science.
48 Ns Erik Enkvist
In a wider context we might note that what compels us to enlarge our field
of vision from sentence-based models to predication-based, cognitive, and
interactional ones is usually the question why? rather than the question what?.
Any proper grammar ought to be able to tell us what the sentences of a text
look like in grammatical terms. But if we go on to ask, why does this text
make use of this particular sentence structure?, we shall have to go from sen-
tence-based models at least to predication-based ones. If we ask, why has the
speaker/writer chosen these particular predications?, we get involved with
cognitive and interactional considerations. But once we start worrying about
causal explanations, we shall find no convenient place to stop until we have
reached the ultimate problems of human cognition and interaction.
To this expansionist challenge, different linguists have responded in differ-
ent ways. Some want to circumscribe their discipline and stay in their own
walled garden. Others follow their problems boldly into strange and alien ter-
ritories. The debate between scholars of the two temperaments, the restric-
tionists and the expansionists, goes on and is perhaps at the heart of a great
deal of the disagreement that prevails in today's linguistics. And in this dis-
agreement, text linguistics and discourse studies provide a lot of ground for
experiments and also for contention. Even the restrictionists have been com-
pelled to recognize the need for admitting certain types of discourse rules into
their grammars. Their usual method is to bring in textual features into the
description of single sentences under various devices such as performative
superstructures, presuppositions, old and new information, and the like. The
expansionists want to face cognitive and interactional problems more directly
(and, as they themselves would say, more frankly). Some of them, particular-
ly those approaching language from outside traditional linguistics, are even
prepared to subordinate linguistic concepts to cognitive and interactional
ones. In the models of Schank, for instance, syntactic parsing is an aid to
cognitive operations. This is an outlook fundamentally different from that of
the syntactician who brings in semantic and pragmatic contexts only when he
must, for instance to disambiguate structurally ambiguous sentences in con-
text.
2 After this presentation of four model types in text linguistics we can ask
ourselves what types of contrastive problems can be studied with the aid of
each type of model.
Sentence-based types are adequate for the study of those text-linguistic
problems that can be defined within the single sentence. Prominent among
these problems is the interplay between the syntactic structure and the infor-
mation structure of the clause and sentence, that is, the way in which the
syntactic structure is brought into harmony with the desired distribution of
old and new information in the sentence and the text. Here we have an area
where textual macrostructures govern, and interact with, syntactic micro-
structures. As the syntactic microstructures of different languages differ, and
often differ more than textual macrostructures, the syntactic exponence of
information structure offers very central topics for contrastive study.
Predication-based models must be brought in as soon as our contrastive
studies compel us to manipulate borders between clauses and sentences. In
translations for instance a translator is often compelled to change the clause
and sentence division of the source text. When this happens we should no
longer compare a sentence of the source text with a sentence of the trans-
lation. We should rather regard the translation as a re-textualization of the
text atoms in the source text. Thus in contrastive studies involving translation
we are likely to profit from predication-based models.
Cognitive models will be needed when we suspect that cognition is struc-
tured differently by speakers of the languages we wish to contrast. Among
other things, cognition involves the structural categorization of the world. If
Benjamin Lee Whorf was right in postulating that speakers of different lan-
guages may live in cognitively different worlds, we can no longer compare
cognitive units as if they meant the same. We must reckon with differences in
the meanings of lexemes, which means that the nodal structures of the
cognitive networks illustrating different languages need not necessarily be iso-
morphic. The associative links between the concepts, illustrated by the paths
that join the nodes in a network, will also be different, as are the culturally
determined schemata or frames or scripts which they reflect. To deal with
such contrasts we need cognitive models. To what extent we may wish to for-
malize them, into networks for instance, is another matter. The choice of de-
grees and levels of formalization depends on what we are trying to do.
Finally, as soon as linguistic contrasts correlate with contrasts in inter-
actional patterns, we shall need interactional models of some kind. Inter-
actional considerations must be brought in when we get involved with dia-
logue, turn-taking, speech acts, style, registers of politeness, considerations of
face, and so on. Experience, particularly in anthropological linguistics, tells us
that interactional considerations increase in importance when the languages
we are contrasting are distant in culture, which often implies distance in space
or in time. But even in closely related languages and in different styles and
sociolects within the same language, interactional patterns may vary greatly.
A good example are conventions of politeness: literal translations of texts
from one sociolect or language into another may lead to absolutely disastrous
results unless the translator modifies the text according to complex rules of
decorum. In a wider sense, then, interactional considerations include all social
traditions in language such as matters of politeness, appropriateness and ling-
uistic manners. They also include conformity to social, narrative and literary
traditions where such conformity is indicated. In this sense all of literary
theory and history enters into the interactional study of textual macrostruc-
tures. And the other way round: some grasp of the interactional effects of
textual macrostructures is necessary for all translators who aim, not only at
cognitive equivalence but also at equivalent effect.
3 In the past several years we have learned that the discussion of word order
must be based on an analysis of textual parameters, and perhaps situational
ones as well. But word-order patterns have also been studied from other points
of view. Today we need an approach integrating a number of theories and
models of word order which have often been dealt with one by one.
Important among these theories to list them merely in the form of brief
reminders are: first, syntactic word-order typology; secondly, information
structure and information dynamics in terms of the distribution of old and
new information; thirdly, those aspects of linearity that fall under what I have
called experiential iconicity, as well as the related concept of salience; and, in
the fourth place, the stylistic variation of word-order patterns. Each of these
problems would of course be a subject for a volume of its own. What I shall
say is in the nature of brief reminders and references to principles well known
to us all. I shall, however, use some space to discuss some conceptual, methodo-
logical and terminological problems which often play havoc with meaningful
discussion. I shall indicate such asides and digressions by using small type.
Like all descriptions of language, those of word order too must be filtered
through descriptive models. And all such models lend their own colour to the
description, the colour springing from the theoretical base with its descriptive
limitations. The models of word order that grammarians have used fall into
three major categories.
First c o m e the word-order descriptions of traditional grammarians which w e might
label as relative. Relative word-order descriptions state the relative positions o f constit-
uents: the subject goes before or after the verb, the modifier goes in front of or behind
its head, and so forth. The ordinary, unmarked order is usually defined as basic, and de-
partures from such a basic order are called inversions. Secondly, structuralist grammarians
of the behaviourist persuasion were fond of slot-and-filler models. They defined posi-
tions in terms of fields or slots and then explained what types of elements could fill the
slots. Finally, generative grammarians arrived at certain word-order patterns as a result of
their phrase-structure rules and those transformations whose primary aim was other than
the adjustment o f linear patterning of constituents. If the basic order was not the same
as the desired surface order, special movement transformations had to be added to the
generative system.
All three models have advantages as well as shortcomings. The relative

model is simple and easy to handle, but fails to give us a picture of the com-
plete pattern. The slot-and-filler model becomes highly delicate given a large
number of slots. But it can also be over-delicate: in simple sentence types,

many slots merge or disappear, and such neutralization makes it difficult to
compare slots in different types of sentences. The generative model compels
the grammarian to deal with word order together with a lot of other things,
including morphemes. This is all for the good as a memento that word order
is part of syntax. But the theory of word order itself has been somewhat of a
stepchild in generative-transformational grammars: it is discussed only when
it intrudes beyond ignoring, but not dealt with gladly in its own right.
First, word-order typology. Since Greenberg's epoch-making book on lin-
guistic universale there has been a prominent line of thought classifying lan-
guages by word-order pattern into SVO, VSO and SOV languages. Two of the
most prominent figures in this movement are W.P. Lehmann and Theo Ven-
nemann (survey and bibliography in Kohonen 1979). The trend has been to
view the order of the main constituents, subject, verb and object, as related
to other structural features of a language such as its use of pre- or postposi-
tions and its tendency to place modifiers before or after the noun. These typo-
logies have also been used in diachronic studies. Changes from one word-order
pattern to another have been regarded as instances of typological change, and
typological change has been referred to as explaining, perhaps even giving the
reasons for, linguistic changes involving word order. Certain sets of word-order
patterns thus become something like syntactic universals.
But as many critics have pointed out, word-order typologies also raise a
host of problems. To begin with, no well-documented natural language seems
to represent a type in pure form. All natural languages seem to be typological
mixes. The mixture of typological features is explained as a result of typo-
logical shift: the language is about to move from one type to another so that
some of its features come from the old pattern and others already conform to
the new one. Of course such explanations greatly complicate the issue, per-
haps even to the point of weakening the explanatory powers of the model.
Instead of well-defined types what we actually have is, at best, trends. But
there have been interesting attempts at explaining differences between lan-
guages in terms of their being in different stages of typological change (Jacob-
son 1979).
Another question has to do with the choice of an optimal level of depth
for the description of word-order patterns. Once we allow a situation where
an underlying structure has an order of elements different from that of the
surface structure we must make up our minds as to which of these different
structures is the typologically important one. When studying German, for in-
stance, which of the patterns, the SVO of the main or the SOV of the sub-
ordinate clause, is the one we should use for typological classification? The
answer of course depends on what criteria we choose. Are these criteria mere-
ly reflections of our own grammatical theories and models and of our own
principles of descriptive economy? Or are they based on diachronic principles
supported by historical data? Or do they reflect some kind of psychological
reality or psychological markedness, or are they founded on statistical counts
which identifies 'basic' with 'most frequent'?
A fundamental query about word-order typologies relates to variation. If a
language always arranges its elements in the same order, well and good. But in
fact languages use numerous alternative word-order patterns. To begin with,
even languages whose word-order mechanisms serve largely syntactic ends
('analytic' languages such as Modern English) allow more variation than we
often realize. Such variation occurs for instance in impromptu dialogue which
is rich in what syntacticians like to call 'deviance' (and which is nevertheless
a wholly normal phenomenon). And there is a lot of variation that might be
labelled as stylistic, whatever that ultimately means. The question is, once
there is variation in word-order patterns, how can such variation be accom-
modated into a word-order typology which by definition implies the domi-
nance of one pattern over the others?
The only road to a satisfactory answer involves a study of the factors moti-
vating the variation. We must identify the forces that cause modifications in
the basic pattern. We must thus make statements such as "language X is SVO
except when forces p, q and are strong enough to motivate the use of other
patterns such as OVS".
As such discussions are often expressed in terms of markedness, another terminological
note may be in order.
One way of speaking about 'markedness* involves a contrast between a form which
has a marker and another form that has not: b is marked and unmarked because b has
a marker, namely voice, which does not have. (Markers can be manipulated, however:
we might also say that is marked and b unmarked because has aspiration and b has
not: some principle of naturalness would be necessary to constrain such manipulation.)
A second way is to look at distribution: -s is the unmarked plural in English and -n as in
oxen one of the marked ones because -s has a wider distribution and is applied to a larger
number of substantives. A third approach is statistical frequency: the most common
forms in a corpus are unmarked, rarer forms marked. A fourth possible approach is based
on informant responses. What native speakers regard as natural, common, ordinary and
'unremarkable' is therefore unmarked, and what they regard as strange or strained is
marked.
Yet another difficulty is that markedness may be both paradigmatic and syntagmatic.
A form such as The apple John ate is marked when seen out of context (at least marked
in some of the senses given above). But in a context such as The apple John ate but the
oranges he only sniffed at, and when pronounced with proper focussing, it may be en-
tirely natural. Thus such a sentence may be paradigmatically marked but syntagmatically
(textually, contextually) unmarked (and there are parallels to this in phonology and
lexis as well).
Thus 'markedness' can mean a number of different things. In discussions we ought to
make clear which type of markedness we are talking about - or, alternatively, use differ-
ent terms for what I have here called different types of markedness.
One force that affects word order is information structure. In context,

every sentence has a contextual (textual as well as situational) fit. The sen-
tence must be accommodated to fit into its contextual envelope; to use an-
other metaphor it must be provided with devices that plug it into the preced-
ing text and make possible a contact with what comes after. Textual fit is
largely a matter of that information structure which signals what in the sen-
tence is old and what is new information. The usual terms used in analyses of
information structure are theme and rheme, topic and comment, and focus
(as well as presupposition). As these terms have been used in many senses,
and as many discussions have been bedevilled and even spoiled by termino-
logical disagreements, a bit of conceptual and terminological analysis is once
again indicated.
As terms, theme/rheme have roots in Europe, topic/comment in America, with notable
exceptions (Susumo Kuno speaks about theme and rheme, Osten Dahl and Petr Sgall
about topic and comment). Some linguists (like the present writer) have use for both
theme/rheme and topic/comment: to me, a topic is the kind of, or part of, theme which
has been fronted through a movement operation called topicalization.
Next a concise taxonomy of theme/rheme (or topic/comment) definitions:

1. Logic-based definitions
a. Theme equals logical subject, rheme equals logical predicate. Such definitions pre-
suppose canonical forms for predications specifying what is logical subject and
what logical predicate.
b. Definitions based on quantifier scope: the most scopeful element (= the element
marked with a quantifier which has maximal scope) is the theme. Compare
(Everybody in this room (speaks (two languages)))
(Two languages (are spoken by (everybody in this room)))
Such definitions can be used in instances such as the above, in which the relative
scopes of the all-quantifier and the existential quantifier are verifiable through our
semantic intuitions. But they are hard to apply to sentences of the type
(Six girls (bought (three books))) / (Three books (were bought by (six girls)))
where there is more ambiguity: did the girls buy three books each, or three books
collectively? how do we know? which is more likely, and why?
2. Psychological definitions
a. Intuition-based definitions: theme is what people think they are speaking or hear-
ing about, rheme is what they think they are speaking or hearing about what they
think they are speaking or hearing about.
b. Definitions based on psycholinguistic models: themes are elements foregrounded
in short-term memory, they are elements on top of a push-down store, etc.
3. Definitions based on information structure: themes express information that is 'old*
in the sense that the speaker/writer thinks the receptor already knows it because of
life under the same stars, a shared cultural or episodic experience, situational recover-
ability or previous mention in the text.
a. Dichotomous definitions: sentences or clauses have two parts, old versus new in-
formation and thus theme versus rheme, and can be cleanly segmented into these
informational constituents.
b. Scalar definitions: the oldness and newness of information form a scale, and there
can be different degrees of thematicity and rhematicity (or communicative dyna-
mism to use the famous term used by the most prominent protagonist of this
view, Jan Firbas of Brno).
4. Definitions based on syntactic form. The theme/rheme contrast can be marked mor-
phologically (by articles, special thematic markers etc.) and by word order (in some
languages one could try defining themes as sentence- or clause-initial elements).
These definitions need not necessarily be mutually exclusive. Indeed one might try the
hypothesis that the 'least marked' thematization occurs when all definitions agree: in
English a 'least marked' theme would then be a sentence-initial subject or adverbial ex-
pressing old information, intuitively understood to signal what the sentence is really
about, having maximal scope (when measurable in spite of the fuzziness of natural lan-
guages), and functioning as logical subject in its canonical predication. This is a problem
I cannot pursue here.
As with 'markedness' we should always make clear in what sense we are using theme
and rheme or topic and comment, or, alternatively, abandon these terms and speak about
logical subjects and predicates, maximally and minimally scopeful elements, psycho-
logical subjects, old and new information, initial and non-initial constituents, etc. An ex-
treme Occamist might even argue that theme and rheme are comparable to the ether of
now-outdated physics: we can do without them if we know what we are really speaking
about.
One of the interesting additions to linguistic typology is the classification

of languages into topic-prominent, subject-prominent, both topic- and subject-
prominent, and neither topic- nor subject-prominent, suggested by Li and
Thompson (in Li 1976). In contrasting a wide variety of languages, including
many from East Asia, Li and Thompson found that different languages rank
the relative structural importance of topic and subject very differently. Some
languages regard topic as primary, whereas others pay more attention to sub-
jects than to topics. Li and Thompson also postulated that certain diachronic
changes can be related to their typology. The reason why this typology is of
interest to students of contrastive linguistics is that it re-systematizes an old
and basic type of observation. In fact, all studies of theme and rheme original-
ly started from linguistic comparison and contrast. Thus Henri Weil compared
Greek and Latin word order with that of French, and Vilem Mathesius that of
Czech and English. Granted, Li and Thompson label all Indo-European lan-
guages as subject-prominent (cf. the SAE of Whorf). But all the same we
know that even closely related languages may have different ways of handling
theme and rheme. Swedish for instance allows object-initial patterns such as
(1) a Sin flicka lskade sjmannen.

*'His girl loved the sailor.'
b Lraren ansg eleverna vara obegvad.
*'The teacher regarded the pupils as untalented.'
c Osten t rt tan i gr.
""The cheese ate the rat yesterday.'
which are disallowed in English. Delicacy could be added to the Li-Thompson

typology to adapt it to the contrastive study of more closely related languages.
In contrastive work we should of course bear in mind that syntax is not

the only influence on word-order patterns that reflects on information struc-
ture.
In English we have four types of devices regulating the information flow in the sentence.
The first is lexical: compare A is B's teacher / is A's pupil, A is older than / is
younger than A. The second involves changes in syntactic structure such as passivization
or extraposition or dislocation. The third is topicalization or commentization, meaning
front or end placement without further change in syntactic structure. And the fourth is
the use of marked focus (see Enkvist 1980), either by clefting or by intonation. The
function of marked focus is to mark new information chosen from among the items in a
presuppositional set; often such marked new information occurs in a place where w e ex-
pect old information, i.e. at the beginning of the clause or sentence.
These different devices are likely to have different markedness values. Lexical themati-
zation by the use of converses often seems intuitively less obtrusive than, say, the use of
marked focus or of topicalizations. This markedness scale would be a nice subject for
contrastive investigation. As Swedish topicalizes more liberally than English, topicaliza-
tions in Swedish may well strike us as less marked (in the sense of 'less noticeable and
strange') than the corresponding ones in English.
In this light, then, the basic word-order typology discussing the relative
order of subject, verb and object (and sometimes the "X" which means 'other
constituents') should be integrated with theme-rheme typology so that the
thematically motivated modifications in basic word order could be made
maximally clear.
Those who set out to clarify these matters will find a number of interesting
problems in their path. One fundamental principle which on the whole re-
stricts and constrains thematic modification of basic word-order patterns is
what I shall call the principle of valency: elements that are intimately bound
to each other by strong valency bonds should also stay close to each other in
the linear sequence. There have been many attempts at stating such movement
restrictions: Ross's constraints, cross-over principles, anaphoric islands and
other such devices have been presented in this very context. But many of
these word-order phenomena in fact also have stylistic correlates. There are,
in other words, styles in which certain word order patterns are permitted
which would be awkward in other styles.
Let me take just one example. Adverbials which are tied to their verbs by
semantic specification are usually hard to move, for instance to topicalize.
Thus the verb to put implies both object and place (you must put something
somewhere). Hence examples such as
(2) a *Put!
b *Put it! *Put your hat!
c Put it down! Put your hat under your chair!
d ??Under the chair he put it. He put it under the chair.
e ?Into a drawer he put the book. Into the drawer he put a

book.
f Into the drawer he put the book he had been reading all
afternoon.
It is, however, possible to improve the acceptability of some such sentences

by contextualization, a process in which we motivate paradigmatic marked-
ness by syntagmatic naturalness and excuse syntax with textual fit. Compare
g Into a drawer he put the book; the manuscripts had to be

thrown away.
h Into a drawer he put the book; into the wastepaper basket
he threw a whole sheaf of manuscripts.
In these instances there is a conflict, a tug-of-war, between different prin-

ciples: the syntactic principle of valency which wants to keep related ele-
ments together and retain the adverbial after put, and some other principle or
principles that want to move the adverbial away from put, for instance up
front. And what are these other principles? I can at once think of three, all of
them textual in that they involve parameters outside or beyond the sentence.
One we have already seen. This is information structure which requires old
information to come first and new information later. If the adverbial expresses
old information and links the sentence to the previous text or the situation it
merits front position and concomitant straining of the valency bonds. This is
even more plausible if the sentence can end with a heavy constituent which
contains a great deal of new information (principle of end weight). Another
principle motivating fronting of valency-bound adverbials is textual iconicism
which relates the pattern of one clause or sentence iconically to that of an-
other neighbouring clause or sentence (as in (2h) and chiastically in (2g)).
Such iconical motivation is of course most common in poetry and in artistic
prose. A third principle is experiential iconicism, as I have called it.
By experiential iconicism I mean an isomorphy between the text and our
experience of the world. In other words, in instances of experiential iconicism
symbols in the text are ordered in the same way as their referents in the world
of things or the world of events. Thus if a text tells us about events in the same
order in which they took place the text becomes experientially iconic, as in
(3) a They got married and had a baby.

b They had a baby and got married.
Compare
c They had a baby after they got married.
d They got married after they had a baby.
(c) and (d) are not experientially iconic because the order of predications in
the text is not identical with the order of events. Generally, whenever we con-
join predications which can be interpreted as having a temporal or causal rela-
tionship, we understand them to be experientially iconic as in (3 a) and (3 b).
'If they are not, the departure from experiential iconicism must be marked, as
in (3c) and (3d), in which after carries the onus of marking the departure.
I have elsewhere (forthcoming in Text, volume 1 number 1) discussed ex-
periential iconicity at some more length and pointed out that it can be of
three major types: temporal (as in (3)), spatial, and social (as in ladies and
gentlemen! rather than gentlemen and ladies!). What is of particular interest
here is the fact that corpus studies show the importance of experiential
iconicism in texts that have to order experience for the benefit of the receptor.
Such texts are instructions, for instance guidebooks and cookery-books. In
cookery-books we find examples in which strongly valency-bound adverbials
are fronted, as in
(4) Into a champagne glass put a lump of sugar, an eggspoonful

of brandy, and . . . (LOB Corpus 19 139)
(5) Peel, core, and slice enough apples to make a dome in the pie
tin, and set aside. In a saucepan put sufficient water to cover
them . . . (Brown Corpus E l l 1640)
The motivation for such fronting, which would seem very strongly 'marked'
in the sense of 'rare' in isolation, is found in text strategy. The writer's aim is
to indicate the order of operations: by (4) she is implying 'take a champagne
glass and put into it a lump of sugar' and by (5) 'take a saucepan and put into
it sufficient water. ..'. Here, then, the text strategy proves strong enough to
override the syntactic inertia and to stretch the valency bonds between put
and its adverbials. In textual corpora such fronting of place adverbials with
put is rare. My pupil Ann Westerlund tells me that she has found some two
examples out of some four hundred potential instances in the Brown Corpus
(while adding that such counts always build on arguable principles of classifi-
cation).
The discussion of the connection between a certain genre or text type,
such as a cookery-book, and sentence structure leads us into stylistics. In
terms of what I have said so far, different styles can be said to offer different
degrees of resistance to the use of different types of devices that regulate
textual fit. In each style and substyle each thematic operation has a threshold
of a certain definite height. In spoken, informal, impromptu dialogue, there is
a low threshold for dislocations of the type John he is a nice chap and She is a
doctor, Susie is. But they do not occur in laws or statutes or legal contracts:
the threshold is too high to let them in. Similarly, the threshold against front-
ing place adverbials with put is very high, except in styles containing operating
instructions, such as a cookery-book style. Here the threshold ist lowered by
the need for experiential iconicity.
This is of course only a metaphorical way of stating what is common and
what is rare in different functional varieties of language. But the images of a
tug-of-war, of a dualistic struggle between different forces, of the forces
stretching valency bonds and of the syntactic inertia being overcome by re-
quirements of textual fit, or of different styles having thresholds of different
heights against different syntactic structures all have one thing in common.
They link up with a view of language as a process rather than simply as a set
of structures. In these terms, the ultimate syntactic form of a sentence, in-
cluding its thematic arrangement, is the outcome of a process which is a result-
ant of different, often opposing, forces. In different situations these forces
acquire different relative weights. One such force is the syntactic inertia of
the sentence. The inertia helps to maintain a canonical, typologically funda-
mental, and in that sense least-marked word-order pattern. But another set of
forces strives to adjust the sentence to its textual and situational surroundings
and to give it an optimal textual fit. What is optimal in textual terms depends
on the speaker/writer's intentions and motives, on the text type and on the
text strategy. These are exposed through textual characteristics that have been
studied, among other things, under the heading of 'style'.
With these few examples and many metaphors I have tried to suggest an
avenue towards resolving the tensions between canonical forms such as those
of language typologies, and the variation that we actually find in texts. In
brief: canonical forms are subjected to various textual and situational stresses
and strains, and in different situations they offer different degrees of resist-
ance. Many of the forces straining the syntax of the sentence by compelling it
into more strongly marked patterns are textual and discoursal. They cannot
be found and understood unless we operate with a textual apparatus. In some
text types and styles the canonical forms tend to prevail more strongly, and
in others they are more freely modified. To extract the maximum value out
of linguistic typologies we should study and learn to define the forces that
compel departures from typologically basic patterns. Once these departures
are reduced to rule, our typologies can be set up with more confidence.
At the same time the concept of a dualistic struggle between canonical
forms and modifying forces helps us to understand what happens in im-
promptu speech and in syntactically deviant literary and poetic texts. The
task of a text is not necessarily to present a receptor with a set of perfect syn-
tactic and lexical patterns. What a text must do is to trigger off a process of
interpretation. In principle any text which succeeds in triggering off a success-
ful interpretive process is a success, irrespective of whether its linguistic form

follows the patterns linguists choose to call 'grammatical'or not. In different
situations interpretability requires different degrees of textual explicitness
and elaboration. In some situations for instance when two people work to-
gether on some familiar task such as pulling up a boat or pushing a car sim-
ple exclamations such as now! or two more feet! or even a grunt may be per-
fectly interpretable. In other situations - for instance in statutes or laws
proper interpretability must be founded on a maximally complete and ex-
plicit verbalization. As little as possible should be left to the reader's arbitrary
interpretation, and in the absence of clues that can be retrieved from the con-
text and situation, the receptor must be given very full and detailed instruc-
tions as to how to interpret the text. Most situations fall somewhere on the
scale between these extremes: laws and statutes that demand maximal ex-
plicitness, and familiar situations in which simple, perhaps syntactically deviant
and linguistically unstructured signals will do the job. Syntactically deviant
poetry seems to build its effects partly on its ambiguity which at best increases
semantic pregnancy, and partly on the pleasure people seem to derive from
extracting meanings out of deviant patterns.
4 If these views are correct, the form of a sentence and therefore also the
form of a text can be seen as results of a set of text-productive processes. In
discussing word order, for instance, one should not only describe structures
as such, but also view them as an output of a set of processes which have been
triggered off by certain definable forces. Similarly the interpretation of a text
is a result of an interpretive process.
Such arguments are becoming so prominent in today's linguistics that one
is tempted to ask: are we in fact moving from a predominantly structural lin-
guistics to a predominantly processual 2 linguistics? Like Humboldt, many
linguists today like to see language as energeia and not only as ergon, as pro-
cess and not only as structure.
I do not want to insist that we have, overnight and through a dramatic
paradigm shift, moved out of structuralism into processualism. Indeed the
history of linguistics shows us many movements which have combined struc-
ture and process. Even in the more extreme variants of behaviorist-structural
linguistics, units such as phonemes and morphemes could be defined pro-
cessually: a phoneme or a morpheme was what one arrived at if one segmented
a string according to certain closely defined processes. And in generative-trans-
formational grammar too, re-writing operations and transformations are pro-
cesses. They explain how one state (or, in linguistics, structure) changes into
another state (or structure). And such a change is precisely what we mean by
a process.
Thus the difference between structural linguistics and process linguistics is

not that the former deals exclusively with structures and the latter only with
processes. Both can make use of both. The difference is rather that in process
linguistics we no longer use processes in order to arrive at structures, as did
the structuralists and transformationalists. We rather use structures to expli-
cate the nature of processes. The difference, then, is one of emphasis and goal:
structuralists use processes to arrive at structures, processualists use structures
to arrive at processes.
This distinction brings with it an important corollary. If processes are mere
auxiliary constructions which we can set up at will to arrive at structures, they
get their shape and justification from the structures. Any process that results
in the desired structure is all right. To cite Andrew Chesterman (1980: 18),
"in the product grammar processes are purely formal, such as those known as
transformation, substitution, derivation, etc., and the motivation for postulat-
ing them is given by the internal requirements of the grammar." In process
linguistics, on the contrary, the processes ought to be 'real' in the sense of re-
flecting phenomena beyond the internal requirements of the grammar. They
ought to be anchored in processes outside the grammar itself. There are in
fact two obvious places where we can look for processes on which we can base
our descriptions of language. One is psychology. To go on citing Chesterman
(1980: 18),
In the psychological grammar [. . . ] processes are behavioural, genuinely dynamic, they
take place in time, and they are thus claimed to have a definite objective existence ex-
ternal to the theory [ . . . ] a grammar would seek to describe actual psychological pro-
cesses and capacities. As a branch of cognitive psychology it will be expected to for-
mulate its claims in such a way that they can be empirically tested like any other hypo-
thesis in the behavioural sciences [ . . . ] . We would expect such a grammar to be experi-
mentally based, closely linked to the study of memory, neurolinguistics, cognition, per-
ception, etc.
The other kind of reality, beside the psychological, to which we can appeal in
the quest for real processes beyond natural language is the world of the com-
puter. If we base process models of language on psychology we are pattern-
ing them on what we know about human intelligence; if we base them on
computer science we are writing grammars patterned on artificial intelligence.
In saying this I am of course begging the question, what precisely is the rela-
tion between human intelligence and computer processes? What can man do
that computers cannot? And what can computers do that man cannot com-
fortably imitate? These are questions easy to put and hard to answer.
A question more relevant to our present context is, what can process lin-
guistics and processual thinking contribute to contrastive linguistics?
Structural linguistics uses structural concepts which vary with different
models. Typical examples of structural units are distinctive features, pho-
nemes, morphemes, lexemes, phrases, clauses, sentences, and suprasentential

text units. They are handled differently in different types of models. Process
linguistics uses processual units such as strategies, plans, and processes.
A strategy can be defined as a set of weighted principles for a goal-oriented,
purposeful choice between alternative modes of action. When a strategy is
made concrete and adapted to a certain specific situation it becomes a plan.
And the modes of action themselves involve processes which change one state
or structure into another.
To describe a strategy we shall therefore need a description of the goal and
purpose of the choices, an inventory of the available alternatives (which in-
volve processes and their output), and an analysis of the various factors and
parameters that affect the choice together with the relative weights of each
parameter. We must state what constellation of parameters signals a strategic
decision point, and thus define the situations in which strategically motivated
choices become necessary. As so-called garden-path sentences such as The
canoe floated down the river sank show, we may also need the concept 'con-
firmation point': a confirmation point is the point at which assumptions
made at previous decision points are shown to be right or wrong. This ter-
minology should of course not be taken to suggest that decisions in text pro-
cessing must be made after conscious deliberation. On the contrary many of
them are made on the basis of our linguistic and communicative experience
and 'competence' without conscious attention.
As decision theorists tell us, choices between alternatives can be modelled
with the aid of different types of models and parameters. A deterministic
model determines decisions fully by the relevant parameters: there are no real
options and no real alternatives, and in that sense no strategies: the only
strategic decisions consists of either entering or not entering the system. A
statistical model determines choices by citing statistics of past experience;
alternatives can be weighted depending on to what extent they favoured or
prevented a relevant goal. A stochastic model is based on estimates of random
choice, perhaps with subjective weightings in the absence of past experience.
And a strategic model proper is founded on estimates of the relations of the
choice to the purpose: a strategic model looks forward to a goal, rather than
backwards to past experience like a statistical model. Different decision
models can of course be combined within the same decision system: some
decisions may combine strategic and statistical considerations, whereas others
may remain stochastic. Making a strategic decision is in itself a process involv-
ing phases such as problem perception, problem definition, inventory of alter-
natives, inventory of relevant parameters, weighting of parameters, the choice
itself, and the realization of the chosen alternative through the appropriate
process or sequence of processes. The output of one process may itself be
evaluated in terms of a new strategy, which in turn triggers off a new plan,
leading to a new process.
A process, as I said above, is the change of one state (in linguistics, struc-
ture) into another. To identify the trigger of a process we need a strategy. Both
strategies and processes form hierarchies: communicative macrostrategies lead
on to microstrategies, processes lead on to subprocesses. In describing strate-
gies and processes we must decide what is the optimal delicacy level for our
descriptions: we can either trace them in major outline, or we can try to work
them out in delicate detail through a large number of intermediate stages. In
practice, our descriptions of strategies and processes can be oriented either
towards message production or towards message interpretation. But we should
not forget that production strategies too are usually receptor-oriented. People
speak and write to be understood, and text strategies are always founded on
estimates of what the receptor already knows or does not know, and how
much information the receptor can process (information density). The Gricean
postulates, theme-rheme studies, discussions of implication and presupposition
and entailment and the like are all concerned with the sender's estimates of
the knowledge and of the message-processing capacity of the receptor. Such
estimates are supremely important in determining the content and shape of a
message.
A concrete example may be in order to show what I mean by weighting of
strategic parameters. Let us assume that we have a choice between three ex-
pressions: it's getting light, the blank day breaks on the bald street, and on
the bald street breaks the blank day. Which should we prefer? If our goal is to
produce ordinary informal conversation over morning coffee, it's getting light
would presumably be the most appropriate alternative. The blank day breaks.. .
would reveal the speaker's poetic vein but maintain the canonical SVO order.
But when composing/ Memoriam Tennyson preferred the third alternative:
(6) He is not here; but far away

The noise of life begins again,
And ghastly through the drizzling rain
On the bald street breaks the blank day.
Tennyson's purpose was to write a rhyming poem, and the requirements of
rhyme become a weighted parameter in his text strategy. There are other
reasons that conspire to bring about the same choice. By fronting on the bald
street Tennyson makes us visualize the street first. Having seen the street we
see the day breaking over it. At the same time the street is old information
because it has been mentioned two stanzas earlier in the poem. Thus at least
three text-strategic factors (rhyme, old information before new, experiential
iconicity) conspire to motivate the process of inversion, and their combined
strength suffices to overcome the syntactic inertia which tries to maintain the
canonical and unmarked SVO pattern. We might add in passing that one cate-
gory of people who have used an approach evaluating text-strategic parameters
are the literary critics. When criticising a literary text a critic tacitly compares
that text with other alternative texts which he, the critic, has either experi-
enced or conjured forth out of his own imagination. It is such a comparison
that provides the foundation of critical judgments.
I have here dressed up my discussion of Tennyson in terms of production
strategies, but of course analyses of strategies and processes can be focussed
on reception as well. One of the areas of linguistics where one can profitably
look for concrete examples of processual views of text reception is parsing, a
process carried out by the receptor according to conventions and strategies
anticipated by the sender. To add another small morsel of flesh to my pro-
grammatic skeleton I shall give an example of a processual approach to one
detail in the contrastive study of English, Swedish and Finnish.
In Swedish there is an optional element which has sometimes been called
'resumptive s' which can be inserted between an adverbial clause and a fol-
lowing main clause and roughly rendered by English then, as in
(7) Om Kalle kommer (s) gr jag till farfar.
'If Charlie comes, (then) go I to grandfather.'
In Finnish the corresponding element is optional resumptive niin, as in
(8) Jos Kalle tulee (niin) min menen kotiin.
'If Charlie comes, (then) I go home.'
Resumptive sd and niin are both more frequent in spoken and in informal
written styles than in formal writing (cf. Platzack 1973: 58ff.).
A structural description would presumably simply note that sd and niin
can occur in the juncture from adverbial to main clause, particularly in spoken
and informal written styles. In the light of parsing strategies and processes,
however, their function appears in a clearer light. One of the main problems
of parsing is, of course, to know at what point a phrase or clause ends and
another phrase or clause begins. This is a problem students of parsing have
discussed in terms of a Principle of Early Closure (according to which the
parser should close a phrase unless the following elements can be construed
as parts of that phrase) or a Principle of Late Closure (whenever possible the
parser should attach an element to the phrase or clause currently being con-
strued) (e.g. Frazier 1979: 33 and 43). A speaker or writer who wants to
maximize the clarity and ease of parsing of his message would obviously do
well in marking closure points such as junctures between clauses. In English
this strategy can be carried out largely through processes involving intonation
and perhaps rhythm and pause patterning. In Swedish and in Finnish the
strategy can be realized through the process of sa or niin insertion. These ele-
ments are parsing cues triggering off a certain type of segmentation and struc-
turing process. In English too we can sometimes use then in a similar function;
in Old English pa had a comparable function (particularly in pa . . . pa ...).
The marking of such closure points is of course especially useful in speech,
where the receptor cannot parse and re-parse the message at leisure. This ex-
plains why s and niin are particularly frequent in spoken discourse and in-
formal writing. That they seem more common than English then in this func-
tion may be connected with the fact that English generally makes more use of
intonation than Finnish or Finland-Swedish (I am not sufficiently familiar
with the Swedish of Sweden to venture an opinion at this point).
Another way of illustrating the functions of resumptive s and niin is to
show how they avoid garden-path phenomena by marking junctures between
clauses in instances such as
(9) a Nr Kalle ter fr (s) fredrar han stekt potatis.

'When Charlie eats mutton he prefers fried potatoes.'
b Nr Kalle ter (s) fr han inte se p television.
'When Charlie is eating he must not watch television.'
(10) a Kun is otti pienet lapset (niin) iti muutti pois.
'When father took the small children, mother moved
away.'
b Kun is otti pienet (niin) lapset lhtivt karkuun.
'When father took a little drink, the children ran away.'
c Kun is otti (niin) pienet lapset lhtivt karkuun.
'When father took [a drink] the small children ran away.'
To sum up: while marking closure of the adverbial clause, s and niin simul-
taneously also mark the continuation of the sentence through the beginning
of the main clause. Their use can be explicated with the aid of a predication-
based text model.
Thus the full structural and contrastive significance of resumptive s and
niin best appears when we see them as cues to parsing and as processing sig-
nals. In this light they no longer seem redundant and in that sense marginally
functional. On the contrary they are part of the receptor-oriented strategies
of the person producing the text.
It should now be clear what use observations of these kinds can be in con-
trastive linguistics. They enable us to compare parsing strategies and to teach
our pupils to observe, and to use, parsing cues in a more systematic way. For
Swedes and Finns who learn English the closure markers of their native lan-
guages signal places where they should learn to use tone-unit boundaries and
an intonational rise in English. And conversely those learning Swedish or Fin-
nish will be helped in their parsing by the junction marking with s or niin.
Such junction markers at the same time increase the redundancy of the mes-
sage and thus make comprehension easier. Further, once we start drawing
attention to parsing cues such as sa and niin we get involved with all the other
devices that Swedish and Finnish use to make clear where are the junctures
between phrases and clauses. We are learning to pose new questions which
unroll the entire field of parsing strategies and processes. And in so doing we
are conforming to the basic principle of process linguistics: we are using
structures and structural cues to illustrate a process, namely parsing. To trans-
lators too, processual thinking is useful. For instance those who wish to trans-
late Tennyson's In Memoriam will have to analyze the original text strategy
and decide how it should be adapted, and how its parameters ought to be
weighted, and perhaps re-weighted, to produce an optimal rendering in the
target language.
It is, then, text linguistics, together with psycholinguistics and cognitive
science, that has helped to bring process linguistics to the fore. Of course the
basic concepts of strategies and processes are also immediately applicable to
another area of applied linguistics, namely language learning. But that is an-
other subject.
Notes
1. To some people the terms 'text linguistics* and 'discourse linguistics' are synonymous.
To others, text linguistics is the linguistic study of texts as such, whereas discourse lin-
guistics also reckons with the situational context: it is the situational envelope that
turns a text into discourse.
2. The term 'procedural' is also in use. I prefer 'processual' or 'process' linguistics because
so many other currents in linguistics have emphasized procedures and been in that
sense 'procedural'. For example the behaviorist structuralists emphasized discovery
procedures and adopted procedural definitions of linguistic units: a certain unit was
what one got if one segmented a string according to a certain procedure.
References
Beaugrande, Robert, de
1980 Text, discourse and process. Toward a multidisciplinary science of texts
(Norwood, New Jersey: Ablex).
Chesterman, Andrew
1980 "Contrastive generative grammar and the psycholinguistic fallacy", Papers
and studies in contrastive linguistics 11:17-24.
Clark, Herbert H. - Eve V. Clark

1977 Psychology and language (New York: Harcourt Brace Jovanovich).
DaneS, FrantiSek
1974 "Functional sentence perspective and the organization of the t e x t " , Papers
on functional sentence perspective, edited by F. Dane? (Prague: Akademia),
106-128.
Dijk, Teun A. van, ed.
1980 "Story comprehension", Special issue of Poetics 9:1 - 3 .
Duncan, Starkey - Donald W. Fiske
1977 Face-to-face interaction (Hillsdale, New Jersey: Lawrence Erlbaum).
Enkvist, Nils Erik
1974 "Theme dynamics and style", Studio Anglica Posnaniensia 5:127136.
1980 "Marked focus: function and constraints", Studies in English linguistics for
Randolph Quirk, edited by S. Greenbaum et al. (London: Longman), 1 3 4 -
152.
1983 "Chairman-rapporteur's summary", Papers from the Nobel Conference on
text processing held at Stockholm in August, 1980, edited by S. Allen
(Stockholm: Almqvist & Wiksell), 6 1 7 - 6 3 6 .
Findler, Nicholas V., ed.
1979 Associative networks (New York: Academic Press).
Frazier, Lynn
1979 On comprehending sentences. Syntactic parsing strategies (Bloomington:
Indiana University Linguistics Club).
Freedle, Roy O., ed.
1 9 7 7 - 8 0 New directions in discourse processing, I - I V (Norwood, New Jersey:
Ablex).
Givon, Talmy, ed.
1979 Discourse and syntax [Syntax and semantics 12 ] (New York: Academic Press).
Goody, Esther N., ed.
1978 Questions and politeness. Strategies in social interaction (Cambridge: Cam-
bridge University Press).
Greenbaum, Sidney - Geoffrey Leech - Jan Svartvik, eds.
1980 Studies in English linguistics for Randolph Quirk (London and New York:
Longman).
Greenberg, J. H.
1966 Language universals. With special reference to feature hierarchies (= Janua
linguarum, series minor 59) (The Hague: Mouton).
Grimes, Joseph
1975 The thread of discourse (The Hague and Paris: Mouton).
Halliday, M.A.K. - Ruqaiya Hasan
1976 Cohesion in English (London: Longman).
Hammarberg, Bjrn, ed.
1979 Kontrastiv lingvistik och sekundr sprkforskning (Stockholm: Institutionen
fr lingvistik vid Stockholms universitet).
Hopper, P. J. - S. A. Thompson
1980 "Transitivity in grammar and discourse", Language 56:251 - 9 9 .
Jacobson, Sven
1979 "The placement of adverbs in principal and subordinate clauses in English
and Swedish", Kontrastiv lingvistik och sekundrsprkforskning, edited by
B. Hammarberg (Stockholm: Institutionen fr lingvistik vid Stockholms
universitet), 3 1 - 3 9 .
Just, Marcel Adam Patricia A. Carpenter
1977 Cognitive processes in comprehension (Hillsdale, New Jersey: Lawrence Erl-
baum).
Kllgren, Gunnel
1979 Innehlt i text (Lund: Studentlitteratur).
Kohonen, Viljo
1978 On the development of English word order in religious prose around 1000
and 1200 A.D. (Abo: Publications of the Research Institute of the Abo
Akademi Foundation, 38).
Li, Charles N., ed.
1975 Word order and word order change (Austin: University of Texas Press).
1976 Subject and topic (New York: Academic Press).
Longacre, Robert E.
1980 "Discourse typology in relation to language typology", Tentative publication
draft in Preprints, Nobel Symposium on Text Processing, Stockholm.
Ohmann, Richard
1964 "Generative grammars and the concept of literary style", Word 2 0 : 4 2 3 - 4 3 9 .
Reprinted in various readers on stylistics.
Osgood, Charles E. - J. Kathryn Bock
1977 "Salience and sentencing", Sentence production, edited by S. Rosenberg
(Hillsdale, N.J.: L. Erlbaum), 8 0 - 1 4 0 .
Platzack, Christer
1974 Sprket och lsbarheten (Lund: Gleerup).
Rosenberg, Sheldon, ed.
1977 Sentence production (Hillsdale, New Jersey: Lawrence Erlbaum).
Schank, Roger - Robert Abelson
1977 Scripts plans goals and understanding (Hillsdale, N.J.: Lawrence Erlbaum).
Schenkein, Jim, ed.
1978 Studies in the organization of conversational interaction (New York: Aca-
demic Press).
Sigurd, Bengt
1977 "Att presentera Buffalo Bill", Tre experiment med text, edited by B. Sigurd
and G. Kllgren (Stockholm: Akademilitteratur), 2 9 - 7 2 .
CLAUS F.ERCH AND GABRIELE KASPER
Ja und?- og hva s ?-a contrastive discourse

analysis of gambits in German and Danish*
1 Introduction
In designing communicatively oriented syllabuses and teaching materials, one

often lacks precise information about how certain discourse functions can be
realized in L 2 , and what learners' problems are when trying to perform such
functions in their interlanguage.1 We assume that in linguistic areas of this
kind, the function of contrastive analysis (CA) can be twofold:
(1) As CA presupposes descriptions of comparable categories and their means
of realization in both L t and L 2 , it involves an analysis of the target variety,
which is a prerequisite for determining specific features of the learner varieties
(interlanguages).
(2) In interlanguage areas which have not yet been described, CA can help to
generate hypotheses about potential learner problems, which can subsequent-
ly be tested out in experimental interlanguage studies.
The discourse functions we shall concentrate on in this paper relate to the

maintenance and regulation of ongoing discourse. They are typically realized
by linguistic devices which do not constitute speech acts in their own right,
and which have consequently been termed "speech act augments" ("Sprech-
handlungsaugmente") by Rehbein (1979). We shall follow Strevens (1972),
Beneke (1975), Keller -Taba-Warner (1976; 1977; 1979) and Edmondson
(1977a, b) in referring to them as gambits.2 However, as this term has not
been used consistently by these authors, it will be necessary to describe in
some detail what discourse functions and devices we have in mind when we
talk about gambits.
2 Gambits defined
We shall define as gambits linguistic items whose primary function is to main-

tain and regulate discourse. By maintaining discourse is meant the establish-
70 Claus Faerch and Gabriele Kasper
ment of coherence either between turns of speech or within turns of speech.

By regulating discourse is meant the distribution of turns of speech among the
discourse participants in terms of uptaking, turntaking, tumkeeping, turn-
giving, and the marking of discourse boundaries.
In other studies, devices which primarily function on an interpersonal level
have also been classified as gambits (e.g. the "cajoler"inEdmondson 1977a, b,
1981; Edmondson - House 1981;House 1980; Kasper 1979,1981;the "soft-
ener" in Crystal Davy 1975). However, interpersonal (expressive, appealing,
attitudinal) functions can be realized by a great variety of linguistic means,
some of which cannot or do not always carry a discourse maintaining or
regulating function at the same time (cf. the 20 odd categories of "modality
markers" found by House and Kasper 1981). Such devices are therefore not
in agreement with our definition of gambits as suggested above (but see also
below section 4).
Both discourse maintenance and discourse regulation can be achieved by
other means than gambits: coherence is also established through the occur-
rence of adjacency pairs; turngiving can be signalled paralinguistically, etc.
This constitutes a (theoretical) problem for a contrastive analysis, as what is
achieved in one language by means of a linguistic item (a gambit) may be
achieved in another language by different means. However, in languages which
are both linguistically and culturally closely related, like Danish and German,
we assume that this is not the case and hypothesize that one can reasonably
limitate the area of discourse functions by concentrating on gambits and ignor-
ing alternative means of expressing identical functions.
If one distinguishes various levels in discourse structure (e.g. following
Sinclair - Coulthard 1975), one can say that gambits belong to the discourse
level of acts. This means that gambits can be part of moves. However,gambits
cannot in themselves constitute a move, as they cannot develop conversation
(Edmondson 1981:6). As a turn contains at least one move, it follows that
a turn can never consist of a gambit only. According to their level in discourse
structure, gambits can thus be defined as the subclass of acts which do not
contribute to the development of a conversation (see also section 5 below).
3 Coherence and discourse regulation
In order to classify gambits according to their coherence creating and dis-

course regulating functions, these two discourse aspects have to be somewhat
specified.
Coherence should be kept distinct from cohesion:
Gambits in German and Danish: a discourse CA 71
where we can establish a propositional relationship across sentences, without regard to

what illocutionary acts are being performed, by reference to formal syntactic and
semantic signals, then we recognize cohesion [ . . . ] where we recognize that there is a
relationship between the illocutionary acts which propositions, not always overtly linked,
are being used to perform, then we are perceiving the coherence of the discourse (Wid-
dowson 1978: 2 8 - 2 9 ) .
Edmondson refers to cohesion as brought about by "devices by means of

which texture is evidenced in a suprasentential stretch of language", whereas
he refers by coherence to "a 'well-formed' text or discourse" (1981: 5). As
expressed in the terminology adopted in the present paper, coherence is esta-
blished between discourse moves, which may either belong to different
(though contiguous) turns (speaker-hearer coherence: "SH coherence") or to
the same turn (speaker coherence: "S coherence").
Among possible discourse regulating functions, we shall distinguish between
the following:
(a) uptaking - signals to present speaker that the channel is still open and
that the message is being taken in ("I follow you", possibly accompanied
by "I react to what you say in this way")
(b) turntaking signals that speaker wants to take a turn (and possibly serves
the function of plugging a conversational gap) ("I want to say something,
I'm thinking")
(c) tumkeeping signals that present speaker wants to keep the floor, this is
of particular importance at a possible completion point ("I want to say
more, please give me time")
(d) turngiving signals that present speaker wants other speaker to take
a turn ("I've finished saying what I wanted to say, please continue")
(e) marking - signals boundaries in discourse, e.g. between opening and
core phase, between core phase and closing phase, between various dis-
course topics (for this distinction of discourse phases, see Edmondson
House 1981).
4 A typology of gambits
In order to distinguish various gambit classes, we shall use a modified version

of a gambit typology established by Edmondson (1977 b). It comprises the
following types:
Receipt, a signal by which its user indicates that he has taken in what his
interlocutor is or has been saying. It can either have the function of a neutral
attention signal (mm, uhu), or it can express understanding of (I see what you
mean) or even agreement to the interlocutor's speech act ((you're) right,
I agree, that's true). The first two functions are sometimes realized by a repre-
sent, i.e. an utterance in which the speaker repeats to himself a part or the
whole of what has just been communicated to him (SI: "these cuts in the
university budget are a catastrophe"/S2: "catastrophe ... yes . . . yes I sup-
pose that's an adequate way of putting it").
Go-on: A special case of a neutral receipt whereby the present hearer signals
to the present speaker that he is listening and that he wants the present speaker
to keep the turn (oh I see, tyes, tmm).
Exclaim : a special case of a receipt which is marked for an expressive func-
tion, i.e. it indicates its user's reaction to his interlocutor's utterance or a part
thereof. Although the discourse functions of an exclaim are identical to those
of a neutral receipt (see below), we shall treat them as two categories because
in both German and Danish they are realized by different classes of tokens,
i.e. either by a receipt carrying an emphatic marker (\oh, \no), or by elements
which function regularly as "interjections" (good grief, wow).
Starter: as a preliminary to a following utterance, it indicates that its user is
going to say something (well, oh, erm).
Underscorer: serves to focus the interlocutor's attention to a point being made
by the present speaker (the point/thing is).
Aside: functions to inform the interlocutor about what the present speaker is
doing while he is not addressing himself to the present hearer (SI: "can I see
you about those references tomorrow"/S2: "tomorrow ... let me see . . .
I've got appointments all morning . . . how about 2 o'clock").
Appealer : functions to elicit a response from the hearer to what the speaker
has just said (tag questions, uh, okay).
Marker: signals boundaries in discourse in either indicating the completion of
an ongoing discourse topic or phase (okay, good) or the opening of a next dis-
course topic or phase (well now, by the way).
We shall also include into the analysis a number of linguistic items whose
primary function is not that of a gambit as defined above, but which serve as
gambits in a given context, i.e. their 'derived' function is that of a gambit.
Thus hesitators which occur at a possible completion point and have thereby
a floor-holding function will be classified as gambits, whereas hesitation
phenomena which occur at non-completion points within a turn will not be
included. Similarly, cajolers (you know, you see, I mean) will only be classi-
fied as gambits if they are used as turn-keepers; if they function on an inter-
personal level exclusively, they will not be included into the present analysis.
Conjunctions obviously have the primary function of creating cohesion in
texts by expressing how what follows the conjunction relates to what pre-
cedes it. However, by virtue of the fact that sentence conjunctions are nor-
mally placed clause-initially, they may be used with a secondary discourse-

regulating function (e.g. with a turn-keeping function: "then she decided to
stop trying and you know . . . " , cf. also Sacks' category of "utterance
incompletor", referred to in Coulthard 1977: 57). In our analysis, we have
only included conjunctions in those cases in which the discourse-regulating
function overshadows the cohesion-creating function (in particular in cases
where the 'conjunction' has lost its 'normal' semantic function, like Danish
men ('but', see further below)).
The gambit types can now be classified on the basis of their primary dis-
course functions outlined above (cf. table 1).
As can be seen, some gambits are monofunctional(e.g.go-ons and starters),
others are multifunctional (e.g. receipts). Furthermore, some gambits can be
used in a way which is derived from their primary use: thus starters may
occur turn-internally between moves to initiate a new move, with a clear turn-
keeping function. These derived uses of gambits are ignored in the classifica-
tion, but they will be touched upon in the contrastive analysis itself.
Table 1: Classification of gambit types
Regulating function Type of gambit Type of coherence
go-on
Uptaking receipt SH
exclaim
starter
receipt
Turntaking
exclaim SH
(conjunction)
underscorer
aside
Turnkeeping
(cajoler) S
(hesitator)
(conjunction)
Turngiving appealer SH
Marking marker S/SH

5 Some problems in the classification of gambits
We should like to mention three problem areas which the analyst is faced with
in setting up a gambit typology and applying it to data:
(1) how many different gambit classes should be established
(2) if and how the individual gambit classes should be subclassified
(3) how to distinguish a gambit from a speech act.
As for (1), our analysis of 'interruptors' may serve as an illustration. It
would be possible to draw a distinction between the discourse function of
'interrupting' and the discourse function of 'turntaking', and consequently to
establish one class of 'interrupting gambits' and one class of turntaking gam-
bits. The reason why we have not done so are (1) that the two functions seem
to be realized by the same types of gambit; (2) that the difference between
the two functions is not so much a difference in the function itself as in the
context in which the function is used: interruptors can be characterized as
turntaking gambits placed 'inappropriately' at non-completion points in the
interlocutor's turn, their turntaking function becomes particularly clear if the
interlocutor yields the floor and the 'interruptor' is followed by a move. We
would therefore say that turntaking gambits have the additional function of
interrupting if placed at a non-completion point; and that turntaking gambits
'misfire' if the speaker does not succeed in winning a turn.
Problem area (2), the subclassification of gambit classes (cf. the general
discussion of "diminishing returns" in Lyofis 1968:152f), will be illustrated
by the way in which we have classified the gambit classes receipt and exclaim
(cf. section 4). If one chooses as a subclassifying criterion the aspect of the
present speaker's utterance which the present hearer focusses on, one can
distinguish between at least four aspects, which can be paraphrased as
"I'm listening" (mm, uhu)
"I understand what you're saying" (/ see)
"I agree to what you're saying" (that's right/true, sure, I agree)
"I react with anger, surprise, joy, . . . to what you're saying" (really, oh no,
how nice).
The first aspect refers to the communicative channel being open, the second
to the present speaker's message being taken in by the present hearer, and the
third and the fourth to the hearer's attitude and reaction, and to the interloc-
utor's utterance, respectively. As can be seen from the typology of gambits
which we have adopted for the present analysis, we have decided to treat the
last-mentioned subtype as a class of gambits in its own right ('exclaims'),
whereas the first three subtypes are all subsumed under the class of 'receipts'.
From a functional point of view there are no compelling reasons for classifying
in this particular way, our reasons for doing so are simply as stated above
that in the languages we have been concerned with, it seems to be the case that
the tokens used to express reactions to the interlocutor's utterance differ sys-
tematically from the tokens used to realize the other three aspects of 'receipts'.
The third aspect of receipts mentioned above ("I agree to what you're say-
ing") provides a link to problem area (3): On what theoretical and empirical
grounds can gambits be distinguished from speech acts, i.e., acts which carry
illocutionary force? For example, if acts which express agreement can be
classified as gambits, doesn't this hold true for acts which express disagree-
ment as well?
The answer to this can be found in our definition of gambits are constitut-
ing a subclass of acts which do not develop the ongoing discourse. They are
thereby functionally distinct from speech acts which, by virtue of their carry-
ing illocutionary force, do contribute to the progression of the discourse, and
can thus function as moves. Expressing unelicited agreement to the inter-
locutor's utterance clearly does not develop the discourse, as the present
speaker can simply ignore it without violating conversational rules. The ex-
pression of disagreement, on the other hand, is likely to elicit a reaction from
the interlocutor and can potentially steer the discourse in a new direction. In
the empirical analysis, the distinction between, e.g., a receipt and the speech
act agree is often not clear-cut, in particular if the same token can be used for
the realization of both functions (e.g. yes/yah/yeah). One criterion for inter-
preting such a token one or the other way is whether or not it has been elic-
ited: if it has, it is likely to function as a speech act, if it has not, it probably
functions as a gambit.
6 Analysis
As data for our analysis of gambits in German and Danish we have used role
plays, recorded at Ruhr-Universitt Bochum within the "Projekt Kommunika-
tive Kompetenz als realisierbares Lernziel" (cf. Edmondson House Kasper
- Stemmer 1982) and at the University of Copenhagen in connection with a
seminar on discourse analysis.3 The German subjects were university students
at the Ruhr-Universitt Bochum, coming from the Ruhr district. The Danish
subjects were university students at the University of Copenhagen, coming
from various areas in Denmark.
The analysis is based on the German and the Danish data in the sense that
the phenomena described are those contained in the data. However, as is well-
known to everybody working with linguistic corpora, these can rarely provide
all the information needed in order to produce a linguistic description. We
have relied on our native speaker competence in German and Danish, respec-
tively, whenever this was necessary in order to fill in a 'gap' in the data or in
order to establish more precise rules for the occurrence of specific gambit
tokens. The data quoted in the following is authentic unless expressly marked
as 'fabricated data'.
6.1 Uptaking gambits

The following gambit types can be used with an uptaking discourse function:
go-ons, receipts, exclaims. Uptaking gambits can be elicited by the present
speaker (cf. ex. 1), or they are non- elicited, occurring normally at non- com-
pletion points (cf. ex. 2).4
(1) x: Is ne 'schne "Fcherkombina\tion wrde ich "\sagen
tn [y: und das hat dir immer gut ge"letgen
(2) x: und wir 'hatten dann 'vorher noch 'zwei 'Jahre [y: \ja]
Fran"zsisch gethabt5
6.1.1 German data6

Tokens:
receipts: ja, nein, hmm, na ja, ach ja, aha, ach so ; phrases expressing agree-
ment like genau, gut, klar; represents (cf. (1), (2) above and (3)
below)
go-ons: tja, thmm (cf. (4))
exclaims: ach, och, oh, ah; oh + represent; Gottseidank, Mensch, ach du
liebe Gte (cf. (5))
(3) x: wann "isn die \Fete
y: die is heute "\abend - ich "wei \nich [x: ah heute
"\abend] das fngt so um "10 oder so \an das fngt ganz
"spt \an
(4) x: \ja das is ja weite eigentlich "weniger - ich meine wel-
che wolltste denn "ha\ben
y: ja "irgendwelche "guten "Pla\tten - wir wollten zu 'ner
Fite gehen und da "brauchen wir \welche [x: /hmm] da
wollten wir welche "aus\leihen
(5) x: wahrscheinlich ist der der "\Hahn irgendwie - der der
"Gashahn irgend\wie nicht ganz dicht da mu das "Gas
ausgestrmt sein [y: hach du "lieber \Gott] ich meine
ich hab auch nich allzu "groe Ahnung da\von aber h
- es "mu eben daran \liegen
We shall postpone our tentative formulation of the rules governing the distri-
bution of the receipts ja and nein to section 6.2.1, as our data provide more
instances of these receipts in a turntaking than in an uptaking function.
The lexicalized uptaking receipts are typically realized with a falling in-
tonation, whereas the non-lexicalized receiving noise hmm is most often used
with a neutral intonation contour. Uptaking receipts with a rise always func-
tion as go-ons, as the rise in combination with the present hearer's non-claim
of the turn produces an inherently eliciting effect, encouraging the present
speaker to continue his speech activity.
Uptakers can combine in various ways:
the same token is repeated, which sometimes conveys exasperation, com-
pare the repeated go-on in (6):
(6) x: h darf ich dann mal um Ihren "Namen bifiten oder be-
ziehungsweise um Ihren er so' 'nalaus fweis
y: \ja \klar \hier mein Name is Wilfried "Steinfberg steht
ja auf meiner Ent"leihkarfte liegt da bei [Ihnen "tvor]
[x: /ja /ja] und die A"dresse steht tdrauf
different tokens belonging to the same gambit type combine; compare the
two receipts in (7):
(7) x: ja ich meine wenn du nun mal - so'n - guten Sound
"haben willst tn und - ah da is ne Platte - vonm ziem-
lich durchschnittlichen Schallplattenspieler "abgespielt
tworden dann "kriste den eben nich mehr so \hin das is
[y: Vja ja "siVcher] das hngt immer mit der "Nadel
zu\sammen
different classes of uptakers combine; compare the sequence receipt - go-
on in (8):
(8) x: tjah tweite das is'n bichen "schwie\rig - ja ich meine
h ich fahr die immer "na \ab tn und die mssen dann
auch na "abgefahren \werden und [y: hah \so /ja] das
is nich so - ganz "einfach
6.1.2 Danish data

Tokens:
receipts: n, ja, jo, nej, hm, ok, ah;
phrases expressing agreement like det' godt \ represents (cf. (9))
go-ons: hm, jaja(ja), ah, aha, na\ represents; phrases like n for fanden,
h for pokker (cf. (10), (11))
(9) x : . . . og de gik og snakkede om hvor interessant de syntes

det havde vceret og sdan [y: \ j a ] noget jeg toenkte . . .
[x: . . . and they were talking about how interesting they
had found it and so [y: yes] on I thought . . . ]
(10) x: ... jeg s mit snit bade for at det kunne vaere meget
interessant nok [y: / m m ] ogs ogs ...
[: . . . I thought both that it could be very interesting pro-
bably [y: mm] also and then . . . ]
(11) x: . . . s er det sgu da ogs din egen fejl om du afleverer
[y: min egen fejl] den syv dage for sent eller. . .
[x: . . . then it's damned well your own fault that you hand
[y: my own fault] it in seven days too late o r . . . ]
As is the case with the German uptakers ja and nein, the Danish up takers ja,
jo, nej and n are best discussed in connection with a description of their turn-
taking function.
Some of the tokens which can be used as receipts can also have the func-
tion of a go-on, and as in German, the difference between the two types of
gambits is expressed by their intonation: receipts are marked by a level or
falling intonation, go-ons by a rising intonation.
The same types of combinations of uptakers as in German are found in
Danish: combinations of the same tokens, combinations of different tokens
belonging to the same gambit type, and combinations of tokens belonging to
different gambit types (cf. (12)(14.
(12) : hvor kommer du egentlig fra ellers -

y: fra Odense - n rhus universitet ja [: n n n] jeg
laeserfransk
[x: where do you come from actually
y: from Odense well rhus university yes [: I see]
I study French]
(13) x: . . . selvom du visker den ren s blir den alts ik'helt s
poen som den var iforvejen [y: ok-jaja] det vil jeg be'dig
huske.. .
[x: . . . even if you rub out everything it still won't be quite
as clean as it was before [y: yes - ok] IH ask you to
remember . . . ]
(14) x: . . . og s smed han mig lige ud i sneen [y: n for fanden
- og hva' s] og rbte til mig at.. .
[: . . . and then he kicked me right out into the snow [y: oh

no and what then] and shouted to me that . . . ] (fabri-
cated data)
6.1.3 Contrastive analysis of uptaking gambits

The relevant contrasts between various receipt tokens will be discussed in con-
nection with the turntaking receipts in 6.2.3.
On the basis of our corpus, we do not find any evidence for further con-
trasts between uptaking gambits. However, a minor contrast in the category
go-on should be mentioned. In German, the conjunction und, when used with
rising intonation, can function as a go-on, sometimes combined with a preced-
ing/a and/or a following dann or weiter, as in
(15) x: also gestern abend bin ich spt nach "Hause ge\kommen
und will die "Tr ausschlieen und merk ich hab den
"Schlssel \nich - erst dacht ich ich htt'en im "Hand-
schuhfach - war aber nicht [y: ja /und] ja und da mut
ich durchs "Fen\ster was sollt ich "ma\chen
(fabricated data)
While Danish can have og hvad /s [and what then] as a go-on, the conjunc-
tion og cannot on its own be used in this function. A tentative explanation
for this might be that in contradistinction to German und, whose phono-
logical structure (vowel+nasal+dental) provides the necessary basis for carry-
ing rising intonation, Danish og [o] lacks such phonological conditions, being
composed of a single vowel. The supplementation of og by elements such as
hvad s is thus obligatory in Danish, whereas the supplementation of German
und by e.g. (was) dann is optional.
6.2 Turntaking gambits

The relevant gambit types to consider are: starters, receipts, exclaims, in addi-
tion to conjunctions, which are potentially relevant as mentioned above.
6.2.1 German data

Tokens:
starters: ja, also, na (cf. (16))
receipts and same tokens as used for uptaking, see above, section 6.1.1,
exclaims: and cf. (17) - (20) for receipts, (21) for exclaims
(16) x: oh \je wer ist das denn heute \abend (opens door)
y: ja guten Abend Frau Bergmann
(17) x: es ist ja nicht /"so da du die [Schallplatten] jetzt die-
sem Typ da \gibst den du berhaupt nicht "kennst
oder \so
y: Yja das "stimmt \schon \aber...
(18) x: . . . nee \du ich "/glaub dat "klappt nich mehr mit mei-
nem '\Bruder [y: ooh] dat wird 'nich mehr "lau\fen
y: Vnee da mten wir ja noch nach "Essen \fahren
(19) x: ich schaff das so meistens zwischen vier und fnf "Stun-
den
y: ah \so - wie raffst du das denn mit Lingu "fistik also
mir fllt das unheimlich "\schwer
(20) x: Gott ja das kommt eben "\vor - wiege'sagt "Calf dum
y: kommt vor kommt vor - und wenn Patienten sich das
nicht "durchlesen...
(21) x: ich mein ich hob da volles Verstndnis fr deine Situa-
tion ich wr da wahrscheinlich - war da wahrscheinlich
"auch ganz froh wenn ich 'n Referat am andern Morgen
zu "schreiben htte wenn ich dann meine "Ruhe htte
also
y: \Mensch wie "soll ich das nur scha\ffen
In both their uptaking (cf. 6.1.1 above) and their turn taking functions, the
two receipts ja und nein (or nee), cf. (17), (18), seem to be distributed in the
following way in German. After a turn which contains no negation, only ja
can be used as a receipt; if nein is used, it is not a receipt but an exclaim (i. e.
it is emotionally marked). After a turn which contains a negation both ja and
nein can be used, as illustrated by (17) and (18). Even in the latter case, there
may be a tendency to use ja rather than nein, but this cannot be confirmed
on the basis of our corpus, as the occurrence of these items is too limited to
allow for quantitative analyses.
Various combinations of these tokens are possible, both within the same
class (ja also) and across the classes (Gott ja). One frequently occurring com-
bination is receipt + starter, cf. the following example:
(22) x: . . . wenn das bei Ihnen nicht "\klappt dann knnen Sie
mich ja mal "an\rufen dann wrd ich die "fra\gen
y: /ja - Vja ich mte mal "\schauen ...
6.2.2 Danish data

Tokens:
starters: ja, jamen, alts, n (cf. (23), (24))
receipts and same tokens as used for uptaking, see above, section 6.1.2,
exclaims: and cf. (25), (26) for receipts, (27) for exclaims
conjunctions: (cf. (28))
(23) x: tror du du du kan finde ud afdet
y: ja det skal jeg nok finde ud af
: ja for jeg vil jo gerne ikk'
[x: do you think you can manage
y: oh sure I can manage
x: because I do want to babysit you know]
(24) x: det er da bed(f>vende ligegyldigt
y: min egen fejl alt det vi har haft her p - eeer -
x: jam hvadfik du at vide dengang du ...
[: it doesn't matter at all
y: my own fault! all the things we've had to . . .
x: but what were you told when . . . ]
(25) x: vi kan snakke lidt sammen
y: ja jeg har alts fet et problem
[: let's talk together a bit
y: yes - I've got a bit of a problem]
(26) x: det viser jo bate at jeg har brugt den ikk' alts
y: jo men bibliotekets b<f>ger ererm aldeles ikk' til at skrive i
x: nej det ve 'jeg godt ik' og. ..
[x: it only shows that I've used it doesn't it
y: yes but library books are not for writing in
x: no of course not]
(27) x: jeg er bare alts jeg er bare skuffet over det
y: skuffet jam det er jo - det er jo...
[x: it's just that you see it's just that I'm disappointed
y: disappointed - but that's - t h a t ' s . . . ]
(28) x: . . . det er simpelthen bare for drligt synes jeg og det
bliver jeg bare ked af
y: men jeg kan godt se at du er ked afdet men eer...
[x: it simply is too bad I think and it makes me feel un-
happy
y: well I do understand that it makes you feel unhappy
b u t . . .]
The tokens n, ja, jo, nej are all used as receipts, both with an uptaking and
with a turn taking function. Of these, the occurrence of ja, jo, and nej is re-
stricted by the proposition^ content of both the preceding turn and - in the
case of the receipts being used with a turntaking function - the move they
initiate, whereas n is neutral and signals "I've heard what you say". This ac-
counts for the fact that n, when used in a situation in which the interlocutor
expects agreement, can be felt as less polite than ja, jo, nej (cf. (29)). 7
(29) x: jeg kommer p naeste fredag

y: n jamen det ergodt
[x: 111 be coming next Friday
y: yeah well that's fine]
N can also be used as a turntaker before a move expressing disagreement, in

order to downtone the disagreement. In this case, n is obviously a more
polite way of initiating the move than by using an element that expresses the
speech act directly (cf. (30)).
(30) x: . . . du kan da altidg derover alts det'da ikke

y: n ja men vi har arbejdet. ..
[x: you can always go over there, can't you, it's not
y: yes but we worked . . . ]
As can be seen from this example, n can combine with the receipts ja, jo, nej,
in which way a following disagreement is more strongly downtoned. Such
combinations of n with another receipt are very frequent in the data.
That AW is a more neutral receipt than ja, jo, nej does not mean that there
is more or less a free choice between n and ja, jo, nej: there are situations in
which n, and not ja, jo, nej can be used, and vice versa. Before we can de-
scribe these restrictions, however, we have to introduce the distinction be-
tween l X-propositions', -propositions', and 'X-Y-propositions': 'X-propo-
sitions' are propositions which X (i.e. the speaker of the first turn) only can
have knowledge about, -propositions' are propositions which Y (i.e. the
addressee of X and the speaker of the second turn) only can have knowledge
about, and 'X-Y-propositions' are propositions which both X and Y can have
knowledge about (cf. the "speaker knows best principle", Foreman 1974).
It is possible to formulate the rule for Danish receipts that ja, jo and nej
can only be used if either the preceding turn or the subsequent move contains
a Y or an X-Y-proposition: if both turns contain X-propositions, only n can
be used (cf. (31)):
(31) : jeg er traet (X-proposition)

y: | n J du er troet (X-proposition)
n l
[x: I'm tired
y: uhum you're tired] (fabricated data)
If the first turn contains an X-proposition and the following move contains
a Y-proposition, both n and jaljo/nej can be used (cf. (32)). 8
(32) x . jeg er troet (X-proposition)
y: (n) deter jeg ogs (Y-proposition)
UaJ
[x: I'm tired
y: uhum so am I] (fabricated data)
If the first turn contains a Y-proposition, only ja, jo, nej can be used as turn-
taking receipts in the subsequent turn (cf. (33)). This may be explained by
the implicitly eliciting function of the first turn, which means that ja/jo/nej
are here borderline cases between gambits proper and speech act realizations
(and n cannot function as a speech act realization).
(33) x: du er trait (Y-proposition)
y: ( *n) jeg er traet (Y-proposition)
ja
or
f * I det er du ogs (X-proposition)
ja I
[: you're tired
y: yes I'm tired or yes so are you] (fabricated data)
With X-Y-propositions, both n and ja/jojnej can be used, but with the differ-
ence that n marks that the hearer does not know, jajjo/nej that the hearer
does know the propositional content of the preceding turn (cf. (34)).
(34) x: Peter hadergule certer (X-Y-proposition)
y: J n 1 g<f>r han det (X-proposition)
*ja
or
*n| det g<j>r han (Y-proposition)
ja J
[x: Peter hates pea soup
y: uhum does he or yes he does]
In the preceding we have referred to ja/jo/nej as constituting one group of
receipts. Let us now consider the restrictions holding for the choice of each
of these. The choice between ja and nej is considerably determined by the

propositional content of the preceding utterance, ja primarily following a
positive proposition and nej necessarily following a negative proposition (un-
less nej is used as an exclaim, cf. what is said about nein above). Jo can be
used as a receipt if the preceding turn contains a negative appealer (usually
ikke, cf. (26)), but ja can also be used (in which case the receipt ties up with
the (positive) proposition, not with the (negative) appealer). The ways the
three receipts can be used in Danish are presented in table 2.
Table 2: Occurrence of , jo, nej as receipts
turn y turn
positive proposition ja
negative proposition nej (ja)
positive proposition . ,. .
+ negative appealer
negative proposition
+ positive appealer
Finally a few words about the conjunction men 'but'. As can be seen from
example (28), it can function as a completely neutral 'starter' without any ad-
versative semantic meaning. Examples of this in the data are rare, however
what one finds more often is men used with some of its original semantic
meaning preserved, and used in combination with gambits proper: men alts,
ja men, na men. Particularly common is the combination receipt + men,
which often has the pragmatic function of downtoning disagreement, contrast,
opposition or the like: "I don't think I can do it!" - "no but don't you think
you could at least try?". Of these combinations, ja + men is frequently used
as a single gambit {jamen/jam), which we have classified as a starter proper
(cf. (24) above).
6.2.3 Contrastive analysis of turntaking gambits

As can be seen from the descriptions, there are strong similarities between the
German and the Danish data, not only with respect to the types of gambit
used with a turntaking function, but also with respect to the tokens used as
realizations of the individual gambit classes. However, significant contrasts can

be found within the class of receipts. First, while German operates with a sub-
system of two semantically opposite elemens, ja and nein, the corresponding
Danish subsystem comprises three elements, ja, jo and nej. German nein seems
to be more restricted in usage than the equivalent Danish nej. Danish jo,
which is used both as a proper act with an illocutionary force of its own ('dis-
agree') and as a gambit ('receipt') corresponds to German doch in its speech
act function only: as a receipt, the equivalent German gambit is ja. These con-
trasts can be illustrated as in table 3 (for comparison's sake, we also include
the responding acts agree/disagree in the table):
Table 3: Contrastive analysis of ja, jo, nej/ja, doch, nein
Gambit: receipt Responding act: agree/disagree
Danish: German: Danish: German:
ia - ja ja (agree)
. } (disagree)
nej nein nej nein J v '
The second area of contrast is the relationship between the multifunctional

Danish receipt n and its German equivalents. There seem to be three German
tokens which correspond to three different functions of n respectively, ja,
ah/ach so, naja.
(1) German ja can be used as a receipt both if it follows or precedes a Y-pro-
position or an X-Y-proposition, and if it follows an X-proposition, thereby
merely signalling 'understanding', but not 'knowing' of X's preceding utter-
ance. In the former case, it corresponds to Danish ja or n only, compare
(35)(38).
(35) x: man das war spannend
y: ja fand ich auch
(36) : det var spoendende
y: ja det synes jeg ogs
(37) x: man das war spannend
y: ja fandst du auch
(38) x: det var spoendende
y: n det synes du ogs
(all fabricated data)
(2) Unlike ja, the token ah/ach so is not ambiguous between 'knowing' and
'understanding', but expresses the latter exclusively: it is used as a receipt
following X-propositions only and thus clearly corresponds to n in this func-
tion, cf. (39).
(39) x: aber ich kenn mich hier ' bichen besser "/aus da - ich
hier "wohfne [y: ah \so] und h auch jetzt wohn ich
hier bei meinen "Eltern und fahr dann tglich "hier\hin
The distribution of Danish ja and n vs German ja and ah/ach so can be
schematized as in table 4.
Table 4: Contrastive analysis of n, ja/ ja, ach so
X-proposition X-Y-proposition Y-proposition
Danish n n n
ja ja
German ja ja ja
ah/ach so
(3) The German equivalent for n as expressing tentative acceptance or as

preceding indirect disagreement (cf. (30) above) is naja, as in (40):
(40) x: ich mein wenn das wenn de mir das Versprechen "nicht
ge/geben httest htt ich mich darum "selber be\miiht
tn "tdann wr die Sache er"le\digt gewesen - aber
"Ho ich mein ich hab mich fest drauf ver"la\ssen
y: na Vja also - knn wer nichts "ma\chen tut mir "\leid
Contrarily to Danish n, which can precede and thus downtone the 'force' of
the elements ja, jo and nej, German na can precede ja only (*nanein).
Apart from the receipts, which constitute the main area of contrast, it
should be mentioned that the Danish conjunction men as a semantically
neutral starter has no formal equivalent in German, as after always retains its
primary meaning as an adversative conjunction. In most cases, the functional
equivalent for men will be ja or another token by which starters can be real-
ized in German.
6.3 Tumkeeping gambits
Whereas the turn-initial position is central to the occurrence of turntaking

gambits, it is obviously marginal to the turnkeeping gambits, as their primary
function is to signal to the hearer that the speaker intends to continue past a
possible completion point. Turnkeepers typically occur turn-internally;how-
ever, they are occasionally also used turn-initially, following a turntaker (cf.
(44) below). The relevant gambit classes are: underscores, asides, cajolers and
hesitators, the latter two classes having a derived turnkeeping function (see
section 4 above).
6.3.1 German data

Tokens:
underscorers: phrases like hr mal/hrnse mal, sieh mal/sehnse mal, guck
mal, warte mal, weit du was, Augenblick, Moment (mal),
halt (cf. (41))
asides: cf. (42)
cajolers: ich mein(e), weit du, verstehst du, komm, ehrlich (cf.
(43), (44))
hesitators: cf. (45)
(41) x: aber "trotzdem hr mal ich hab das "frher \mal - da

war ich noch "nich \so da hab ich "auch immer meine
Platten aus\geliehen und von fiinfen hab ich dann mei-
stens nur drei zu"rckge\kriegt
(42) x: (filling in a French form for y): und du hast immer sehr
ordentliche "Leistungen getzeigt und ich glaub das das
kannst du "auch \schaffen \ja so dann wollen we mal
das heutige "Datum auch ein/tragen signature "/date
heute ham we den "/Via den wievielten "ham we denn
\heute
y: zweiter Au'%gust
(43) x: ja Gott der "E\ggert der war heute schlechter als "\sonst
aber ich "meine - der Lippens is "\gut der Lippens "is
\gut und wenn man den nur "foulen kann dann "foult
man den \eben ich meine der Eggert hat getan was er
"konn\te
(44) x: . . . das ist dann "sch\ner als wenn man vom Anfang
gleich so iso "liert und da rum\geht ich
y: \ja also ich mein - ich hab da'n bichen "mehrErfah-
rung . ..
(45) x: ich meine ich hab auch nicht allzu "groe Ahnung \da-
von h aber es mu eben daran \liegen - am besten wir
rufen sofort "\an
Sequences of turnkeeping gambits occur frequently in the German data, com-
prising both elements of the same (cf. (46)) and of different classes (cf. (47)).
(46) x: ["stt dann einer Mran und so] dann wird'n "Bierglas
bern Apparat gekippt
y: ja "\komm ich \mein es gibt ja die Appa " Trte wo man
da 'n "Deckel drber \macht tn
(47) x: ja "tnun aber ich /mein ku\mmal Vhm ich mein wir
"wohnen ja jetzt hier wohl doch falle zwar mehr oder
weniger zu"\sammen aber "immer\hin tn ich mein du
kennst mich ja doch jetzt eigentlich schon etwas \lnger
tn
Moreover, receipts and starters are sometimes used turn-internally, thereby as-
suming a 'secondary' turnkeeping function (cf. (48)). A special case of a
turn-internal receipt is the 'self-receipt' as illustrated in (49), whereby the
present speaker 'takes up' her own preceding speech activity. Thus, discourse
functions which are normally realized in a dialogue structure (interlocutor
taking up the speech activity of interlocutor A) are transformed into a mono-
logue, by means of which the present speaker makes sure that he keeps his turn.
(48) x: na ja is \klar - na - ich mcht ja auch nich eure "Fite
deswegen - aufn Hund kommen \lassen aber - nun na
Yja also hm - ganz alte knnt ich dir "mittgeben aber
dann mt ich wirklich auch selbst "mit\kommen tn
(49) x: . . . und "langweilt man sich zu \Tode dann geht man
schon gar nicht mehr in die Vorlesung "\hin
y: ja ja "\eben das ich "mei\ne das is mir schon zweimal
pa"\ssiert aber bei dem wollte ichs "nicht grad \machen
das isn bichen "eigenartig mmh Vja - he ihr trinkt hier
"\Bier kannste mir "auch 'n Bier tgeben
6.3.2 Danish data

Tokens:
underscores: phrases like det vil jeg sige 'this is what I want to say',
(men) hva' skal jeg sige '(but) what can I say', men hva' 'but
what', ved du hvad *you know what', h<j>r lige 'just listen'
(cf. (50)).
asides: cf. (51)
cajolers: du ved you know', jeg mener mean' (cf. (52), (53))
hesitators: cf. (54)
(50) x: det var nok mulig men hva' er - - jeg syns der er
sdan lidt - kedeligt her syns du ikke [y: jo] til festen...
[: that was possible I suppose but what er I find it a
little bit boring here don't you think [y: yes] at the
party . . . ]
(51) x: n og s er det fru Hansen - lad mig lige se [looks into
her file] - ja De var her sidst. ..
[x: well so now it's your turn Mrs. H. let me just see [ . . . ]
ok you were here l a s t . . . ] (fabricated data)
(52) x: . . . jeg har fet ny ven s du ved jeg vil jo jeg vil jo
osse gerne med ham ik'
[x: . . . I've got a new friend so - you know I also want to
go out with him of course]
(53) x: hvad tror du for (frvrigt mine chancer er for at eer det
overhovedet gr igennem alts jeg mener det er jo for
sent ik'
[x: by the way how good a chance do I stand do you think
that er it will work at all well I mean it is too late isn't
it-]
(54) x: mm - har du snakket med nogen andre alts erm - og
hva'de havde syns om det
[x: mm - did you talk to some of the others - erm to
hear how they like it ]
The Danish data happen not to contain sequences of either different tokens
belonging to the same gambit class or tokens belonging to different gambit
classes (cf. the description of sequences in the German data above, section
6.3.1). This is no doubt completely accidental, and the following (fabricated)
example would be a perfectly possible turn in Danish:
(55) x: . . . men erm alts sdan en bog der alts ved du hva' -
det vil jeg sige dig - den kan du alts ikke aflevere
[x: . . . but erm such a book - well you know what - I can
assure you you can't possibly return it] (fabricated
data)
(55) can also be used to illustrate the occurrence of a starter (altsa) turn-inter-
nally, with a secondary turnkeeping function (cf. also (53), (54)). Examples
of 'self-receipts' (cf. section 6.3.1) are frequent in the Danish data:
(56) x: . . . tror du ikke de har lukket her om aftenen ja det kan

jeg godt sige dig de har. ..
[x: . . . don't you think they are closed here at night yes I
can assure you they are . . . ]
6.3.3 Contrastive analysis of turnkeeping gambits

Most of the gambits used with a turnkeeping function are phrase-like tokens,
and apart from the obvious lexical differences between underscores, asides
and cajolers in German and Danish, no significant differences were found in
the functions of formally comparable tokens. It is theoretically possible that
contrasts of a quantitative type exist between the two languages, but the data
have not been subjected to quantitative analyses.
6.4 Turngiving gambits

The turngiving function in both German and Danish is realized by appealers
exclusively. However, not every occurrence of an appealer is indicative of
turngiving: speakers might use an appealer in order to elicit an uptake (irre-
spective of whether or not the elicitation is successful) during their turn, i.e.
without inviting their interlocutor to take over. Appealers thus function as
turngivers in turnfinal, but not in turn-internal position (cf. (57)).
(57) x: so ne halbe Stunde "rumgegurkt da /ne der einzige der

bei "euch wat gebracht hat dat war vielleicht der "Lip-
\pens /ne - aber der Eggert is halt im Moment - nich
so "\stark /ne mu vielleicht viel studiern (laughs) - ich
"wei \nich
Appealers are primarily used with declarative sentences, as these are not inher-
ently turn-giving: the appealer adds an eliciting function to an utterance which
does not in itself invite the interlocutor to react. With interrogative and
imperative sentence types, the need for using appealers is obviously very
limited as the utterance of these sentence types themselves has a direct elicit-
ing function if an appealer is added in such cases (which, incidentally, is
found in neither the German nor the Danish data), the function is to reinforce
the inherently eliciting element in the utterance ("kommst du mit - /ja").
6.4.1 German data

Appealers are realized by the following tokens: ne/m/nich, ja, oder, oder was,
nich wahr, was, hmm, okay, cf (58) and (59):
(58) x: ja wissense denn jemand ne "Freundin oder so der -
jemand der jetzt einsteigen tkann so r Sie nur ein-
"tmal sonst wrden Sie ja "\kommen /n
y: \ja sonst wrd ich "\kommen aber das - hat sich gerade
so er"ge\ben
(59) x: ja ich werd mich drum be"\mhen "wirk\lich aber - das

eine Mal knnen wir ja noch mal davon "ab\sehen / j a
y: \gut
Usually, a rising intonation contour is obligatory with appealers in German.

This rule becomes optional, however, if the preceding speech act has in itself
an eliciting effect. Thus:
(60) x: du kommst doch "mit \oder (fabricated data)
As concerns the distribution of ne/njnich, ja and oder, our informants show

a strong preference for the first group of tokens. This preference seems to be
governed by the constellation of the following parameters:
(a) what illocutionary function the appeal has,
(b) whether the preceding proposition is positive or negative
(c) the speaker's presuppositions vis--vis the proposition and the hearer's
reaction.
As for the first parameter, we shall distinguish between three types of ap-
peal: appeal for information, appeal for agreement, and appeal for acknow-
ledgement:
appeal for information , the speaker does not know (although he may have
some assumption about P), and appeals to the hearer for information about
("Peter kommt nicht toder")
appeal for agreement: The speaker infers P, wants the hearer to express agree-
ment to
("das ist ein furchtbar bldes Buch ?ne")
appeal for acknowledgement: The speakers knows P, wants the hearer to ex-
press acknowledgement (or understanding, appreciation,...) of this
("Ich hob nicht viel geschafft weil ich die ganze Woche krank gewesen bin
\ne")
According to these appeal functions, to whether the preceding utterance
has a negative or a positive proposition, and to the speaker's presupposition,
the rules for the distribution of ne/m/nich, ja and oder can roughly be stated
as follows:
ne/m/nich can be used in all three appealing functions, following both a posi-
tive or a negative proposition. Thus
- appeal for information:
"du hast den Film schon gesehen tne"
"du hast den Film noch nicht gesehen tne'''
- appeal for agreement:
"das ist 'ne ziemlich langweilige Vorlesung hier tne'''
"das ist keine besonders spannende Vorlesung hier tne"
- appeal for acknowledgement:
"derSearle verwechselt eben Sprechakte und illokutive Verben \ne"
"der Searle unterscheidet eben nicht konsequent zwischen Sprechakten
und illokutiven Verben \ne"
In contradistinction to the 'universal' applicability of the negative appeal-
ers, the use of ja seems to be restricted to
- appeals for information, following a positive proposition:
"du hast den Film schon gesehen tja"
- appeals for acknowledgement, following a positive proposition:
"der Searle verwechselt eben Sprechakte und illokutive Verben \ja"
Whereas in these cases the low acceptability of the sequence negative pro-
position + positive appealer might be due to grammtical co-occurrence restric-
tions holding in German, the restriction on the realization of an appeal for
agreement by the token ja is pragmatically determined: the use of ja as an
appealer carries a strong bias on the part of the speaker that the state of af-
fairs he refers to in the proposition is true, and that his interlocutor's response
wl be in accordance with this presupposition. If one appeals for agreement,
however, such a bias is interactionally weird, as the speaker then expresses
himself what he in fact had intended to elicit from his interlocutor. It seems
to be precisely this bias or presupposition in the use of ja which makes our
subjects prefer ne over ja where both tokens are possible, as ne is neutral in
terms of the speaker's presuppositions.
The token oder can be used interchangeably with ne in both positive and
negative appeals for information and agreement. It cannot, however, be used
as an appeal for acknowledgement, as it is open for a potential hearer's re-
sponse which expresses disagreement.
From what has been said above, one can tentatively suggest that there is a
cline in terms of the speaker's presupposition holding for the proposition of
his utterance, and the expected hearer response: oder has the lowest speaker
bias and leaves most room for the hearer's disagreement, ne/m/nich is neutral
in both these respects, and ja implies that the hearer shares the speaker's pre-
suppositions and reacts accordingly.
Finally, the restrictions holding for the use of okay should briefly be
stated: okay presupposes that the speaker and hearer have reached a common
decision, opinion or agreement on a state of affairs; cf. (61).
(61) x: ja pa "Xauf ich geb dir dann die "drei vonn Beatles
und "eine vonn Stones \mit o/kay
y: fgut herzlichen \Dank
6.4.2 Danish data

Appealers are realized by the following tokens: ikkejvel, hva', okay, or by
phrases like synes du ikke/vel 'don't you think', ikke ogs (cf. (62)(64)).
(62) x: du kommer ikke til festen i aften /vel

y: jo mske, det kommer an p ...
[: you're not coming to the party tonight are you
y: well perhaps that depends...]
(63) x: du er ogs blevet sulten /hva'
y: ja - jeg skal jo have lidt her...
[x: so you also got hungry did you
y: yes I certainly need something here . . . ]
(64) x: . . . jeg har knoklet dagen lang - for at tjene enge til
resten of ret /ikke
y: det g<f>r jeg da ogs
: hm
[: . . . I've been slaving all day - to save money for the rest
of the year haven't I
y: so have I
x: hm]
The appealers ikke/vel clearly belong to one and the same group, the only dif-
ference between them being that ikke follows a positive proposition, vel a
negative proposition (as can be seen from examples (62), (64)).
The distribution of ikke/vel and of hva' is only partially identical, there be-
ing situations in which either of the two subcategories is unacceptable.
With appeals for information (cf. the discussion in section 6.4.1 above),
ikke/vel only seem to be possible (cf. (62)):
94 Claus Faerch and Gabriele Kaspei
Peter kommer ikke I tvel I

Vthva]
[Peter isn't coming is he]
Peter kommer alligevel \ tikke I

(*?hva j
[Peter is coming after all isn't he]
That hva' is not possible in these examples should not be taken to mean that
it cannot co-occur with the type of sentence exemplified above but that the
reading of the preceding sentence is different if hva', and not ikke/vel is
added. With ikke/vel added, the utterance is used in a situation in which the
speaker is not sure whether Peter is coming and where he presupposes that the
hearer will know hence the appeal is an appeal for information. With hva'
added, the utterance implies that the speaker has reasons to believe that Peter
is coming and that he wants the hearer to agree to this. Hence hva' is used as
an appeal for agreement, something which is clearly brought out by (64): the
proposition contained in X's turn is a Y-proposition (cf. section 6.2.2 above),
the speaker assumes that Y is hungry and requests Y to confirm this assump-
tion. In such cases - in which the speaker could not have direct knowledge
about the state of affairs which his utterance refers to - only hva' can be
used as an appealer. If the appeal for agreement relates to a propositional con-
tent which could logically be a X-Y-proposition, both hva' and ikke/vel are
possible (with the difference that hva' more clearly than ikke/vel expresses
that the proposition is tentative, based on inferring from something else).
The final type of appeal appeal for acknowledgement is inevitably ex-
pressed by ikke/vel in Danish, as shown in example (65). In this case the speaker
states the proposition as something he knows/believes, and simply appeals for
the interlocutor to communicate that he has taken in this information. One
could perhaps very tentatively say that ikke/vel are primarily used in connec-
tion with factual information, either information which the speaker requests
the hearer to provide (appeal for information) or information which the
speaker already possesses and which he wants the hearer to agree to or to
acknowledge the existence of (appeal for agreement and appeal for acknow-
ledgement, respectively), whereas hva' is predominantly used in connection
with non-factual information, in particular in situations in which the speaker
can only make assumptions about the state of affairs and appeals to the inter-
locutor for confirmation (appeal for agreement). We can summarize this as
presented in table 5.
Table 5: Occurrence of appealers in Danish
type of appeal X-Y specification
appeal for information ikke/vel X does not know P, expects Y to know
hva' X assumes/believes/experiences , expects Y

to confirm
appeal for agreement
ikke/vel X believes/knows P, expects Y to agree to
appeal for acknow- ikke/vel X believes/knows P, requests Y to acknow-

ledgement ledge that X believes/knows
6.4.3 Contrastive analysis of turngiving gambits

There are major contrasts between the tokens by which appealers are realized
in German and Danish, as can be seen from tables 6 and 7 below.
Table 6: German appealers
appeal for appeal for appeal for

information agreement acknowledgement
positive ja ne/n/ oder ne/n/ oder ja ne/n/

proposition nich nich nich
negative ne/n/ oder ne/n/ oder ne/n/

proposition nich nich nich
speaker pre- + +
+ - + - + -
supposition
("+, , " refer to the strength of the speaker presupposition as discussed in 6.4.1)
As indicated by tables 6 and 7, Danish has no direct equivalent to the 'uni-

versal' German appealer ne/n/nich. The formal equivalent for these negative
appealers, ikke, is more restricted in its range of applicability, as it can only
be used after a positive proposition. Further, in the case of the appealers for
information and for acknowledgement, Danish does not differentiate between
Table 7: Danish appealers
appeal for appeal for appeal for

information agreement acknowledgement
positive
ikke ikke hva' ikke
proposition
negative
vel vel hva' vel
proposition
speaker pre-
supposition + - + + - + -
("+, , " refer to the strength of the speaker presupposition as discussed in 6.4.1)
various speaker presuppositions. This distinction is only made in the case of

appealers for agreement, where both German and Danish distinguish between
a neutral speaker presupposition and one which has a low bias towards the
speaker's own belief.
6.5 Markers
Markers have the discourse function of signalling boundaries in discourse be-
tween discourse units. They may occur both turn-initially, turn-medially, and
turn-finally, as can be seen from examples (65), (67), and (72).
Whereas many of the gambit types discussed so far functioned as coherence
establishing devices (cf. table 1 above), markers serve to create cohesion be-
tween various phases of a discourse (cf. our distinction between coherence
and cohesion in section 3 above). In referring back to a preceding discourse
phase and marking its termination, they provide an anaphoric link, and in
marking the opening of a new discourse phase, they function as cataphoric
linking devices.
6.5.1 German data

The following tokens are used as markers in an anaphoric function: gut, okay,
naja, (alles) klar, schn, often preceded by ja, (cf. (65)(67)). Cataphoric
markers are realized by brigens, so, and starters like ja, also, h, usually car-
rying a stress marker (cf. (68)).
(65) : na hrt sich ja alles ganz schn und gut "\an ich mein -
"\gut - wenn ich Ihnen also so weit "glauben Marfwas
machense denn film "Vorschlag wo kann man denn
"hin\gehn
(66) x: na \ja das wird in der Dunkelheit ja 'ne schne Qule"\rei

sein 'n "Bauern\hof da - ber den "FeldStweg zu finden
aber es wird einem ja nichts andres "brig \bleiben
y: ja na"tr\lich s es 's is auch fir 'ne "\Frau 'ne Qulerei
da h "rumzuMiegen
x: ja \ja immer diese "\Spitzen ja o\kay dann hau ich
jetzt ab
(67) x: ja ich will diesmal noch drber hin "weg\sehn wenn das
in "Ordnunggebracht \wird tja
y: das das "Bade\zimmer
x: ja das "Bade\zimmer
y: ja das Badezimmer soll nich mehr "vorkommen "\nein
x: \gut dann is es in "Ord\nung
y: \gut - dann ja - schn dann freu ich mich jetzt auf mei-
nen "Ka\ffee
(68) x: wir ham heute "a\bend wolln wir wohl auf so'ne "Fete
\gehn von som Freund vom "\Hans - ich wei "auch
\nicht ich kenn den "auch nicht \unbedingt tn [y:
/hmm] soll wohl ganz "nett tsein soll ganz "gut Ver-
den - ach "bVrigens Yja - fllt mir nmlich "auch \ein
h (.. .) du hast sollst doch auch irgendwie unheimlich
klasse "Pla\tten so "weit schon
As can be seen from the examples, both co-occurrence of different tokens (cf.
(67) and combination of markers with other gambits, e.g. starters (cf. (68)),
are possible. There seem to be no selectional restrictions operating on the ana-
phoric markers: they can be used in order to terminate the opening phase of
a discourse, various topics within the core phase or the core phase as a whole,
i.e. before the opening of the closing phase.
With cataphoric markers, the situation is slightly more complicated: starters
in a marking function and the token so can be used as initiators to the core
phase and new topics within the core phase as well as to the closing phase.
Moreover, they can even be used at the beginning of the opening phase, thus
marking the transition from a non-contact to a contact situation as in (69):
(69) (x entering a room)

: "so Frau Bdinger - ich bin "fertig hab meine "Sachen
alle gepackt und jetzt wollt ich mich noch von Ihnen
ver"abschieden
Co-occurrence with opening signals (greetings) is also possible, cf. (70), (71):
(70) x: tna - gr \dich

y: ha\llo wie "\gehts
(71) : "Yja Tach "Wal\ter - na wie "gehts \dirdenn
y: ach hallo "\Franz - "gr \dich - och wie soils mir
"\gehn
The token brigens, however, has a far more restricted range of applicability:
it can only be used for marking the transition from the opening phase to the
core phase (cf. (68)) or the introduction of a new discourse topic, but it can
neither mark the beginning of the opening nor of the closing phase.
6.5.2 Danish data

The following tokens are used as markers in an anaphoric function: godt,
okay, fint (cf. (72)). As cataphoric markers are used/or resten, for <j>vrigt, and
starters like jamen, alts (cf. (72), (73)).
(72) x: ... s kan vi ogs sidde derhenne ifred og ro og

y: ja og snakke og [x: ja] s ' lidt andet [: ja] okay skal
vi ikke g(f>re det
: jo
y: det er fint
x: jamen det er moegtigt fint [y: ja] s okay
[: . . . and then we can also sit there peacefully and
y: yes and talk and [x: yes] so [x: yes] okay why don't
we do that
x: yes
y: that's great
x: okay that's really great [y: yes] so okay]
(73) x: ja men ellers skal jeg nok gd op og fortoelle at at det er
min skyld og jeg aldrig skal g<^re det mere
y: hvad tror du for ^vrigt mine chancer er for at eer det
overhovedet gr igennem .. .
[x: okay and otherwise I'll certainly go and tell them that
it's my fault and that I shall never do it again
y: by the way how good a chance do I stand do you think

that er it will work at all. . . ]
Example (72) illustrates how a series of markers can be used by both inter-
locutors to signal that the conversation is coming to a stop. The same example
also illustrates how adjectives like fint 'great' and godt 'good' are used in such
a sequence with a marking function, summarizing the speaker's attitude to the
content of the preceding discourse. There are no compelling reasons why det
er fint 'that's great' should not be considered a discourse marker; however, we
have decided not to do so in the present analysis, where only the short forms
fint, godt etc. are treated as gambits.
The Danish data contain a few examples of underscores with a clear mark-
ing function; these are obviously cataphoric, and typically combine with an-
other marker (cf. (74)).
(74) x: . . . du siger selv der er lukket - p eksamenskontoret -
y: ja det er selvf^lgelig rigtigt -
: ved du hvad alts - okay - lad os se lidt cool p det her
ikk'du laver.. .
[x: . . . you say yourself the examination registration office
is closed
y: yes that's true of course
x: do you know what okay let's play it cool you make . . . ]
6.5.3 Contrastive analysis of markers

Apart from the obvious formal differences between the tokens that can be
used with a marking function in German and Danish, no contrasts were esta-
blished in the analysis. Whether a more extensive analysis than the present
would bring out such contrasts is difficult to tell, but the total number of
markers in the data we have analysed is admittedly very small.
7 Conclusion
In the preceding section of the article we described the use of gambits with
uptaking, turntaking, turnkeeping, turngiving and marking functions in
German and Danish. Although this description is necessarily limited in various
respects (see further below), it should provide useful information about an
area of discourse which, with one notable exception, namely a CA of gambits
in German and English carried out by House (1980), has not previously been
subjected to cross-linguistic analysis. As mentioned in the introduction, we

consider this one important contribution of CA.
The individual descriptions of the five discourse functions in German and
Danish, respectively, were compared and contrasts pointed out. One striking
result of this is that in most cases there seems to exist functional equivalence
between formally related gambit tokens. This result could be interpreted as
an indication that the area of gambits is not likely to lead to problems for the
Danish learner of German or the German learner of Danish. For two reasons,
this would be an overhasty conclusion to draw. First, it is by now a well-esta-
blished fact that the extent of learning problems cannot simply be assessed by
measuring out the amount of contrast holding between the learner's Lj and
the L 2 : learners do not necessarily transfer because of similarity between L t
and L 2 ; rather, the degree to which they transfer depends on their assumptions
about the transferability of the linguistic item in question (cf. Kellerman 1978
for a comprehensive discussion of this point). That learners do perhaps not
experience the way in which pragmatic and discourse functions are realized as
something which is highly transferable is evidenced by a study of German
learners of English, in which the learners in certain cases did not utilize the
possibility for positive transfer of pragmatic and discourse elements but chose
to reduce instead (Kasper 1981).
Second, the CA carried out above did reveal a number of contrasts between
German and Danish. Following the line of argumentation just suggested, the
existence of such contrasts need not represent particularly difficult learning
problems: it is conceivably almost as difficult to learn when to transfer as to
learn when not to. What makes the situation potentially difficult for the
learner is that all of the gambit tokens in German and Danish have formal equi-
valents in the other language - and most of the functions of these tokens are
apparently identical. This means that //learners transfer, there is a risk that
they will use the gambit tokens in the L 2 in the same way as their formally
related tokens can be used in the Lj. We therefore hypothesize that the con-
trasts we have described in section 6 do represent learning problems, and for
this reason we shall briefly summarize these contrasts (tables 8 and 9). The
results are presented in the following way: the first three columns describe
the effects of transferring gambits from the L t in cases in which such transfer
leads either to an error (second column) or to an overrepresentation of the
formally equivalent gambit in the L 2 (third column). In the fourth column
we describe gambits in the L 2 which we hypothesize will be underrepresented
because of lack of possibility for transfer from the Li.
The validity of the results presented in tables 8 and 9 is affected by a number
of limitations which pertain to the data basis and to the theoretical frame-
work of our research.
Table 8: Learner problems for Danish learners of German,

hypothesized on the basis of CA
Danish German
over- under-
error representation representation
jo doch
Receipts nej nein
n na
ja
Starters (ja)men (ja)aber
ikke ne, n, nich

Appealers vel wohl
ja
oder
Table 9: Learner problems for German learners of Danish,

hypothesized on the basis of CA
German Danish
over- under-
error representation representation
ja ja ja
jo
Receipts
nej
n
jamen
Starters ja ja men
ne, n, nich ikke ikke

ja ja
Appealers oder eller
vel
hva'
First, on the basis of our two corpora we have not been able to describe
frequency differences in the use of various gambit types and tokens. Similarly,
we have only given rather sketchy rules of distribution for individual gambit
elements. Without an analysis of the frequency with which various gambits
are actually used, and without a more detailed specification of the pragmatic
constraints on their applicability, our predictions about potential over- and
underrepresentation of gambit types and tokens in German-Danish or Danish-
German interlanguage are necessarily very tentative. Second, it is difficult to
determine to what extent our results are representative for a majority of
Danish and German native speakers with comparable social features, inter-
acting in similar situations, and to what extent they are due to idiosyncracies
in our subjects' use of gambits. It seems to be the case that these linguistic
phenomena are more subject to individual variation, and more directly related
to a speaker's socio-psychological characteristics, than many other aspects of
language. What one would like to know, for descriptive as well as for pedagog-
ical purposes, is which gambits are in fact used and accepted by most native
speakers under given pragmatic conditions, which are less frequently used,
but widely accepted, and which gambits are so strongly idiosyncratically
marked that they create communication problems (irritation and/or misunder-
standing) for most interlocutors. Here we touch upon the problem of linguistic
norms and tolerance margins, which is of course another highly relevant issue
both for the description and evaluation of learner language and for teaching
purposes. Third, we feel that we have failed to find a satisfying procedure in
establishing a tertium comparationis for our contrastive analysis. Rather, we
have started by using a classification system which was developed on the basis
of English, and which we modified for the present analysis. Especially with
the various subfunctions within a particular class of gambits, we have the
suspicion that the lack of a more general frame of reference which is neutral
between the languages under investigation obscures functional equivalences
which are more intricate and complex than the ones which we have established
in a rather ad-hoc manner. However, the issues of universality and specificity
in discourse have hardly begun to be investigated, and their formulation into
a model for contrastive discourse analysis has not been attacked at all. We
would like the reader to understand this closing remark both as an excuse for
the shortcomings of the present article, and as a suggestion for future research.
Notes
* We would like to thank the participants at the 4th International Conference on Con-
trastive Projects, Charzykowy, for a long and illuminating discussion of a preliminary
version of the present article.
1. Cf. Faerch - Kasper 1980: 92 for a brief overview of learners' discourse problems as
discussed in the literature. A detailed study of pragmatic and discourse problems of
German learners of English has been presented by Kasper (1981).
2. Other terms used in the literature for reference to the same or similar linguistic func-
tions are: "speaker/hearer signals" (Duncan 1974; Schwitalla 1976), "attention sig-
nals", "linking signals" (Leech - Svartvik 1975), "connectives" (Crystal - Davy 1975),
"greasers" (Fillmore 1983), "fumbles" (Edmondson 1981).
3. We would like to thank the students who participated in the seminar for their coopera-
tion in collecting and transcribing the data.
4. As receipts, exclaims and represents can also be used as turntaking gambits, the occur-
rence of one of these at a possible completion point would most likely be interpreted
by the present speaker as a turntaking gambit.
5. The following conventions are used in the transcriptions of the data:
" primary stress
1
secondary stress
\ fall
t rise
V fall - rise
(These symbols precede the syllable concerned.)
- pause relative to speaker's speed of delivery
I . . . ] simultaneous speech activity of speaker who has not got the turn
( . . . ) extraverbal activity
6. For a more detailed analysis of the German data, see House 1980.
7. Thus Danish children are sometimes told not to use but ja or nej as a receipt as we
often conveys the connotation of "I don't care".
8. The restrictions we have described here hold true for receipts with both an uptaking
and a turntaking function, but not for n and ja when used as go-ons. In this case, ja
can be used as an uptaker even if the hearer does not know the content of the speak-
er's S-H proposition: "x: og s m^dte jeg Lone [y: tja\ og hun fortalte mig . . ."
x: ' and then I met Lone [y: tyes] and she told me . . . ' .
References
Beneke, Jrgen
1975 "Verstehen und Miverstehen im Englischunterricht", Praxis des neu sprach-
lichen Unterrichts 2 2 : 3 5 1 - 3 6 2 .
Coulthard, Malcolm
1977 An introduction to discourse analysis (London: Longman).
Crystal, David - Derek Davy
1975 Advanced conversational English (London: Longman).
Duncan, Starkey Jr.

1974 "On the nature of speaker-auditor interaction during speaking turns", Lan-
guage and society 3 : 1 6 1 - 1 8 0 .
Edmondson, Willis
1977a "Gambits in foreign language teaching", Kongressdokumentation der 7. Ar-
beitstagung der Fremdsprachendidaktiker Gieen 1976, edited by . Christ
and . Piepho (Limburg: Frankonius), 4 5 - 4 7 .
1977b Gambits revisited. Unpublished manuscript, Bochum.
1981 Spoken discourse: A model for analysis (London: Longman).
Edmondson, Willis - Juliane House
1981 Let's talk and talk about it. A pedagogic interactional grammar of English
(Mnchen: Urban und Schwarzenberg).
Edmondson, Willis - Juliane House Gabriele Kasper Brigitte Stemmer
1982 Kommunikation lernen und lehren. Berichte und Perspektiven aus einem
Forschungsprojekt (= Manuskripte zur Sprachlehrforschung 20) (Bochum
and Heidelberg: Groos).
Fasreh, Claus - Gabriele Kasper
1980 "Processes and strategies in foreign language learning and communication",
Interlanguage studies bulletin 5 : 4 7 - 1 1 8 .
Fillmore, Charles L.
1983 "Remarks on contrastive pragmatics", this volume.
Foreman, Donald
1974 "The speaker knows best principle", Papers from the Tenth Regional Meet-
ing of the Chicago Linguistic Society (Chicago: Chicago Linguistic Society),
162-176.
House, Juliane
1980 "Gambits in deutschen und englischen Alltagsdialogen", Sprache und Ver-
stehen, vol. 2, edited by W. Khlwein and A. Raasch (Tbingen: Narr), 1 0 1 -
107.
House, Juliane - Gabriele Kasper
1981 "Politeness markers in English and German", Conversational routine. Ex-
plorations in standardized communication situations andprepatterned speech,
edited by F. Coulmas (The Hague: Mouton), 1 5 7 - 1 8 5 .
Kasper, Gabriele
1979 "Errors in speech act realization and use of gambits", The Canadian modern
language review/Revue Canadienne des langues vivantes 3 5 : 3 9 5 - 4 0 6 .
1981 Pragmatische Aspekte in der Interimsprache. Eine Untersuchung des Engli-
schen fortgeschrittener deutscher Lerner (Tbingen: Narr).
Keller, Eric - Sylvia Taba-Warner
1976, 1977, 1979 Gambits 1, openers; Gambits 2, links; Gambits 3, responders,
closers and inventory. A series of three modules (Ottawa: Government of
Canada).
Kellerman, Eric
1978 "Transfer and non-transfer: where we are now", Studies in second language
acquisition 2 . 1 : 3 7 - 5 7 .
Leech, Geoffrey - Jan Svartvik
1975 A communicative grammar of English (London: Longman).
Lyons, John
1968 Introduction to theoretical linguistics (Cambridge: University Press).
Rehbein, Jochen
1979 "Sprechhandlungsaugmente", Die Partikeln der deutschen Sprache, edited by
. Weydt (Berlin: de Gruyter), 5 8 - 7 4 .
Schwitalla, Johannes
1976 "Dialogsteuerung. Vorschlge zur Untersuchung", Projekt Dialogstrukturen,
edited by F. J. Berens, K.-H. Jger, G. Schnk and J. Schwitalla (Mnchen:
Hueber), 7 3 - 1 0 4 .
Sinclair, John - Malcolm Coulthard
1973 Towards an analysis of discourse. The English used by teachers and pupils
(London: Oxford University Press).
Strevens, Peter D.
1972 British and American English (London: Collier-Macmillan).
Widdowson, Henry
1978 Teaching language as communication (London: Oxford University Press).
RUDOLF FILIPOVIC
What are the primary data for contrastive analysis?
1 Introduction. In preparing theoretically the work of the Yugoslav Serbo-

Croatian-English Contrastive Project (YSCECP) and the Zagreb English-Serbo-
Croatian Contrastive Project (ZESCCP), I dealt with a whole range of funda-
mental and methodological questions in order to form a methodology for
Contrastive Analysis (CA) and to fix the programme of work for both pro-
jects. One of the questions which deserve our full attention is what are the
primary data for CA. If we analyse how the work of the YSCECP proceeded
from its first phase, 19681971, through its second phase, 19721975, and
finally to its third phase, 19761980, we will go through several stages of CA
and at the same time come across several kinds of primary data that the anal-
ysis of the two language systems was based on.
2 Grammatical description of Ly and L2. One of the basic principles of CA

methodology says that both systems have to be described equally well, espe-
cially if the two languages confronted are genetically and typologically differ-
ent. In the first stage of our work on the YSCECP the above principle was
followed in each article and analysers who took various grammatical units as
their topics studied them in both languages.
For English they consulted about thirty standard works on English gram-
mar (Filipovic 1969a:1617) and all available articles and other sources on a
particular problem. Their analyses, however, were written not only on the
basis of the literature available, but also on the analysers' experience and intu-
ition. The same was done for each topic in Serbo-Croatian (S-C): several S-C
grammars (Filipovic 1969a: 17) and specialized literature on the problem were
consulted. Thus we obtained the required description of a topic in both sys-
tems. Then the topic was examined contrastively and we got what we call the
first report on the CA of a grammatical unit (Filipovic 1969b1976).
3 Learners' errors. During the first phase of the YSCECP we came to the
conclusion that the decision made at the very inception of our project, to
108 Rudolf Filipovic
organize two teams of researchers, one working on CA and the other on error
analysis (EA), was of great importance for our future work on the project.
First we never faced the dilemma, which vexed some other contrastive pro-
jects, which of these two analyses (CA or EA) should be given preference,
which is more useful, and which can give better and more lasting results.
Secondly, through a very careful and exhaustive error analysis we identi-
fied all the topics that deserved thorough treatment in both projects. For the
YSCECP, error analysis was carried out within the framework of a project
called "Morphological and syntactic errors in the speech of students of English
in Serbo-Croatian speaking areas" (Andrassy 1971; Bilimic 1971; Kranjfievic
1971). This work helped us to ascertain several types of deviation and to
identify various causes for these deviations. For this purpose we collected a
great number of learners' errors typical for students of English in the Serbo-
Croatian speaking areas in Yugoslavia.
For the ZESCCP, error analysis was not carried out on such a large scale.
However, the choice of topics that were examined and analysed contrastively,
was done through error analysis based on spoken and written material col-
lected at the S-C courses for English speaking students from Britain and the
U.S.A. (Nakic 1975a-d; Bauer 1978;Mazalin 1978).
The second stage of error analysis was carried out in the form of tests; all
the results of CA achieved in the YSCECP were checked with the aid of spe-
cial tests in experiment of 500 learners (Filipovic 1973-1974b). In this way
additional material about the errors made by Serbo-Croatian speaking learners
was supplied. Through the results achieved in A we got acquainted with
learners' errors in both directions and this was of great use in the first phase
of CA in the YSCECP and in the ZESCCP.
4 A corpus in Li and L2. For both projects we decided to apply the same
method: the translation method. This decision led us logically to a corpus
since this method requires work on language material, i.e. a corpus. A com-
plete CA based on the translation method requires a detailed corpus in both
languages with translations into the other language. Only a corpus constructed
in this way enables us to carry out a complete investigation of some phenom-
enon in both languages.
Even in my first paper outlining our programme of work (Filipovic 1967)
I emphasized our intention of developing a bidirectional corpus of 100,000
English sentences translated into S-C and 100,000 S-C sentences translated in-
to English. Soon after that I laid down specific principles for the construction
of these two corpora. We intended to include both British and American
authors, non-fiction and fiction, along with S-C translations. We also planned
to record some spoken language and have it translated into S-C. Similarly, the
The primary data for CA 109
Serbo-Croatian to English part would include writers representing the differ-

ent geographical variants of the literary language, with English translations
made by both Englishmen and Americans.
We still believe that only such a bidirectional corpus, consisting really of
two corpora: a corpus of L t and a corpus of L 2 , can ideally satisfy all the re-
quirements for the primary data for CA. However, the scope and complexity
of this work: (a) the excerpting of materials, and (b) the processing of this
huge corpus (or rather two corpora) impelled us to give up the bidirectional
idea and to seek one of the already existing corpora of English, even though
this would give us only a unidirectional one. Clearly, both bidirectional and
unidirectional corpora can be considered the primary data for CA. Therefore
we will proceed with our analysis of the primary data for CA and try to prove
the value and necessity of a corpus for CA.
5 A unidirectional corpus as primary data for CA. In our projects, as we have

seen, 'reports' were written on the basis of data from grammars, errors, and
previous knowledge. They were not meant to be the final products of the
contrastive analysis done in both projects. Analysers were next supposed to
write 'studies' for each grammatical unit. These studies were envisioned as
final products and were to serve as a basis for writing a contrastive grammar
of S-C and English (Filipovic 1978, 1981b). They are substantially different
from the previously written reports because they are based on 'bilingual con-
trastive concordances' (Bujas 1975 ; Filipovic 1981a) derived from the corpus.
This means that all the results achieved in CA of a grammatical unit in the
first two phases of the project are further documented and tested with ex-
amples from our bilingual corpus.
This brings us to the main problem we want to examine in this paper: How
does a corpus and its translation fit into the primary data for contrastive ana-
lysis? The best way of examining this problem seems to be to discuss the
relation between a corpus and the intuitions of native speakers in contrastive
analysis. In this discussion we want to give the reasons for each choice - that
is, whether to use a corpus or to use native informants.
6 The task of the German-English Contrastive Project - PAKS, which was

directed by Professor G. Nickel, was to provide "an adequate description of
the German and English languages on the basis of the generative-transforma-
tional theory" and to "contribute to the further development of T. G. theory"
(Nickel 1972:225-226). In his description of the project, Nickel gave an ex-
planation of his method, which was based on the T-G and which did not en-
vision the use of a corpus.
110 Rudolf Fipovic
7 The German-English Contrastive Project which was located for a while in

Mainz and which was directed by Professor B. Carstensen, was based on
T. G. theory, but also involved the use of large standard dictionaries of both
languages as well as the most recent structural and transformational descrip-
tions of their syntax. This project not only planned to use large dictionaries,
but also planned to analyse a large corpus of the two languages being con-
trasted.
In order to complete that task, this project envisioned the use of the com-
puter to obtain statistical information about the frequency of use and the
degree of regularity of certain linguistic elements, either in a single discourse
or in a series of discourses. Professor Carstensen envisioned the use of inform-
ants that is, intuitions of native speakers - because in his opinion certain
sorts of information could only be obtained from informants. This project
did not accomplish a single aim of its programme, it very quickly ceased to
exist, and it has not left any significant results. We mention it only because it
had an interesting, although unproductive programme.
8 The Polish-English Contrastive Project also used the T.G. model and did
not envision the use of a corpus. This project did in its beginning stages begin
to collect a corpus; but this corpus was considered to be only an aid in the
work. The director of the project, Professor J. Fisiak, confirmed later that the
Polish-English Contrastive Project considers "the corpus to be of help only in
some cases and unnecessary in other cases, because the corpus is not in itself
our goal [. . .]. If the corpus provides only a few examples of the unit being
investigated, then one must look for examples outside the corpus." The main
reason for this view towards the corpus was that the Polish project was using
the T. G. model which does not envision the use of a corpus, but only the use
of native speakers as a source of information about the language.
9 Although the Hungarian-English Project (Hegeds 1971, Dezs 1971; Dezs

- Stephanides 1976) was not based on the T-G model, it still did not develop
any large standard corpus which could be of use in contrastive analysis. The
need for a corpus was recognized in the project, but only for a limited and
unsystematic one.
10 In its contrastive analysis the English-Romanian Project (Slama-Cazacu

1971; Chijoran 1976) planned to use a corpus consisting of a vocabulary of a
few thousand English units. This corpus was analysed on the basis of the poly-
semy of these elements and their role in grammatical constructions.
11 All the projects mentioned above, regardless of the linguistic model ac-
cording to which they were working, had some kind of corpus, but none of
them had a standard corpus. For us, 'standard' means that the corpus can be
used in comparative studies in which it is necessary to have the same collec-
tion of data for each language. Standard can mean that the corpus is suitable
for further work preparing and presenting other data in English and other
languages (Filipovic 1969:49).
12 The fact is that each contrastive project has a corpus: a) those projects
based on the T.G. model, according to which one uses intuitions of native
speakers and does not need a corpus, use a corpus anyway generally refer-
red to as a 'helping' corpus; b) those projects which are based on some other
model have a corpus, more systematic than the 'helping' corpus used by the
first group, but still not 'standard'; and c) YSCECP is the only project to
work with a standard and representative corpus. This shows that contrastive
analysis, regardless of the linguistic model on which it is based, regardless of
whether or not it relies on native speakers, must use some sort of corpus, pre-
ferably one which is standard and representative.
13 If the answer to our question whether to use a corpus or not is positive,

then we must further show (a) why it is best to base one's analysis on a stand-
ard and representative corpus and not on one of some other sort, and (b) how
the corpus can best be processed so that it can form the basis for the con-
trastive analysis of two languages. Since YSCECP till now is the only contrastive
project to use a standard corpus, that is, the Zagreb version of the Brown
Corpus (Filipovic 1976), we can best illustrate the use and value of a standard
corpus in contrastive analysis by considering a few examples from the con-
trastive analysis of S-C and English carried out by this project.
14 Work on individual grammatical units from the syntax for example the
analysis of English possessive pronouns and their equivalents in S-C (Spalatin
1970) - has shown that without a standard corpus the investigator cannot
complete his analysis. This analysis would begin with an examination of the
uses of these pronouns in basic literature and the author's own knowledge
and experience with the language. However, as soon as the investigator began
to seek formal semantic correspondences in the second language, he would
immediately feel the lack of a corpus. As at that time computer processing of
our corpus was not yet complete, the investigator would compile a small pilot
corpus, and from this he would work out a table showing the equivalents. All
these data remained provisional and statistically incomplete as long as the
analysis did not proceed from the computer processing of the corpus and so
long as these data were not supported by contrastive concordances. Further-
more, so long as the investigator based his analysis on a pilot corpus, he could
qualify his conclusions only with phrases such as Very little', 'almost always',
Very rarely', etc. By using the whole corpus, he would be able to provide
relevant statistical data and reach more precise conclusions.
15 A further example illustrating our point is the analysis of relative clauses

which went through both phases of our work, i.e. the 'report', based on
literature and the pilot corpus of 1,000 relative clauses (Maiek 1970), and a
later study based on 30,000 examples from the contrastive concordance of
the Zagreb version of the Brown Corpus (Maiek 1975). These two phases of
the project demonstrate that it is not sufficient just to have a corpus, that is,
any sort of corpus, but one must have a truly representative one, i.e. suffi-
ciently large, representative and standard in our sense of the word. It happened
that in the pilot corpus certain translational equivalents simply did not appear.
In addition, numerical relations were also of importance to us, since contrastive
analysis must show the frequency of certain uses of relative pronouns, in
order for our results to be used in the pedagogical materials (Filipovic 1971c
-1974).
16 The investigator in the study mentioned above showed that without a

standard and representative corpus, one could not accomplish the following
goals of the analysis: a) to determine when and how often relative pronouns
are used and when they may be omitted; b) to examine the grammatical and
semantic nature of the antecedent to which the relative pronoun was related;
c) to determine the use and placement of prepositions with relative pronouns;
d) to provide a detailed study of the function of relative pronouns; e) to in-
vestigate separately the use of relative pronouns in restrictive and non-restric-
tive clauses.
17 Equally useful for our thesis regarding the usefulness of a representative

and standard corpus in contrastive analysis are the observations we have made
about the work of projects based on the T-G approach. Such projects began
their work with a contrastive analysis of equivalent rules in both languages.
Even after successful completion of such investigations, these projects still
found use for some sort of corpus. So far they have not followed our view-
point about the need for a representative and standard corpus, but have com-
piled their own smaller, ad hoc corpus in order to a) examine the functioning
of the rules that have been established, and b) obtain examples (which they
could not obtain from intuition or from the use of native speakers) on the
basis of which they could carry out further investigation and formulate new
rules (Knig 1971). This problem becomes especially important because of
increasing doubt about the extent to which we can rely on the informants on
whom investigations within the T-G model depend. Ilse Lehiste (1971:79)
maintains that today we can rely less and less on grammaticality judgements
of native speakers because of an increasing amount of variation among speakers
and an increasing amount of similarity between native and non-native speakers.
18 At the Zagreb Conference on English Contrastive Projects, in summarizing

the results of this conference (Filipovic 1971:79), I observed that the work of
the conference had shown that each contrastive project, regardless of their
linguistic model, was relying on some sort of corpus either from the very be-
ginning or later on when their analysis required it. In answer to one of the
questions in the discussion, I emphasized that the need for a corpus did not
mean the abandonment of all theory. The material from the corpus serves
first to verify the conclusions based on the theory and second, to provide a
means to collect data in areas where the theory is inadequate.
Thus it would be a mistake to talk of a choice between theory and corpus.
Instead we must consider the degree of their interdependence: with the
structuralist theory and the translation method, the corpus is indispensable,
and with the generative approach to contrastive analysis, the corpus has been
shown to become more useful as we proceed in our analysis from the general
to the more detailed (Filipovic 19731974a).
19 Similarly, it would be mistaken to exclude native speakers no matter

what approach is being used. Native speakers can be viewed as a sort of 'living
corpus' and are often quite useful in confirming theoretical results. Native
speakers must be consulted in cases in which a) in the analysis of certain units
results are obtained from the corpus which must be confirmed by another
method; and b) the corpus cannot provide adequate help in reaching conclu-
sions. This can happen even with a large corpus, as representative and stand-
ard as possible, especially since it is very difficult to have a processed corpus
which would cover the most recent linguistic usage. Accordingly, although the
corpus is indispensable, it does not exclude any other method which would
be useful to us in contrastive analysis.
20 Our long experience in the contrastive analysis of languages has shown us

that such analysis cannot be carried out without the use of a corpus. Today it
is generally accepted that not one important part of language can be con-
trastively analysed without precise data on distribution. We cannot obtain
such data from just any sort of language material, collected in an ad hoc man-
ner, but only from a well organized corpus. Such data on distribution cannot
be obtained from native speakers, because, for psychological reasons, they
can give data of only one individual, and these data often reflect the bias of
the investigator.
Distributional information obtained from grammars is not reliable. The
grammarian usually seeks examples that illustrate a theory; consequently,
there is the danger that he has used only those examples that support his
theory and ignored all the rest. In the systematic use of the corpus, this is not
possible.
21 By working on a good corpus, it is possible to investigate contrastively

the stylistic value of some construction and to determine its statistical signif-
icance and representativity. This is possible because the corpus contains long
stylistically homogeneous extracts from continuous texts.
22 As one of the aims of contrastive analysis is the application of its results

in pedagogical materials, the use of a representative corpus in contrastive ana-
lysis has direct pedagogical applications, in that it offers better and more
varied linguistic material than do the individual examples which are often
used to illustrate rules in theoretical T-G discussions.
23 In conclusion, we can state that there are several primary data that the
contrastive analysis of the two language systems can be based on. We have
discussed the following: a) grammatical descriptions of Ll and L2 which are
very suitable for the initial stage of CA; b) learners' errors which help to de-
velop further the analysis and its scope; c) a corpus of Li and L 2 , i.e. a bi-
directional corpus, which can ideally satisfy the requirements for CA; d) a
unidirectional corpus which does not satisfy the requirements for CA as per-
fectly as a bidirectional one. However, we have shown that even a unidirec-
tional corpus has an important place among the primary data for CA, since:
a) a valid contrastive project cannot be considered complete before its results
have been verified and completed with the help of some representative corpus;
b) only a corpus can verify certain cases of doubtful grammaticality; c) fre-
quency and distribution can be established only on the basis of a corpus;
d) without a corpus we could not analyse the stylistic value, i.e. stylistic levels
and registers, of certain forms; e) the corpus is necessary for the component
of 'use', which with the taxonomic and generative components constitutes
the 'contrastive mix' (Ivir 1971:167), without which it would be difficult to
imagine the successful completion of contrastive analysis; f) without a corpus
it would be impossible to obtain a more or less complete list of all units which
belong to some part of speech; such a list is important for contrastive analysis
and its practical applications.
References
Andrassy, V.
1971 "Errors in the parts of speech in the English of learners from the Serbo-
Croatian-speaking area", YSCECP, Pedagogical materials 1, edited by
R. Filipovic (Zagreb: Institute of Linguistics), 7 - 3 2 .
Bauer, I.
1978 "Verbal aspect: error analysis", ZESCCP, Contrastive analysis of English
and Serbo-Croatian, Vol. II, edited by R. Filipovic (Zagreb: Institute of
Linguistics), 2 7 4 - 3 2 5 .
Bilinic, J.
1971 "Errors in the morphology and syntax of the verb in the speech of learn-
ers of English in the Serbo-Croatian-speaking area", YSCECP, Pedagog-
ical materials 1, edited by R. Filipovic (Zagreb: Institute of Linguistics),

32-60.
Bujas, Z.
1975 "Computers in the Yugoslav Serbo-Croatian-English contrastive project",
Bulletin of the Institute of Linguistics (Zagreb) 1 : 4 4 - 5 6 .
Carstensen, B.
1972 "Contrastive syntax and semantics of English and German", Active meth-
ods and modern aids in the teaching of foreign languages, edited by
R. Filipovic (London: OUP), 2 0 6 - 2 1 6 .
Chitoran, D.
1976 "Report on the Romanian-English Contrastive Analysis Project", 2nd
International Conference on English Contrastive Projects, edited by
D. Chitoran (Bucharest: Bucharest University Press), 1 1 - 3 4 .
Desz, Laszlo
1971 "Contrastive linguistics project on English and Hungarian in Hungary",
Zagreb Conference on English Contrastive Projects, edited by R. Fili-
povic (Zagreb: Institute of Linguistics), 1 2 4 - 1 2 9 .
Desz, Lszl - Eva Stephanides
1976 "Report on the English-Hungarian Contrastive Linguistics Project", 2nd
International Conference on English Contrastive Projects, edited by
D. Chitoran (Bucharest: Bucharest University Press), 5358.
Filipovic, R.
1967 "Contrastive analysis of English and Serbo-Croatian - Theory and prac-
tice", Studio Romanica et Anglica Zagrebiensia 2 3 : 5 - 2 3 .
1968 "Uloga kontrastivne analize u lingvistiSkom istraiivanju" [The role of
contrastive analysis in linguistic research], Filoloiki pregled (Beograd)
3-4: 1-10.
1969 a "Poietne faze rada na projektu 'Kontrastivna analiza hrvatskosrpskog i
engleskog jezika* " [The initial phase of the project 'Contrastive analysis
of Serbo-Croatian and English'], Prilozi i gradja, edited by R. Filipovic
(Zagreb: Institute of Linguistics), 3 - 2 5 .
1969b- YSCECP, Reports, 1 - 1 0 , edited by R. Filipovic (Zagreb: Institute of
1976 Linguistics).
1969c- YSCECP, Studies, 1 - 6 , edited by R. Filipovic (Zagreb: Institute of
1975 Linguistics).
1969d "The choice of the corpus for the contrastive analysis of Serbo-Croatian
and English", YSCECP, Studies 1, edited by R. Filipovic (Zagreb: Insti-
tute of Linguistics), 3 7 - 4 6 .
1970-1971 "Problems of contrastive w o r k " , S / M Z 2 9 - 3 2 : 1 9 - 5 4 .
1971a "The Yugoslav Serbo-Croatian-English contrastive project so far (Papers

and discussion)", Zagreb Conference on English Contrastive Projects,
(Zagreb: Institute of Linguistics), 3 1 - 8 6 .
1971b "Contrastive analysis and error analysis in pedagogical materials", YSCECP,
Pedagogical materials 1 (Zagreb: Institute of Linguistics), 1 - 7 .
1971c- YSCECP, Pedagogical materials 1 - 2 , edited by R. Filipovic (Zagreb:
1974 Institute of Linguistics).
1973- "The use of a corpus in contrastive analysis", SRAZ 33-36:489-500,
1974a and in Trends in kontrastiver Linguistik I, Das Zagreber Projekt zur ange-
wandten kontrastiven Linguistik, edited by . Raabe (Mannheim: Institut
fr Deutsche Sprache), 5 1 - 6 6 .
1973- "Testing the results of contrastive analysis", Rassegna Italiana di Linguis-
1974b tica Applicata (Roma), 2:155-163. Also in: ILA Third congress - pro-
ceedings Vol. 1, Applied contrastive linguistics, edited by G. Nickel (Hei-
delberg), 9 7 - 1 0 9 ; Trends in kontrastiver Linguistik I. Das Zagreber Pro-
jekt zur angewandten kontrastiven Linguistik (Mannheim: Institut fr
Deutsche Sprache), 1 7 9 - 1 9 1 .
1975a "Objectives and initial results of the Zagreb English-Serbo-Croatian con-
trastive project", ZESCCP, Contrastive analysis of English and Serbo-
Croatian, Vol. 1 (Zagreb: Institute of Linguistics), 3 - 2 5 .
1975b "The second phase of the Yugoslav Serbo-Croatian-English Contrastive
project", SRAZ 3 9 : 1 7 5 - 1 9 1 .
1975c "The Yugoslav Serbo-Croatian-English contrastive project at the end of
its second phase", YSCECP, Studies 6 (Zagreb: Institute of Linguistics),
1-22.
1976 "The Yugoslav Serbo-Croatian-English contrastive project from the
Zagreb Conference to the present", 2nd International Conference of
English Contrastive Projects, edited by D. Chitoran (Bucharest: Bucharest
University Press), 3 5 - 5 1 .
1978 "The application of CA and EA to a contrastive grammar", SRAZ 4 5 -
46:337-348.
1981a "Some basic methodological questions of contrastive analysis", Atti del
XII Congresso Internazionale del la LSI, Linguistica Contrastiva, (in press)
1981b "On some problems of writing a contrastive gtammsa", Forms and func-
tions, edited by J. Esser and A. Hbler (Tiibingen: Gunter Narr Verlag),
271-275.
Fisiak, J .
1971 "The Poznan Polish-English contrastive project", Zagreb Conference on
English Contrastive Projects, edited by R. Filipovic (Zagreb: Institute of
Linguistics), 8 7 - 9 6 .
Hegeds, J .
1971 "Two questions of Hungarian-English contrastive studies", Zagreb Con-
ference on English Contrastive Projects, edited by R. Filipovic (Zagreb:
Institute of Linguistics), 1 0 1 - 1 2 1 .
Ivii, V.
1971 "Generative and taxonomic procedures in contrastive analysis", Zagreb
Conference on English Contrastive Projects, edited by R. Filipovid
(Zagreb: Institute of Linguistics), 1 5 6 - 1 6 7 .
Knig, .
1971 "Transformational grammar and contrastive analysis", Zagreb Conference
on English Contrastive Projects, edited by R. Filipovic (Zagreb: Institute
of Linguistics), 1 2 9 - 1 4 6 .
KranjSevic, S.
1971 "Errors in the syntax of the sentence in the speech of learners of English
in the Serbo-Croatian-speaking area", YSCECP, Pedagogical materials 1,
edited by R. Filipovic (Zagreb: Institute of Linguistics), 6 0 - 8 1 .
Lehiste, I.
1971 "Grammatical variability and the difference between native and non-
native speakers", Papers in contrastive linguistics, edited by G. Nickel
(Cambridge: Cambridge University Press), 6 9 - 7 4 .
MaCek, D.
1970 "Relative pronouns in English and Serbo-Croatian", YSCECP, Reports 3,
edited by R. Filipovic (Zagreb: Institute of Linguistics), 105127.
1975 "Relatives in English and their Serbo-Croatian equivalents", YSCECP,
Studies 6, edited by R. Filipovic (Zagreb: Institute of Linguistics), 2 7 - 6 2 .
Mazalin, D.
1978 "Word order: error analysis", ZESCCP, Contrastive analysis of English
and Serbo-Croatian, Vol. II, edited by R. Filipovic (Zagreb: Institute of
Linguistics), 2 7 4 - 3 2 5 .
Nakic, A.
1975a "Word order (error analysis)", ZESCCP, Contrastive analysis of English
and Serbo-Croatian, Vol. I, edited by R. Filipovic (Zagreb: Institute of
Linguistics), 9 7 - 1 0 4 .
1975b "Complement clauses (error analysis)", ZESCCP, Contrastive analysis of
English and Serbo-Croatian, Vol. I, edited by R. Filipovic (Zagreb: Insti-
tute of Linguistics) 146-148.
1975c "Relative clauses (error analysis)", ZESCCP, Contrastive analysis of
English and Serbo-Croatian, Vol. I, edited by R. Filipovic (Zagreb: Insti-
tute of Linguistics), 1 4 8 - 1 5 4 .
1975d "Lexical errors (error analysis)", ZESCCP, Contrastive Analysis of English
and Serbo-Croatian, Vol.1, edited by R. Filipovic (Zagreb: Institute of
Linguistics), 1 5 5 - 1 6 6 .
Nickel, G.
1972 "Project on applied contrastive linguistics: a report", Active methods and
modern aids in the teaching of foreign languages, edited by R. Filipovic
(London: Oxford University Press), 2 1 7 - 2 2 6 .
Slama-Cazacu, T.
1971 "The Romanian-English language project", Zagreb Conference on Eng-
lish Contrastive Projects, edited by R. Filipovic (Zagreb: Institute of
Linguistics), 2 2 6 - 2 3 4 .
Spalatin, L.
1970 "The English possessive adjectives my, your, his, her, its, our, their and
their Serbo-Croatian equivalents", YSCECP, Reports 2, edited by R. Fili-
povic (Zagreb: Institute of Linguistics), 9 4 - 1 0 2 .
CHARLES J. FILLMORE
Remarks on contrastive pragmatics
1 Introduction
My purposes in this paper are: first, to sketch out what I take to be a work-
able notion of linguistic pragmatics, and to formulate in its terms a distinction
between 'large' (or systematic) and 'small' (or item-by-item) differences in the
pragmatic practices of different linguistic communities; and second, to illus-
trate methods of discovering and describing some of the item-by-item kinds
of pragmatic knowledge, in particular those related to formulaic expressions
in English and pragmatic particles in German. In both cases what I have to say
is merely programmatic and illustrative. The material on formulaic expressions
is part of a collaborative effort with Lily Wong Fillmore, a work that will not
be finished soon. The remarks on German modal particles are nothing more
than an illustration of the difficulties one faces in doing work in linguistic
pragmatics, responding to a number of insights by Jan-Ola stman, Knud
Lambrecht and Dorothea Franck.
2 Pragmatics in linguistic theory
It must be pointed out at the outset that there is no characterization of lin-

guistic pragmatics on which linguists are in agreement, nor is there, in fact, a
universally convincing case that such a field exists. Several major traditions of
linguistic theorizing (some, among others, originating in London and Prague)
have been getting along quite well without a separately articulated theory of
pragmatics. To Halliday, for example, the interpersonal and textual com-
ponents of the semantic system do not make up a separate layer of conven-
tionalization of the sort that one finds in theories that speak of pragmatics as
a distinct level of linguistic analysis (Halliday 1974). Within several of the
schools of syntactic theorizing that have risen out of the American generativist
tradition, syntactic principles are being formulated that begin with appeals to
120 Charles Fillmore
speech act typologies, pragmatic purposes, and principles of rhetorical organi-

zation; to these scholars syntax is not to be cleanly separated from pragmatics
(Van Valin - Foley 1981). Accounts of indexical elements in language - in-
escapably a matter of semantics - beginning with pronouns and tenses and
demonstratives, and continuing on to politeness forms and discourse anchoring
devices, speak against a clear distinction between semantics and pragmatics
(Fillmore 1973). And among those scholars whose layered idealizations of the
workings of language do maintain a field of pragmatics separate from syntax
and semantics proper, there is disagreement on how and whether a theory of
pragmatics needs to incorporate the processes of speech comprehension and
the role of belief systems in communication and understanding, in addition to
those more traditional notions of pragmatics related to the phenomena of lan-
guage use.
In the tradition of scholars who accept a separate layer of pragmatics, a
schema like the following captures a near-standard view of the interworking
of the various proposed levels of linguistic abilities.
In the diagram pale-face squares and bold face squares contain rules and
representations respectively. Box 2 represents the integration of morpheme
meanings and lexical structure into word meanings. Box4 represents the assign-
ment of semantic structure to complex syntactic structures. Box 6, containing
the main principles of compositional semantics, represents the apparatus for
integrating lexical semantics with the semantics of grammar, resulting in mean-
ing representations (the sentence meanings of circle 7) which are incomplete
only in that their indexical elements and their potential referring expressions are
not yet anchored to the entities of coordinates of any particular 'possible world.'
Box 8 represents a function which maps a sentence's referring expressions
to entities, places and times in some possible world. The meaning representa-
tions in box 9 may be called 'utterance meanings'. (Utterances, unlike sen-
tences, are capable of being true or false in a world.) Box 10 stands for the set
of processes by which one situates an utterance in a context, associating it
with the goals and intentions of the participants in the associated speech act
and with the physical and social 'location' of the utterance's producer.
Challenges to the validity of models representable by the diagram can be
made at almost every point. For example, in certain kinds of indirect speech
acts an utterance's contextualization is required before a referent-identifying
interpretation can be provided. (Consider "A certain young man seems to
have forgotten to brush his teeth.") That is, inferences of the sort assigned to
box 10 appear to be required before it is possible to carry out the assignment
function otherwise assumed to be in box 8. Certain kinds of contexted inter-
pretations are sometimes prerequisite to assigning semantic content to a sen-
tence. (This is the point of George Lakoff's notion of Amalgams; see Lakoff
Contrastive pragmatics 121
1974.) The context of an utterance sometimes constrains or motivates the ap-

plication of particular word-formation principles. (See Zimmer 1972 and
Downing 1977.) Certain syntactic forms seem to be dedicated to particular
pragmatic purposes, permitting the view that there may be a direct link be-
tween pragmatics and syntax that is not mediated through semantics. (Here
I have in mind such 'syntactic idioms' as exemplified by "Who doesn't know

that?" or "Who can't do that?".) And instead of positing an abstract assign-
ment function for what is here called 'anchoring', it makes more sense to
point out that the context in which an utterance is produced automatically
provides such assignments.
For reasons such as those mentioned in the last paragraph, I suspect that
the final complete description of human linguistic abilities will not contain a
separate pragmatics compartment. Nevertheless, I believe that for certain
practical purposes we can usefully separate grammar proper (phonology,
word-formation, syntax and semantics) from pragmatics much in the same
way that we can separate both physical and functional descriptions of a set of
tools from an account of how those tools are actually used in building, main-
taining, and destroying things. It would not be misleading to compare a gram-
mar of a language to a tool factory, and then to think of pragmatics as a de-
scription of the skills and practices of the workmen who use these tools. (My
analogy will work best if, perhaps giving me more cooperation than I deserve,
you are generous in your understanding of the word 'tool'.) If we felt it neces-
sary to find the components of a grammar within the analogy, we might sug-
gest that phonology corresponds to the knowledge of how, in the making of
these tools, particular materials can be molded and shaped or fitted together
in making the simpler tools, that word-formation and syntax correspond to
the knowledge of how complex tools and machines are assembled, and that
semantics corresponds to the knowledge of the purposes for which the indi-
vidual tools were constructed.
There are two characteristics of my analogy which make the view suggested
here different form certain common views of the structure and function of
language. (1) First is the suggestion it incorporates that semantics has basic-
ally to do with knowledge of what the tools of language are for. I believe that
semantic knowledge feeds fairly directly into pragmatics, by way of relating
the reasons why tools exist with knowledge of how they are typically used.
I do not accept a commonly held view on the nature of the difference between
lexical-semantic knowledge and encyclopedic knowledge, the view that, for
example, whatever we know about the 'meaning' of the word carpenter must
be kept sharply distinct from our knowledge of what it is that carpenters
actually do. (In some idealizations of meaning, the analyst associates with the
noun carpenter some smallest set of semantic features capable of distinguishing
the concept of carpenter from all other concepts in the contrast sets within
which carpenter is understood.) I do indeed accept a distinction between lin-
guistic and encyclopedic knowledge, but I feel that the boundary has often
been drawn in the wrong place. True 'encyclopedic' (and hence not linguistic)
knowledge about carpenters would have to do with their numbers in a partic-
ular time, their pay scale, etc. (2) A second intended characteristic of the ana-
logy is that it forces a certain way of talking about texts. In particular, it does
not allow us an easy way of viewing a text as a complex structure having
phrases and sentences as its parts. This is so because we cannot easily think of
a text as an assembly of tools (Fillmore 1983).
We do, of course, need to be able to talk about texts, especially if we are
agreed that linguistic pragmatics can most usefully be formulated as know-
ledge about how to associate texts with their potential or actual contexts of
use. Following Halliday, we can distinguish 'text as process' from 'text as pro-
duct' (Halliday 1983). Within our analogy, it is easy enougji to think about
text-as-process: the process is an activity in which somebody is engaged
(telling a story, insulting a neighbor, proving a theorem), the tools of language
(words, phrases, sentences) being the means for carrying out that activity.The
novelty of the analogy is that it requires us to think of text-as-product as
something like a record of the particular tools used in carrying out a partic-
ular activity, and the sequence in which those tools were used.
The critical characteristic of the analogy shows up when we think of what
it is to understand a text. If a text is a record of the tools used in carrying out
an activity, then the process of understanding that text is best understood in
terms of the steps one goes through in figuring out, from the list of tools, just
what that activity was. An interpreter achieves comprehension of a text by
knowing the 'semantics' of the tools, by knowing conventional ways of using
tools in the local workplace, by appealing to common sense assumptions about
what people ordinarily do (and why they do what they do), and so on.
To make my point more concrete,I offer a short 'text' of non-linguistic tools:
hot water
shaving brush and shaving cream
razor
hot water
washcloth and towel
tincture of iodine
bandage.
I have suggested that it should be possible to interpret such a 'text' by know-
ing what the various tools are for. (You wl recall my asking you to be gener-
ous about the category 'tool': not everybody thinks of water or bandages as
tools.) In trying to figure out what chain of activities underlies this record, we
use both our knowledge of typical uses of tools and typical activities of
people. Once we know what such a list stands for, we will readily find ourselves
constructing 'envisionments' of the activity revealed by our list, probably at
more than one 'level', these levels corresponding to the degree of certainty we
feel about the details of the envisionment. At the lowest level (the level at
which we feel the most secure about our hypotheses) we are likely to infer
that somebody was shaving, that while shaving that person drew blood, and
that he or she later patched up the wound with a bandage. If our envision-
ment goes further, we might find ourselves imagining events the record does
not clearly warrant, for example, that the shaver was male, that the area of
the body that got shaved was the face, that this activity took place in the
morning as a part of a more general activity sequence (a man's morning toilet);
and, embellishing further, we might even imagine that feelings of exaspera-
tion or utterings of unsavory pieces of language make up parts of the total
scene.
For any given tool record, of course, the conclusion we draw could be
wrong. With language, however, there is something else we can count on: the
knowledge of communication as an activity with its own methods and pur-
poses. I have made the present record more like a linguistic text than an
ordinary record of tool use might be, since I have deliberately constructed
this list in order to communicate an idea, and you know that that is what I
was doing. Since my argument would not work at all if I were being unco-
operative, you can be sure that the scene I intended you to construct was very
much like the one I described,perhaps without the later embellishments. You
can be sure, too, that if I had decided to surprise you by describing a thorough-
ly bizarre situation that happened by sheer coincidence to call for the tools
I listed here, it would have been because I intended you to be surprised. You
would still know that I intended you to create the interpretation I spelled out
at first, because if I had not, there would have been no point to the surprising
revelation at the end.
Any ongoing interpretation of a text of this kind or the more familiar
kind can be marked with guesses and uncertainties. Such uncertainties, un-
fulfilled expectations, briefly maintained hypotheses, etc., are an important
part of the interpretation process, because they give significance to new infor-
mation as it comes along. In our tool text for example, there was the assump-
tion that the shaving was part of a morning ritual.That assumption would have
been quickly and easily supported if we later found out that our text was an
excerpt from a larger text in which what immediately followed the items in
our passage were a dressing gown, a new ap er, and a coffee pot.
The importance of the tool record analogy is that it provides a clear case
of activities (shaving), natural cause-effect relations (contact with a blade,
getting cut), the functions of artifacts (razors, shaving brushes), standard uses
of natural objects (hot water for washing and moistening), etc., and for which
it is absolutely essential to find coherence in the record by assuming that it re-
presents the traces of some activity and by figuring out what that activity was.
Let us now consider, from the same point of view, a simple linguistic text.
It is a conversational text, which means that in our representation of it, we
need to indicate that different tools were used by different workers, that is,
that different sentences were uttered by different speakers. Here is the text:
A: Can you see the clock?

B: It's stopped.
From the 'semantics' of the complex 'tool' "Can you see the clock?" we infer
that its use must have counted as an instance of A's asking whether some
object that in their surroundings can be uniquely identified as 'the clock' was
visible to B. The situation we most easily envision is one in which A and
and a clock are situated relative to each other in such a way that A cannot see
the clock and cannot tell whether or not can see it.
Now while it is possible to imagine that in the first line of our text A is
merely inquiring about B's visual experiences, it is more likely that A's utter-
ance is a way of accomplishing something else. We know what clocks are used
for, and we know enough about patterns of indirectness in English speech to
suspect that A's utterance was a way of asking to let A know what time was
indicated on the clock's face. Our awareness of that intention, by turn, given
the assumption that the clock is working, gives the utterance the meaning that
A wants to tell A what time it is.
B's answer - "It's stopped" - now has to be interpreted. It does not directly
respond to any of the question-asking intentions we were able to impute to A,
that is, either Can you see it? What time does it show? What time is it? We
have to assume that perceived A's question as a request to know the correct
time. should be able to answer that question if he can see the clock and if
the clock is working; but it isn't, and he informs A of that fact. Assuming,
further, that is being cooperative, we assume he has no other easy way of
determining what time it is; so B's answer can be taken as informing A that
cannot tell A what time it is.
In order to interpret the text, we needed to fit our understanding of what
these sentences are primarily for into some reasonable account of what the
participants in the conversation could be doing. The coherence of the text
could not have been detected had we not been able to accomplish that. And
again, there is another level, having to do with my purposes as author of the
text. I invented this little text in order to make a point. If you came up quick-
ly with the interpretation I sketched out, before I described it myself, it may
have been partly because you knew that I was using the text to support my
view of utterances and utterance fragments as tools.
It is clear from these examples that knowledge of word meanings (know-
ledge of the uses for which tools exist), knowledge of what the world is like
(knowledge of what kinds of things people use tools to build), and knowledge
of possible communicative intentions, all contribute simultaneously to the
'picture of the world' we get when we interpret a text. How much of that
knowledge, we need to ask, can be thought of as pragmatic knowledge? A rea-
son we need to concern ourselves with this question is that we need to know
if there could in principle be a 'contrastive pragmatics'. A notion of 'con-
trastive pragmatics' makes sense if for different languages either distinct con-
ventions exist for 'reading between the lines' (that is, for discovering the mes-
sage in the pairing of an utterance and a context), or if special and distinct
pragmatic purposes have linguistic means dedicated to them in different lan-
guages.
Pragmatic competence, briefly put, consists in being able to make judg-
ments on questions of the following form: In such-and-such a setting, what
could a speaker say which would produce such-and-such an effect? Pragmatics
comprises judgments on the fittingness of particular expression types (words,
grammatical forms, intonations, etc.) to particular situations. These situations
can be roughly (but not mutually exclusively) grouped into those in which
what is predominant is (1) the speaker's feelings or attitudes, (2) the character
of the activity in which the utterance plays a role, and the degree of speciali-
zation of the utterance form to the activity, (3) the nature of the social and
spatial relationships between the participants in the communication event,
and (4) the development of the speaker's activities (and the interpreter's ex-
perience) through time, i.e., from the beginning of the text to the end. Within
the field of pragmatics we find such topics as: expressive language, illocution-
ary force indicating devices, routine formulas, activity-bound speech acts,
politeness levels, registral varieties, style markers, forms of poetic diction,
genre types, evidentiality, presupposition, indirectness, topic marking, devices
for emphasis or downplaying, markers of cohesive links, word order variation,
etc., etc. Whether such a miscellany can be seriously regarded as a single field
is unclear, but that this list incorporates a very great amount of what is in-
volved in knowing how to use a language is beyond doubting. From the reality
that all of the items in this list do indeed vary across languages, they obvious-
ly make up an important part of what needs to be considered in contrastive
linguistics.
3 "Applied pragmatics"
There have been several good reasons for the relative rarity of pragmatic
studies in contrastive linguistics in the past. First, this kind of knowledge has
looked like knowledge that could freely cross linguistic boundaries and that
could vary arbitrarily among speakers of the same language. Hence it was
thought of as having more to do with what people are like, or what cultures
are like, than with what languages are like. Second, this kind of knowledge
appeared to be easier to learn 'in the world' than in the classroom. If it is dif-
ficult to teach something in the classroom but easy to learn it in the world,
that can be taken as a pretty good reason for not devoting a great deal of time
to it in a language teaching program. Third, pragmatic knowledge seemed to
presuppose mastery of the lexical and grammatical resources of a language, so
grammar and lexicon clearly had to take first place. And fourth, it has proved
difficult to provide, at least for many areas of pragmatics, not just an ade-
quate formalism, but even sufficiently satisfying informal subjective descrip-
tions of just what the facts are. Fifth, it has been in general hard to find out
what the components of pragmatic abilities are. The data for pragmatics are
the linking of sentences with the contexts of use in which 'performances' of
them are welcome and fitting, and the evaluation of the nature of the fit be-
tween the sentence and its contexts. Precisely because the notion 'context'
includes so much, attempts to correlate particular linguistic choices with spe-
cific aspects of context are so difficult.
3.1 Listable ragmatic practices

For the purposes of this paper I need to make a distinction between general
pragmatic patterns (large facts) and special pragmatic practices (small facts),
and concentrate on some of the special ones. The large issues include polite-
ness systems, patterns of indirectness, repertories of registral differences, pat-
terns in the rhetorical organization of discourse, the special devices languages
use for constructing narrative texts, and so on. Each of these large issues de-
fines a whole field of its own. The small issues, the ones I prefer to talk about
here, concern things that need to be learned one at a time. They are or (one
would think) ought to be relatively teachable. Here are some examples of
what I mean by small pragmatic facts:
Speakers of Swedish use past-tense verbs in current-experience exclama-
tions. Some Swedes learning English carry the pattern into their new language.
A man you meet out of doors on a very cold morning saying "It was cold to-
day" can create a very strange impression, and surely ought to be told that
English insists on the present tense in that situation. We have here a small
pragmatic practice that can get wrongly transferred from one language to an-
other.
Speakers of English find it perfectly natural to formulate suggestions as
negative w/y-questions, as in "Why don't we go to the opera tonight?". A
structurally identical question in German, I understand, would sound offen-
sive. On the other hand, Japanese negative permission questions are considered
polite ("Gohan moo sukosi itadakemasen ka?" 'Can't we receive some more
rice?'), but their English translations sound sarcastic and rude. Maybe the
phenomena I have pointed out here reflect, in the context of their languages,
really 'large' questions, but it seems nevertheless that they can be treated here
as 'small' questions. They appear to be among the things that, from a lan-
guage learner's point of view, are best learned one piece at a time.
A Japanese acquaintance of mine, rejecting (or not knowing) the English
formula "I am pleased to meet y o u " used instead an English translation of the
standard Japanese formula spoken when meeting someone for the first time.
It came out in English as, "This is the first time I have seen you". While the
remark is unexpectedly flattering to people who assume themselves to be well
known, it strikes others as being very mysterious.
One of the most telling signs of German-speaking background for some-
body speaking English is the uncommonly high frequency of use of the word
already. This comes, of course, from translating German schon as English al-
ready. I have some German friends who are so self-consciously aware of the
mismatch between these two words that they avoid using the word already al-
together. Many people would have reason to be grateful if teachable insights
into the subtle pragmatic differences between these two words could be dis-
covered and publicized.
We are dealing here with linguistic practices that do not 'translate' well;
put differently, we are dealing here with sources of 'nonphonological accent'.
3.2 English formulaic expressions

The most striking kind of small issue in pragmatics is the formulaic utterance.
I have argued elsewhere that formulaic language is of great importance in the
total picture of linguistic competence (Fillmore 1979).
Several years ago my wife and I offered a course, at a summer Linguistic
Institute in Honoloulu, on the subject of 'speech formulas'. In it we explored
ways of describing those phraseologisms which could be seen as having prag-
matic functions. We ended up working within a descriptive framework which
provided for each formula its form, its setting, the antecedent event which
gave rise to its use, the speaker's attitude which its use makes known, and its
function in the ongoing discourse. In addition, we provided usage notes, re-
marks on uses and users that did not get included in any of the categories just
mentioned, as well as a prototype example of the use of the formula, i.e., an
example of a use of the formula in which the context fits the conditions, we
also sought to find a deviant use of the formula, especially for those cases in
which the formula, by being used in a creatively new way, could produce a
humorous response.
As an illustration, take the expression YOU should talk, with YOU stressed
and the other two words destressed. The setting for using this utterance is one
in which A (the speaker of our formula) and (A's interlocutor) are discus-
sing C. The antecedent event is that has just said something unflattering
about C. The speaker's attitude revealed in A's uttering this formula is that
B's criticism of C is unfair, since the same undesirable qualities can be found
in B. The function of the utterance is to scold for remarking on C's faults
while ignoring B's own identical faults. In the usage notes we might add that
the expression is common, at least in the United States. As a prototype ex-
ample we might bring the following conversation.
B: C never has a kind word to say about anybody.
A: You should talk.
(A collection of formulaic expressions annotated in this style is included in
the appendix to this paper.)
We were fortunate in this class to have groups of speakers of several other
languages present, especially Japanese and German, and we were therefore
able to make a number of item-specific as well as systematic comparisons
among several languages. We learned, for example, that Japanese has elab-
orate speech routines in areas calling for apologies, which both German and
English lacked; that English (at least American English) has a large repertory
of non-serious formulas that played an important role as conversational
'greasers', particularly important in a society in which people become super-
ficially intimate very quickly, whereas both Japanese and German seemed to
be 'impoverished' in that regard; and we learned that Yiddish has whole genres
of speech formulas which English, German and Japanese lack, in the realm of
curses, wishes, blessings, and the like, as described in Matisoff's recent book
(Matisoff 1979).
We became aware of three important characteristics of formulaic language.
The first is that it comprises expressions whose pragmatic functions are rela-
tively easy to bring to consciousness. When we asked native speakers to con-
struct situations that called for the use of particular formulaic expressions we
tended to get recognizably similar reports each time and found that most of
what we learned from the informant interviews or questionnaires was relevant
to the pragmatic description of these forms. They tend to be expressions de-
dicated to particular pragmatic purposes. The second is that the 'effect' of
their use can be very subtle, is seldom identical from one language to another,
and can often be the occasion of cross-cultural misunderstandings. We heard
many anecdotes of American travellers offending their foreign hosts in com-
pletely innocent ways. Two separate anecdotes involve the formula I thought
you'd never ask It's a fairly innocent teasing expression in American English,
but it could easily be taken as insulting by people who did not know its spe-
cial status as a routine formula. In one case a European man asked an American
woman to join him in dance, and she, being playful, said, "I thought you'd
never ask". Her potential dancing partner withdrew his invitation in irritation.
In another case a European hostess offered an American guest something to
drink, when he, unilaterally assuming a teasing relationship, said, "I thought
you'd never ask". He was asked to leave the party for having insulted his host.
The implications for contrastive pragmatics are obvious.
These observations lead naturally to the third important characteristic of
formulaic language: since it almost always involves a kind of secondary con-
ventionalization of language, its presence in a text does not stand out. Gen-
erally, if you have misunderstood something because of its 'formulaic' con-
tent you will have no reason to believe that there was anything in the text
that you missed. If somebody uses a noun or a verb that you don't know, you
can know that you've missed something, and you can ask questions to have
the thing you missed explained. In the cases that I have in mind, however,
what you heard can be given an interpretation. It's just that the people who
know conventional formulaic ways of saying things understand aspects of texts
that the more innocent interpreter has no reason to suspect are even there.
3.3 German pragmatic particles

This unawareness of the pragmatic effect of linguistic material in the text is
easy to find in other areas of language, too. James McCawley (lecture, Berke-
ley, California, 1980) has recently proposed a theory of children's language
acquisition in which he attributes to children the theory' that adults in
general have serious speech defects, the main symptom of which is that they
are always putting into their speech little noises that have no communicative
function at all. Mommy, doggy, bed, milk, these are real words, and this is
what language is all about; but there are all those other little noises in adult
speech that mean nothing at all and just have to be ignored. During the period
in which this theory is never disconfirmed, children are not in any way aware
that there is anything they are missing. Gradually, of course, the theory has
to get abandoned, as the children come to realize that these little noises
the, of, i f , at, etc. are not so random and senseless after all.
A year or so ago I realized that in practice, my own attitude toward
German modal particles was very much like that of McCawley's language-
learning children. The occasion was that I found that I had read fifteen or
twenty pages of a German text without ever feeling the need to consult a
dictionary. I had the wonderful feeling that I had understood absolutely
everything in the passage that I had read, and I regarded it as a milestone in
my progress toward competence in this impossible language. In order to rein-

force my delight, I went over the passage again, and, to be sure, I continued
to believe that I understood everything that I saw. But this time through I
became aware of the great number of small words in the text that I had been
willing to skip over. I had paid no attention to them at all, because, as far as
I could tell, I didn't need them: they contributed nothing to the meaning of
the text. These were words like doch, zwar, etwa, nmlich, ja and denn.
Since the linguist in me knew that these words could not actually be mean-
ingless, I decided to do what I could to figure out what they meant. I first de-
cided to consult dictionaries and native speakers. The dictionaries to which I
had access, I quickly learned, gave me no help at all; almost nothing that
I could find in them fitted the examples in my text. My next strategy, which
I was sure would fail, was to see if any of my German-speaking friends could
tell me, in response to direct questions, what these words meant. I got, of
course, the usual kinds of answers about euphony and rhythm, put-them-in
where-they-sound-nice, and so on.
What I did next was to examine English translations of German texts, to see
how these words got translated. Since I have copies of Kafka's Der Prozess in
both languages, I took them out and laid them side by side. In two pages run-
ning I found about a dozen and a half instances of these words in the German
text, but essentially none of them got translated into English. I found myself
wanting to believe that the translator didn't know what they meant, either, or
that they really are meaningless.
In the end I decided to do something with German texts using the analytic
techniques some colleagues and I had been developing in connection with a
research project on the characteristics of the reading process. The technique
involves the presentation of short texts and interviews with native speaker
interpreters designed to discover the ways in which each element of the text
affects the construction of the interpreter's final envisionment of the world
of the text. The technique involves piecemeal presentation of the text, com-
parison of minimally different texts, and so on. In the end we seek to find
principles of text interpretation - involving compositional semantics, con-
sciousness of genre differences, guesses as to authors' goals, recognition of
registral conventions, and so on - which in their workings satisfy us as ac-
counting for the details of the envisionments. (See Fillmore 1982 for a de-
tailed description of the assumptions and methods of the project.)
For this exercise, I chose two little collections of German jokes, and found
some Germans willing to sit down and go through a few of them with me.
Most of the jokes were easy to understand, and in essentially all of them their
humorous point was discernible by interpeters ignorant of the semantic con-
tributions of the particles. The collection contained 577 jokes, in all, of which
318 contained one or more pragmatic particles. In those 318 jokes were
115 instances of doch, 95 of denn, 86 of mal, 49 of ja, 27 of noch, 15 of
wohl, 15 of eigentlich, 12 of eben, 11 of nmlich, and then a smaller number
of instances of blo, einmal, etwa, halt, zwar, jetzt, allerdings, as well as a few
others of whose status I was unsure.
There was too much material, of course, so I concentrated on ja, doch and
nmlich. Here is one of the jokes from p. 54 of Die besten Witze der Zeit
(Kortmann 1977):
Ein zum Tode Verurteilter wird morgens um fnf Uhr geweckt.

"Haben Sie noch einen Wunsch?"
"Ja. Ich mchte gern eine Tasse Kaffee trinken."
"Und wieviel Zucker nehmen Sie?"
"Zucker? Sind Sie verrckt? Ich bin doch Diabetiker."
It is the doch of the last sentence which interests me.

I noticed that in people's reported envisionment of this text the jailer and
the prisoner knew each other, and the jailer was assumed to have known that
the prisoner was diabetic. That knowledge may not need to depend on doch,
of course, since the question, "Are you crazy?" surely suggests something
like, "You should know better than to ask such a stupid question!" I asked
one informant if doch could be replaced by nmlich and was told that it
wouldn't really fit. Then I asked another informant to retell the story with
nmlich in it instead of doch; what happened is that the last line got slightly
changed. It became, "Keinen Zucker, bitte. Ich bin nmlich Diabetiker." And
this time the interpretation was that the jailer was unfamiliar with the pri-
soner's diabetic condition.
The conclusion, based as much on compatibility with the surrounding
textual material as on direct contributions to the text's envisionment, is that
with nmlich one is giving out information which one's interlocutor is assumed
not to have, while with doch one is asserting something which one believes
one's interlocutor wrong or stupid for not knowing or remembering. In either
version of the joke, the words doch or nmlich do not need to be translated
for the joke to work; the rest of the text is sufficient. Their role in these sen-
tences is more to 'fit' the context than to communicate new information.
The next phase of my inquiry was to give informants texts differing only
in the selection of one or another particle and to ask them to invent contexts
for them. Because of the importance of textual surroundings, the most flex-
ible of such 'texts' were in fact single sentences. Here are three that I tried out
for ja, doch and nmlich:
Ich bin ja dein Vater.

Ich bin nmlich dein Vater.
Ich bin doch dein Vater.
All of the contexts people provided for me sorted the examples out along two
dimensions: position in the discourse, and speaker's assumptions about the
hearer's beliefs or feelings. In all three cases, but most explicitly in the case
of nmlich and doch, the utterance was heard as part of an explanation offered
in response to some antecedent event showing a need for an explanation. (One
cannot begin a conversation with, say, the second of the above sentences.)
Other differences had to do with 'mutual knowledge' assumptions. Using A
for speaker and for hearer, we can say that with ja, A knows that knows
the fact which the sentence expresses; in the case of nmlich, A believes that
does not know that information; and with doch, A suggests that the situa-
tion indicates that has beliefs which are incompatible with what the current
sentence expresses, and that should not have those beliefs.
These descriptions, of course, are very much along the lines of description
my earlier project gave to speech formulas. English has some, but not many,
simple words with functions like those of the German particles; but in their
place it has a number of phrasal expressions that appear to have similar func-
tions. I therefore wanted to see whether there were American English collo-
quial expressions that had essentially the same pragmatic conditions. It seemed
that translations that preserved (my understanding of) the pragmatic condi-
tions of the German sentences would be something like these:
I am your father, you know, (ja)

You see, I'm your father. (nmlich)
I am, after all, your father, (doch)
Very many of the particles in my joke books remain opaque to me, and the
collection contains probably as many instances of doch and ja which do not
fit my generalizations as those which do. The pragmatic points to be made
here are (1) that the German particles correspond (however imperfectly) to
fairly subtle parenthetical formulaic expressions in colloquial English, and
(2) that these expressions, if used in English as often as their counterparts
are used in German, would produce very mannered speech. A 'large' prag-
matic fact about German is that the colloquial language welcomes (one might
almost say 'requires') pragmatic particles that reflect choices among the
numerous ways in which individual utterances can be situated in their dis-
course context. English has a number of such expressions, too, but an im-
portant difference is that the corresponding forms in English cannot be gram-
matically incorporated into surface clauses. This difference seems to exist by
virtue of the fact that the German forms are 'particles'while the English forms
are 'formulas'.
3.4 Conclusion
In conclusion it seems that with the 'small' pragmatic principles and practices
that language learners need to know, the elements of their description require
one to pay attention to essentially all of the ingredients of that mysterious
box in the diagram called context: speaker's attitude, speaker/hearer relation-
ships, discourse purposes, institutional setting, events in the surrounding
world, position in an ongoing dicourse, mutual knowledge assumptions, and
all the rest. Even when the 'small' facts are described carefully and accurately
one at a time for two languages, the contrastive aspect of the total description
remains seriously incomplete. These 'small' facts are too often instances of
'larger' practices by which the two languages differ more seriously. The de-
scription of these 'larger' patterns of use appears to me to be about as diffi-
cult to come by as a description of the two cultures. It was not for want of
space that I chose not to deal with such problems in this paper.
4 Appendix
Formula IT'S NOT WHAT YOU THINK.
Setting A innocently finds himself in a situation which, on

the surface, looks, or might be interpreted as being,
incriminating
Antecedent event A has come upon and has noticed that has ob-
served A's situation
Speaker's attitude embarrassment, desire to explain and justify
Function of utterance a request that not form any conclusions based on

the appearance of A's situation; could be followed
by an explanation
Usage notes common; mostly used jocularly, I suspect
Prototype example A's colleague C has invited A into C's hotel room to
examine a manuscript; C's beautiful sister has been
in the room visiting her brother; A and C's sister
leave the room together; A notices B, a close friend
of A's wife; A says, "It's not what you think".
"Creative" departure
Similar formulas DONT JUMP TO CONCLUSIONS
Miscellaneous observations
Formula THIS HURTS ME MORE THAN IT HURTS YOU
Setting A (parent) is about to do something which will be

painful to (child)
Antecedent event -
Speaker's attitude A wishes to assure that A is not being unkind,

that A does not enjoy inflicting pain on B; the act
is necessary
Function of utterance A would say this just while, or just before, perform-
ing the act which will be painful to
Usage notes usually associated with punishment
Prototype example (child) has done something that deserves punish-

ment; A (parent) is about to spank B, but wants it
to be clear that his spanking comes from necessity,
not anger or pleasure (it's for B's own good); A says,
"This hurts me more than it hurts you."
A, about to remove a sliver from B's finger, wants to
warn that it will hurt and wants to share B's pain
"Creative"departure A (dentist), just before drilling B's teeth (B being an

adult patient of A), says, "This hurts me more than
it hurts you." Obviously false, but it serves to cheer
up.
Similar formulas
Formula IT TAKES ONE TO KNOW ONE
Setting A and agree that to call someone an X is derogatory
Antecedent event has just described C as an X; or has just called A

anX
Speaker's attitude A feels that himself has the negative qualities of

anX
Function of utterance A judges B; A chides for attributing negative qual-

ities to others which possesses himself
Usage notes common, probably seldom used seriously
Prototype example says, "I think C is a little queer" OR says,

"You're an idiot, A." A says, "It takes one to know
one."
' 'Creative " departure
Similar formulas LOOK WHO'S TALKING / IT'S A CASE OF THE

POT CALLING THE KETTLE BLACK / CON-
SIDER THE SOURCE
Formula - PRESENT COMPANY EXCEPTED -
Setting conversation, typically involving several people; A is

in the process of making, or has just made, an un-
flattering generalization about a class of people that
includes members of the conversation group
Antecedent event
Speaker's attitude the generalization was too broad
Function of utterance parenthetical remark; A wants it clear that A is not

criticizing the people in this conversation group
Usage notes common; probably mostly humorous
Prototype example A says, "Linguists present company excepted, of

course tend to be extremely boring people."
"Creative"departure A says, "Geniuses, as a class, tend to be very ec-

centric. Present company excepted, of course."
A says in the presence of linguists, "Linguists are
generally brilliant and entertaining people present
company excepted."
Similar formulas
Formula I'LL TELL YOU WHAT (TELL YOU WHAT)
Setting A and have been negotiating about some issue, X,

and have so far not been able to reach any agreement
Antecedent event -
Speaker's attitude what A is about to say will solve the impasse
Function of utterance to introduce a compromise proposal
Usage notes commonly associated with a bargaining context
Prototype example A and have been haggling over the price of a

watch; has offered $10, A has been insisting on
$50; A says, "111 tell you what. I'll let you have it
for $30."
"Creative " departure
Similar formulas I'LL TELL YOU WHAT I'M GONNA DO / HERE'S

MY FINAL OFFER
"I'll tell you what" could be used by either party in a bargaining setting;
"I'll tell you what I'm gonna do" would be used only by the salesperson. I
associate these expressions only with males.
Formula YOU SHOULD TALK
Setting A and are in conversation; is discussing C;

both A and understand that to say X of some-
body is to be critical
Antecedent event has just described C as being X
Speaker's attitude B's criticism is unjust, since too is X
Function of utterance A chides for commenting on C's faults without

recognizing his own
Usage notes common
Prototype example says, "C never has a kind word to say about any-
one." A says, " Y o u should talk."
"Creative " departure
Similar formulas YOU'RE A FINE ONE TO TALK / IT TAKES

ONE TO KNOW ONE / WHO ARE YOU TO
TALK? I LOOK WHO'S TALKING! / PEOPLE
WHO LIVE IN GLASS HOUSES SHOULDNT
THROW STONES
Use of this formula seems to include A's acknowledgment of the aptness of
B's description of C as X. "It takes one to know one" does not include that.
References
Downing, Pamela
1977 "On the creation and use of English compound nouns", Language 53.4:
810-842.
Fillmore, Charles J.
1973 "May we come in?", Semiotica 9:98-115.
1979 "On fluency", Individual differences in language ability and language be-
havior, edited by C. J. Fillmore et al. (New York: Academic Press), 8 5 - 1 0 2 .
1982 "Ideal readers and real readers", Georgetown University Roundtable on Lan-
guages and Linguistics 1981, edited by Deborah Tannen (Georgetown:
Georgetown University Press), 248-270.
1983 "Discussion of Halliday's 'How is a text like a clause?'", Nobel Symposium
on Text Processing, edited by Sture Allen (Stockholm: Almqvist & Wiksell
International) , 2 4 9 - 2 5 8 .
Halliday, . A. K.
1974 "Modes of meaning and modes of expression", Functions and context in
linguistic analysis, edited by D. J. Allerton et al. (Cambridge: Cambridge Uni-
versity Press), 5 7 - 5 9 .
1983 "How is a text like a clause?", Nobel Symposium on Text Processing, edited
by Sture Allen (Stockholm: Almqvist & Wiksell International), 2 0 9 - 2 4 8 .
Kortmann, Erhard
1977 Die besten Witze der Zeit (Hamburg: Hoffmann und Campe).
Lakoff, George P.
1974 "Syntactic amalgams", Papers from the Tenth Regional Meeting of the
Chicago Linguistic Society, 224-244.
Matisoff, James A.
1979 Blessings, curses, hopes and fears: psycho-ostensive expressions in Yiddish
(Philadelphia: Institute for the Study of Human Issues).
Van Valin, Robert D., Jr. - William A. Foley
1980 "Role and reference grammar", Syntax and semantics 13: Current approaches
to syntax, edited by Edith Moravcsik and Jessica Werth (New York: Aca-
demic Press), 329-351.
Zimmer, Karl E.
1972 "Appropriateness conditions for nominal compounds", Working papers on
language universal 8 (Stanford: Stanford University), 3 - 2 0 .
MARIA GRZEGOREK
English sentences with introductory there

and their Polish counterparts
Some examples:
(1) There are no ghosts.
Nie ma duchow.
(2) There was a dog in the garden.
Wogrodzie byl pies.
(3) There developed a serious crisis.
Rozwinqi sif powazny kryzys.
(4) There entered a strange looking man.
Wszedl dziwny czlowiek.
(5) There were beautiful flowers all over the place.
Wszdzie byiy pigkne kwiaty.
(6) There was a rumour that the president was killed.
Byty pogtoski, ze prezydent zostal zabity.
(7) On the platform there were many women who waited for
their husbands.
Na peronie byto wiele kobiet, ktore czekaiy na swoich
m$z0w.
(8) There were several people sick after the party.
? Po przyjeciu kilku ludzi byto chorych.
Po przyj^ciu kilku ludzi zachorowalo.
(9) There was a girl dancing on the roof of one of the cars.
Na dachu jednego samochodw tanczyla (jakas)
dziewczyna.
(10) There were fifty people injured during the riots.
Podczas rozruchow zraniono pi$cdziesiqt osob.
Podczas rozruchow zostalo zranionych pigcdziesiqt osob.
144 Maria Grzegorek
(11) There is no time to lose.

Nie ma czasu do stracenia.
(12) There is a difference between these two signs.

Jest roznica pomifdzy tymi dwoma znakami.
(13) There is that book about roses, of course.
Jest jeszcze, oczywiscie, ta ksiqzka rozach.
(14) There stepped out in front of the car a small child.
Przed samochod wyskoczyto mate dziecko.
(15) All of a sudden there burst in his younger sister Jane.
Nagle wpadla jego mlodsza siostra Janka.
0 An introductory note
Since rAere-constructions are called themeless in this paper, it may be useful

to explain how this term is used and how the term theme is defined. 1
Following Halliday (1967) we treat theme as the starting point in the
organization of the message. If we agree that on the linguistic level basic mes-
sages correspond to clauses, we can characterize the notion of theme for each
language in syntactic terms. In English, theme is the expression functioning
as the left-most constituent of a clause if this expression is capable of refer-
ential use or at least capable of being contextually bound. According to this
definition all expression phrases such as Thank God, Damn it, etc. and all sen-
tence adverbials in clause-initial position do not count as themes because they
are syntactically outside the clause and they are incapable of referential use.
Introductory there and it, even though they are syntactic constituents of
the clauses in which they occur, cannot be treated as themes because they
cannot be used to make reference to extralinguistic reality or to the preceding
text. There and it have, however, thematic relevance because they mark
clauses as themeless, i.e. they signal that the speaker chooses not to begin the
message with any content word and that he prefers to present the whole in-
formation in the rhematic part treating all components of the message in the
same way as far as the starting point is concerned.
1 English sentences with introductory there
Sentences with there in initial position do not constitute a homogeneous class

either from the syntactic or from the semantic point of view. Also, factors in-
/Aere-sentences in English and Polish 145
fluencing the speaker's choice of a themeless fftere-construction instead of a

corresponding thematic structure are of several different types. There are,
however, things that all tfiere-sentences have in common: 1. their subject NPs
are focus elements, 2. their verbs have a lower degree of communicative dyna-
mism, i.e. the order is: verb - subject.
There is another thing that all f/iere-constructions have in common (related
to the focal status of the subject NP) the subject NP, whether definite or
indefinite, is nonanaphoric in a broad sense, i.e. its referent is not in the
listener's consciousness at the moment of speaking. 2 In fact, the purpose of
uttering a there-sentence is to bring the referent of the 'displaced' subject into
the listener's consciousness, either by asserting its existence or simply by call-
ing the listener's attention to it (cf. van Oosten 1978).
Superficially all there-sentences represent the same thematic type ac-
cording to our definition they are themeless utterances. A more serious con-
sideration shows that we can distinguish two main types of there-sentences:
1. those which are obligatorily themeless by virtue of rules of grammar
(grammar of English offers no other choice),
2. those tfzere-constructions which have thematic counterparts, and the
choice between thematic and non-thematic construction depends on typical
factors such as context and the way the speaker presents a given event.
Type one is represented by bare existential sentences and probably also by
locative existential sentences (examples 1 - 7 ) . Type two is represented by pre-
sentational there-sentences such as those in examples 14 and 15.
The order of elements in existential sentences follows the degree of com-
municative dynamism: be which contributes very little to the message pre-
cedes the information focus - the indefinite NP. This order ist not contextu-
ally motivated because both be and the indefinite NP are contextually inde-
pendent. We may risk the statement that it is the meaning of the verb be, and
other verbs of being, that determines the order of syntactic elements in
existential sentences.
It is not clear how existential sentences should be derived. We may assume,
as many linguists do, that there are at least two general sets of rules respon-
sible for final ordering of words in a given language: 1. linearization rules by
which the basic order (SVO, VSO, etc.) is assigned to declarative sentences,
2. postcyclic rules which are responsible for pragmatic rearrangements (cf.
Pullum 1977). As far as the derivation of existential sentences is concerned,
several solutions can be proposed.
1. Word order in existential sentences is assigned by linearization rules.
This means that apart from linearization rules which assign the order SV(0)
in English, a different rule assigns the order VS for sentences with verbs of
being. In terms of PS rules it means that we need an additional rule: S -* V
146 Maria Grzegorek
NP if V is a verb of existence. There is inserted according to a general surface

structure constraint for English: no declarative sentence may begin with a
finite verb.
2. Another solution is that linearization rules assign the order SV(O) to all
English sentences. The order in existential sentences gets reversed by the post-
cyclic rules. There is then inserted on the basis of the same surface structure
constraint as above.
This solution is undesirable for several reasons. The first one is that the as-
sumption is made that (a) and (b) sentences in examples 16 and 17 are synon-
ymous, which, as we claim in one of the next sections of this paper, is not true.
(16) a. There is a strange man outside,

b. A strange man is outside.
(17) a. There is a mouse in the kitchen,
b. A mouse is in the kitchen.
The second problem with this solution concerns the so-called sourceless
existential sentences such as 18 and 19.
(18) There is a space in this room.

* A space is in this room.
* No ghosts are.
Another reason why this solution is not the best one is that it suggests that
the change of order in existential sentences is caused by the same type of
rules as other pragmatic reorderings such as stylistically motivated word order
or emphatic thematization. As is well known, existential f/2 ere-constructions
are stylistically and otherwise communicatively unmarked. In fact, it is their
alleged sources which are marked.
3. The third solution is to allow the element there to be the subject in the
underlying sentence, i.e. there could be introduced by the PS rules (as in
Jenkins 1975). Apart from theoretical objections (cf. Milsark 1974), some
other objections are that this approach eliminates the difference between
existential and presentational there-sentences and that it hides the fact that
the verb in presentational sentences is put in the position preceding the sub-
ject for strictly pragmatic reasons.
Summing up, we choose the first solution as the best one and claim that
the order in existential sentences V NP is due to basic linearization rules of
English.
there-sentences in English and Polish 147
Presentational sentences are, on the other hand, postcyclic rearrangements

of the basic SV(0) order. Accordingly, for every presentational sentence we
have a corresponding thematic sentence.
(20) There stepped in front of the car a small child.
A small child stepped in front of the car.
(21) There came into his view Irene.
Irene came into his view.
As predicted by this approach, presentational sentences are the marked ones
when compared with their counterparts with themes. This claim is supported
by their embedding behaviour. Presentational sentences behave like 'root per-
mutations', whereas real existential sentences are easily embeddable. (Cf.
Aissen 1975 and her examples, 22 and 23:
(22) a. *The driver regrets that as he was about to stop, there

stepped in front of his car a pedestrian,
b. *The driver regrets that as he was about to stop from
behind the bushes stepped two children.
(23) We regret that there is no job here.
Syntactic classification
77jere-constructions have been classified in several ways by various linguists
(cf. Ross 1974, Milsark 1974, van Oosten 1978). Here we will only mention
the main criteria which have to be taken into account in any classification of
these sentences.
1. Type of the main verb
a. be3
b. verbs other than be but of related meaning, e.g. appear, emerge, occur,
develop, remain4
c. some stative verbs describing location (hang, lie, stand) and some verbs
of directional movement (step out, burst in, walk into).
2. The structure of the NP following the verb
a. without a codicil, e.g. unmodified NP as in examples 1 and 2; NP with a
prenominal modifier (examples 3, 4, 10); NP with a complement (16).
b. with a codicil. Codicil can be of several types:
- prepositional phrase s (examples 12,13)
- relative clause (7)
- adjective6 (8)
148 Maria Grzegoiek
- present or past participle of the verb together with the complement

of that verb (9,10)
infinitive (11).
3. Presence of a locative phrase (or time adverbial)
a. locative (time) phrase in postnominal position (examples 2, 5, 8, 10)
b. locative in thematic position (7)
c. locative in prenominal (before the subject NP) position (14).
4. Type of the determiner preceding the subject NP
a. definite NP (3, 5)
b. indefinite NP (all examples except for 3 and 5).
Semantic classification
The only f/jere-constructions which are considered to be unquestionably
existential are the so-called "bare existential sentences", i.e. sentences in
which the subject NP is not followed by any locative or time adverbial, and
also sentences with locative or time adverbial in the thematic or final position.
But even among sentences which look like bare existential sentences there are
those which do not assert existence as such. For instance, as noticed by van
Oosten (1978) sentences such as 24:
(24) There is even a garden.
do not assert that a garden exists somewhere in the universe. By uttering sen-
tence 24 the speaker wants to draw the listener's attention to the existence of
a garden in the frame of reference of their current discourse. In a similar way,
sentence 25 is interpreted:
(25) There's plenty of time.
Either one of the two functions or both (because they are related) underlies
the use of f/zere-constructions:
1. to assert the existence of the referent of the displaced subject NP,
2. to raise the referent of the subject to the addressee's consciousness.
True existential sentences assert existence (plus also location or time). The
non-existential f/iere-constructions, such as presentational sentences or lists,
perform primarily the second of these two functions.
If we extend the notion of 'asserting existence' to cover also cases of bring-
ing into existence in the listener's mind, we can (following the idea of van
Oosten 1978) establish a-scale of existential sentences according to the de-
gree to which the existential assertion is expressed. 7 If we start from the
"pure" existential we have the following order:
rAere-sentences in English and Polish 149
1. existential sentences
there - be indef. NP codicil
2. locative-existential sentences
there - be - indef. NP - codicil locative
3. existential sentences with verbs of existence other than be
there verb indef. NP codicil locative
4. presentation /iere-constructions with stative verbs
there - verb - | ^ ^ j NP - locative
or
there - verb - locative - NP
5. list r/iere-constructions
6. presentational sentences with active verbs

there verb - locative - | ^ j 6 ^ J NP
Below we discuss in a more detailed way the semantics and pragmatics of
these different types of fAere-constructions.
1.1 Existential sentences
Bare existential sentences, i.e. those in which the subject NP is not followed
by a locative phrase, are always themeless. The speaker has no other choice,
except to choose a different verb:
(27) *No ghosts are.
*Ghosts are not.
(28) Ghosts do not exist.
26 and 28 are not textually equivalent. In 28 the noun phrase ghosts is ana-
phoric, probably mentioned or implied in the context or co-situation. Accord-
ingly, ghosts in 28 carries a lower degree of communicative dynamism than
the verb following it.
Generally bare existential sentences assert the existence or non-existence of
some object with some property described in the coda or prenominal modi-
fier. Locative-existential sentences assert the existence of a given object in
some place, i.e. they assert both existence and location. The locative phrase
may belong to the focus, in which case intonation pattern (a) is used, or it
may be contextually bound, in which case the intonation pattern (b) is used.
150 Maria Grzegorek
Only in the latter case the sentence is informationally equivalent to a sentence

with a thematic locative phrase such as 31.
a -
(29) There's a strange looking woman in the house.
(30) There's a strange looking woman in the house.

(31) In the house, there is a strange looking woman.
Locative-existential sentences present two problems:
1. What is their relationship to sentences in which the focal NP is in the
semantic position, i.e. what is the relationship of 30 to 32:
(32) A strange looking woman is in the house.
2. What is the relationship of locative-existential sentences with the locative
phrase in thematic position such as 31 to those in which there has been
omitted, as in 33:
(33) In the house is a strange looking woman.
Many linguists claim that both types of sentences (30 and 33) are existential
(cf. Breivik 1975, Kirkwood 1977, Erdmann 1978) whereas others treat only
sentences with introductory there as existential. Some of them even claim
that the locution there is by itself introduces the existential interpretation (cf.
Allan 1971).
We will argue here that a sentence of the type:
a. indef. NP - be locative
(An account book is on the table)
is not existential and hence not synonymous to:
b. there - be indef. NP locative
(There is an account book on the table.).
The crucial difference between these two types of sentences is that only (b)
asserts existence of the referent of the subject NP, and in fact this assertion is
the main purpose of uttering such sentences. In (a) the purpose of the speaker
is not to assert the existence of the referent of the indefinite NP (an account
book) but rather its location. The existence of the referent and its identity are
established by the addressee on the basis of the relationship between the sub-
ject and the predicate phrase.8 For instance, it has been observed by Breivik
(1975) that for sentences of type (a) to be acceptable at all the locative
phrase must be [+ deictic], i.e. it must refer to some well-defined, usually
small area, known to the addressee. Only then is the spatial relationship be-
tween the referent of the subject and the location clearly defined and hence
the referent of the subject is identifiable. Notice that the first sentence in ex-
amples 3436 is unacceptable, wheras the (b) sentences are good English sen-
tences because in /zere-constructions the requirement that the referent of the
subject be identifiable by the listener does not hold (you do not have to iden-
tify the object if your purpose is merely to assert its existence).
(34) a. *A book is probably on the table.
b. There is probably a book on the table.
(35) a. *Lions are in Africa.
b. There are lions in Africa.
(36) a. *A famous cathedral is in Guildford.
b. There is a famous cathedral in Guildford.
(examples from Breivik 1975)
Another requirement on sentences of type (a) is that the subject NP must be

capable of referential use, i.e. it must be specific rather than abstract. There is
no such condition for r/z ere-sentences. 9 Consider these examples:
(37) a. I I is in the manger.
b. There is | s P a c e I in the manger.

[ corn J
(38) a. I i l ? ^ " 8 is in this house.

[ *No sign of life J
b. There is I n o ^ n 8 J in this house.

I no sign of life j
Those sentences of type (a) which are acceptable are rare and are stylistically
marked. They are strange out of context. They require very specific contexts
such as, for instance, stage directions and all contexts in which there is a de-
tailed physical setting for the referent of the subject NP(as in the example 39,
from Breivik's article):
(39) We are looking at an extraordinary picture painted by X.

A middle-aged man and three boys are seated on chairs or
stools at a spindly table. An account book is on the table.
The problem of lack of corresponding sentences of type (a) for each sentence
of type (b) has always bothered the proponents of the there-insertion hypo-
thesis, according to which both sentence types have the same underlying
structure: indef. NP be locative. Only sentences with participial coda al-
ways have counterparts with indefinite NPs in thematic position.
152 Maria Grzegorek
(40) a. There was a demonstrator killed by a policeman in Oak-

land.
b. A demonstrator was killed by a policeman in Oakland.
(41) a. There is a girl playing on the roof of your house.
b. A girl is playing on the roof of your house.
It is clear, however, that the (a) members of these examples do not assert the
same as the (b) members, i.e. (b) sentences are not existential. Furthermore,
(a) examples are ambiguous between two readings: 1. assertion of the occur-
rence of an event, in which case they have the meaning close to (b), 2. asser-
tion of existence or calling to the addressee's attention the referent of the in-
definite NP, in which case the sentence is paraphrasable with the verb exist (A
demonstrator who was killed by a policeman existed). That sentences such as
40 a are ambiguous can be shown by two types of contexts in which they can
occur. For instance, 40a can be followed by either (c) existential reading,
or (d) 'event' reading, as the prosentential it indicates.
(40) a. There was a demonstrator killed by a policeman in Oak-
land.
c. He was my friend's brother.
d. It happened on Sunday.
According to our approach in this paper, there-constructions and their counter-
parts with indefinite NPs in thematic position are not cognitively synonymous
because they are used to make two different types of assertions - there-sen-
tences assert existence (plus location), whereas sentences with indefinite sub-
jects assert location. Thus these two sentence types are not thematic variants
of the same proposition.
Some locative-existential sentences have, however, several thematic pos-
sibilities:10
1. a themeless clause, e.g.
(41) a. There are many good theatres in London.
(42) a. There is a big hole in Jim's jacket.
2. with a preposed locative phrase
(41) b. In London there are many good theatres.
(42) b. In Jim's jacket there is a big hole.
3. with a locative in thematic position functioning as subject
(41) c. London has many good theatres.
(42) c. Jim's jacket has a big hole in it.
The pragmatic conditions on the use of either of these three versions are dif-
ferent. Only in case 1 can the locative phrase be in the scope of the focus, i.e.
presentecLas part of the new information. In cases 2 and 3 (examples 41b,
41 c, 42b, 42c) the locative phrase is presented as the given information, but
in each case the speaker has a different view of the state of affairs. In 4 2 b the
jacket is viewed more as a location and in 42c more as an object with some
property.
The relationship of sentences with preposed locative phrases with there to
sentences with preposed locative phrases without there is not clear. Breivik
(1975) noticed that there cannot be omitted under the same conditions
under which the 'paraphrase' of existential sentences with thematic indefinite
NP cannot be used. Hence sentences 4 3 - 4 5 are unacceptable (just like sen-
tences 34a, 35a and 36a are).
(43) *In the house was no sign of life.

(44) *In Africa are lions.
(45) *On the table is probably a book.
There can be omitted if location is specific enough. Consider for instance this
example from Breivik (1975):
(46) The old rocks which make up Wales stretch eastward into
the English counties of Shropshire and Herefordshire. . . .
In the North are the rugged mountains (several exceeding
3,000 feet in height) of Snowdonia; in the centre and South
are extensive moorlands.
Breivik (1975) treats sentences with omitted there as variants of those with
there, i. e. sentences 47a and 4 7 b express the same proposition. 11
(47) a. On the table was a book.

b. On the table there was a book.
It seems to me, however, that the same restrictions on sentence types:
Indef. NP - be locative
(Two books are on the table)
Locative - be - indef. NP
(On the table are two books)
point out to the fact that these two are thematic variants of the same propo-
sition. None of them expresses an existential statement. On the other hand,
the two types of sentences below:
There - be NP locative
(There are two books on the table)
154 Maria Grzegorek
Locative there - be NP
(On the table there are two books)
are thematic variants of the same existential proposition (providing on the
table is outside the scope of the focus NP, because only then is it informa-
tionally equivalent to the preposed locative phrase).
1.2 77iere-constructions as 'lists'

In the list use of /iere-contructions the focus NP can be definite or indefinite,
e.g-
(48) A: How could we get there?
B: There's the trolley . . .
(49) A: What's worth visiting here?
B: There's the park, a very nice restaurant, and the library.
Even if the focus NP is definite, still the general condition of tfiere-construc-
tions is met because this NP is non-anaphoric. Rando and Napoli (1978) ex-
plain this fact in the following way:
'Existential' there-sentences typically allow only indefinite NP arguments,
while 'list' there-sentences accept both definites and indefinites. The rea-
son for this difference is that the argument of an existential there-sentence
is the NP itself; but the argument of a list i&ere-sentence is the list, not the
individual members comprising that list. (Rando Napoli 1978:300)
Thus, even if the items of the list are all anaphoric, the list, i.e. the choice of
the items, is the new information. This is true also if the list consists of only
one member as in 48. List /fcere-sentences do not assert existence. They can
be paraphrased as "one could mention . ..", i.e. they only bring the referent
of the focus NP to the addressee's consciousness.
1.3 Presentational i/iere-sentences

As we have already mentioned it is the presentational f/zere-construction that
is 'marked' when compared with its counterpart without there. A presenta-
tional construction is used when the speaker wants to present things in the
order in which he perceives (or perceived) them. According to Kimball (1973)
the referent of the subject NP is seen from the speaker's point of view as
"coming into the perceptual field of the speaker" (Kimball 1973:265). Kim-
ball's proposal was aimed at explaining why only a specific group of verbs al-
lows the presentational order. Indeed active verbs such as run out, step in,
burst in indicate the appearance on the scene and they can be used in there-
there -sentences in English and Polish 155
constructions. The same explanation can be extended to stative verbs (cf. van
Oosten 1978).
If presentational sentences are looked upon as representing the speaker's

point of view in the way referred to by Kimball, it becomes clear why they
are sometimes preferred to their counterparts with normal word order, for in-
stance why 50a is preferred to 50b:
(50) a. There stepped out in front of the car a little girl,

b. A little girl stepped out in front of the car.
A little girl is the last element of the picture perceived by the speaker. It could
not be put in the thematic position because it was not 'available' when the
speaker started his message.
2 Counterparts of English /ftere-constructions in Polish
Even without a more detailed analysis two things can be said about the struc-
ture of Polish counterparts of English there-sentences, especially existential
sentences:
1. There is no empty subject in Polish in any counterpart of any type of
English i/iere-constructions,
2. In both languages the relative order of the verb and the notional subject is the
same, i.e. verb - subject and follows the principle of arranging lexical items
according to the increasing degree of C.D. (communicative dynamism).
ad 1. - the problem of empty subject
The occurrence of empty subject in English existential sentences and its non-
occurrence in the Polish counterparts of these sentences are instances of more
general rules of English and Polish grammars. The scope of these rules is
broader than the domain of existential sentences. These rules can be inform-
ally stated in the following way:
English. The basic word order of English declarative sentences is SV(O).
Any re-arragements of lexical items (motivated by pragmatic considerations
such as context, co-situation, speaker's intentions, choice of discourse topic,
etc.) tend to re-establish that order, e.g. Passive I, Passive II, Pseudopassive,
to some extent also Subject-to-subject Raising and Tough-movement.
The statement that no English declarative sentence may have the order
verb - subject would be too strong because this order is found in the so-
called "root permutations" to which also presentational r/iere-sentences be-
156 Maria Grzegorek
long. A more adequate formulation is that no English declarative sentence

may begin with a finite verb, i.e. if there is no content word preceding the
verb a lexically empty formative functioning as a slot filler is introduced.
Polish. Although the basic word order is the same as in English: SV(O),
after pragmatically motivated rearrangements of lexical items according to
their relative degree of C.D., the verb may be shifted to the initial position
and it may be left there because in Polish the structure subject verb in this
order does not have to be preserved. No slot filler is necessary.
It is worth mentioning that although generally the function of word order
in English is to signal syntactic functions of lexical items, and in Polish to
signal the communicative function of the lexical items (old versus new in-
formation), in the case of sentences with introductory there and it (i.e. theme-
less clauses) the function of word order is not as much grammatical as com-
municative. Besides, the evidence that 'demotion' of the notional subject in
English changes its syntactic status is nonconclusive (cf. Green 1977), and the
semantic interpretation rules have to interpret the demoted subject as subject
even if it is in post-copular position.
ad 2. the problem of word order
The relative order of subject and verb in existential sentences is the same in
English and Polish (and in some other languages). Although this order dis-
agrees with the basic word order in declarative sentences in both languages, it
is still the unmarked order of existential sentences in English and Polish, i.e.
it is not motivated by contextual factors but rather by inherent semantic
features of verbs of being which are such that these verbs always have a lower
degree of C.D. than any noun accompanying them.
Since in polish word order always tends to reflect the degree of C.D., the
fact that the order of existential sentences is verb - subject is nothing ex-
ceptional in this language. The only thing which seems worth stressing is the
fact that in existential sentences this particular order is lexically rather than
contextually governed. But if in a particular context the verb bye 'be' be-
comes the information focus, it is put in the position after the subject (i.e. ac-
cording to the degree of C.D.) in Polish whereas in English such change is not
possible:
English. A: Do you think that there is no God?

: No, I think that ( God exists
< ? there is a God >
[ * God is j
Polish. A: Czy myslisz wife, ie nie ma Boga?

B: Nie, sqdzf, it ( Bog istnieje \
\ ? ? jest Bog i
I Bog jest J
Since, however, without specific contexts such as this, the order subject
verb is unacceptable in Polish existential sentences, the same linearization rule
as for English should be postulated : S -> V NP (if V = verb of being).
2.1 Counterparts of English existential sentences
2.1.1 Bare existential sentences
English: There - V ex - NP
Polish: Vex -NP
or
lexical
function 1 2
Examples:
(1) There are cannibals.
S%/istniej$ ludozercy.
(2) There are no roses without thorns.

Nie ma rz bez kolcow.13
(3) There was a shot.
PadIstrzai. (fall')
(4) There was a long silence.

Zapadlo dtugie milczenie. ('fall')
(5) There was no frost.

Nie czulo mrozu. ('feel')
(6) There was no sound of music.

Nie stychad by I muzyki. Chear')
2.1.2 Locative-existential sentences
English: There - be - NP - { ^ f * j
158 Maria Grzegorek
locative
- there - be- NP
time
Locative I
Polish: - bye - NP
time J
or
lexical
function
Examples:
(7) There were many old women on the platform.

Na peronie bylo wiele starych kobiet.
(8) On the platform there were many old women who came to
wait for their husbands.
Na peronie bylo wiele starych kobiet, ktore przyszty tutaj,
azeby czekac na swoich mzow.
(9) In the Middle Ages there was a great difference between the
poor and the rich.
W Sredniowieczu byla wielka roznica pomi$dzy bogatymi
i biednymi.
(10) There is no longest sentence in a natural language.

Wjzyku naturalnym nie ma najdluzszego zdania.
(11) There was a passage here.
Znajdowal si? tu pasaz. ('be present')
(12) There was a wedding last week.
Tydzien temu odbyto sif wesele. ('be held, take place')
(13) There was a bookcase in the corner of the room.
W kqcie pokoju stata biblioteczka. ('stand')
(14) All over the place there were bits of paper.
Wszgdzie lezaly kawaiki papieru. ('lie')
(15) There was a map on the wall.
Na scianie wisiaia mapa. ('hang')
(16) There wasn't a single star in the sky.
Na niebie nie swiecila anijedna gwiazda. ('shine')
(17) There was a stream along the road.
Wzdluz drogi ptynî strumien. ('flow')
fAere-sentences in English and Polish 159
(18) There was a swing band on the radio.

W radio grata orkiestra taneczna. ('play')
(19) There was grass on the plain.
Rownine porastaia trawa. ('grow')
(20) There wasn't a drop of rain in two weeks.

Przez dwa tygodnie nie spadia ani kropla deszczu. (fall')
(21) There were patches on the slope.
Na sharpie widad by to plamy. ('see')
The differences between English and Polish existential sentences concern two
problems:
1. the position of the adverbial,
2. the lexical value of the verb.
ad 1 - position of the adverbial

It is generally accepted that adverbials in terms of their communicative func-
tion can be divided into several groups: 1. setting adverbials which provide the
background information about the place or time of the event (but the main
interest of the speaker and the hearer lies in the event itself), 2. specification
adverbials 14 which are part of the new information, 3. those adverbials which
function as information foci.
Setting adverbials may be either contextually bound or context independ-
ent but they always have a low degree of C.D. In both English and Polish they
are usually put in sentence initial position, whereas specification adverbials and
those functioning as information foci are in the rhematic part of the sentence.
In locative-existential sentences the locative (or temporal) adverbial is not
a setting adverbial because it is indispensable, i.e. these sentences are used to
assert both existence and location of the referent of the subject NP. Thus ad-
verbials in existential sentences are part of the new information (but not in-
formation foci). The important distinction is between those which are con-
textually bound and those which are unbound. In English a contextually
bound locative phrase may occur in clause initial position (marked construc-
tion) or in clause final position (neutral word order). A contextually unbound
adverbial occurs in English in clause final position, with a different intonation
pattern (cf. remarks in part one). In Polish the adverbial in existential sen-
tences occurs always in initial position. Probably because of its low degree of
C.D. the adverbial is chosen as the 'starting point', i.e. theme, whereas the
160 Maria Grzegorek
notional subject being nonanaphoric (or noncoreferential in Szwedek's terms)

occupies the clause final position and bears the main sentence stress.The orders:
V locative - NP (By to na peronie wiele kobiet)
V NP locative (Byio wiele kobiet na peronie)
are stylistically highly marked.
Existential sentences with subjects in initial position are unacceptable be-
cause in this position noun phrases allow only anaphoric interpretation in
Polish (subjects of existential sentences are always nonanaphoric). Even noun
phrases which are overtly marked as nonanaphoric by words such as jakis
'some', pewien 'a certain' are rarely acceptable in this position:
(22) ?? Jakies stare kobiety byly na peronie.
'Some old women were on the platform.'
(23) ?? Jakas mapa wisiala na scianie.
certain map was on the wall.'
The only exception are noun phrases which are inherently indefinite. For
these nouns and pronouns the order subject - verb - locative is better than
the typical order for locative-existential sentences in Polish, e.g.
(24) Ktos byl przy drzwiach.
'Someone was at the door.'
(25) ? Przy drzwiach byl ktos.
'At the door was someone.'
ad 2 the lexical value of the verb

Many Polish equivalents of English existential and locative-existential sen-
tences have a verb with 'full meaning' instead of the lexically empty (or al-
most empty) bye 'be'. l s In fact, only a limited class of verbs can replace bye.
Which verb is chosen depends on the semantic features of the subject noun
phrase.
Generally these verbs can be classified into four types:
a. verbs of location
Among these verbs we can distinguish two subtypes:
synonyms of bye: znajdowac si$ (used when the subject NP denotes a
physical object as in example 11), odbyc si$ (used if the subject NP denotes
an event as in example 12).
verbs which characterize the typical manner of spatial existence of the
referent of the subject NP (examples 13-15):
f/iere-sentences in English and Polish 161
slup, szafa - stoi

'pillar, wardrobe stands'
obraz, lampa - wisi
'picture, lamp - hangs'
papier, ksiqzka - lezy
'paper, book - lies'
b. verbs which characterize the typical action or state of the referent of the
subject NP (examples 1620):
gwiazdy - swiecq
'stars - shine'
orkiestra - gra
'band - plays'
trawa - rosnie
'grass - grows'
deszcz - pada
'rain drops - fall'
strumien - piynie
'stream - flows'.
c. verbs of perception in impersonal form.
The verb corresponds to the senses by means of which we perceive the referent
of the subject (existence is asserted indirectly via sensory effects examples
21,5,6):
plamy - widac
'patches - one can see'
muzyka - slychac
'music one can hear'
mroz - czuc
'frost - one can feel'.
d. verbs which are idiomatically associated with the subject noun, i.e. in
collocation with a given noun the verb functions as a synonym of bye:
cisza - zapasc
'silence - fall'
strzal - pasc
'shot - fall'.
Similar observations concerning the occurrence of lexical verbs in Russian

existential sentences were made by Babby (1978). 1 6 Babby observed that in
Russian (and the same is true in Polish), given the meaning of the subject and
162 Maria Grzegorek
the fact that the sentence is existential, the choice of the lexical verb is pre-
dictable. From the point of view of their communicative function these verbs
have a lower degree of C.D. than the subject NP and are thus put in the posi-
tion preceding the subject, like the verb bye.
2.2 Counterparts of English itere-constructions with present and past parti-

ciple following the subject NP
2.2.1 Present participle

English: There - be NP - pres. part, of V - locative
Polish: Locative - V - NP
Examples:
(26) There is something strange going on in that room.

W tamtym pokoju dzieje sif cos dziwnego.
(27) There is a man dancing on the roof of your house.
Na dachu twego domu tanczy (jakis) cztowiek.
(28) There were a lot of idiots swimming in that river.
W tej rzece plywato wielu idiotow.
(29) There is operating in our language, as in most forms of
human behaviour, a principle of least effort, in this case a
principle of maximum illocutionary ends with minimum
phonetic effort (Searle 1971:50).
W naszym jgzyku, jak w wigkszosci form ludzkiego zacho-
wania, dziaia zasada jak najmniejszego wysitku.. .
2.2.2 Past participle

English: There - be - NP - past. part, of V (by NP) - loc
Polish: Loc - NP - zostac - past part, of V (przez NP)
or
Loc Vjnpa-sonai NP
Examples:
(30) There were several people hit by tomatoes at the concert.
Podczas koncertu uderzono kilka osob pomidorami.
(31) There were two children injured during last riots.

Podczas ostatnich rozruchow zostalo zranionych dwoje
dzieci.
Podczas ostatnich rozruchow zraniono dwoje dzieci.
(32) Last year there were three black men killed by the police.
W ciqgu ostatniego roku policja zabita trzech czarnych
m^zczyzn.
W ciqgu ostatniego roku zostali zabici przez policy trzej
czarni mzczyzni.
(33) At the meeting there were discussed various problems which
were new for John and for his colleagues.
Podczas spotkania dyskutowano wiele problemow, ktore
byly nowe dla Janka ijego kolegow.
The English Mere-constructions with participial forms can have two readings:
the assertion of the occurrence of the event (in nonthematic form) and the
assertion of the existence of the referent of the indefinite NP. 1 7
The possibility of two interpretations is due to the structural ambiguity of
these constructions. The present participle may be treated as part of the pro-
gressive form (the event reading) or as a postnominal modifier of the inde-
finite NP (the existential reading). Similarly, the past participle may be ana-
lyzed as part of the passive form of the verb (the event reading) or as a post-
nominal modifer of the indefinite NP (the existential reading). All English
Mere-constructions with participal forms have thematic counterparts, with
the indefinite NP functioning as theme (but only for the event reading), e.g.:
(28) a. There were a lot of idiots swimming in that river,

b. A lot of idiots were swimming in that river.
(31) a. There were two children injured during last riots,
b. Two children were injured during last riots.
Sentences 29 and 33 are different. The word order is different than in the rest
of sentences with participial forms because in these sentences the requirement
of an independent principle, i.e. the 'heavy end' principle, motivates the shift
of the subject NP to the position following the participle. It seems that such
sentences allow only the event interpretation.
77iere-sentences with participial forms are translated into Polish like other
English sentences with the progressive aspect or passive voice, respectively.
Thus sentences with present participle are translated in Polish as sentences
with verbs in imperfective form, whereas sentences with past participles are
164 Maria Grzegorek
translated as sentences with the verb in passive form with the auxiliary zostac
(rarely with the auxiliary bye).
The order of lexical elements in Polish counterparts of English there-con-
structions with participial forms signals their relative degree of C. D. plus the
fact that the subject NP is nonanaphoric (hence generally is in postverbal po-
sition, corresponding to the postcopular position in English). The only thing
worth mentioning is that the Polish equivalents do not allow any existential
interpretation because the verb bye does not occur in them.
If in the Polish translation of an English there-sentence with past participle
the passive auxiliary bye is used, the existential interpretation is possible. It
seems, however, that bye rather than zostac is used if the past participle re-
sembles adjectives, i.e. if there is no przez-phrase {by-phrase). Moreover, the
restrictions here are similar to the restrictions on the occurrence of post-
nominal adjectives in sentences with expletive there and their Polish counter-
parts, i.e. only those past participles are allowed which are followed by a
complement:
(34) There were many children left at home.
Pozostawiono wiele dzieci w domu.
? By to wiele dzieci pozostawionych w domu.
(35) There were many children left without food and water there.
By I tarn wiele dzieci pozostawionych bez jedzenia i wody.
2.3 Counterparts of English presentational sentences with introductory there

Examples:
- with active verbs:
English: There - V - locative - NP
(36) Then there ran out of the bushes a grizzly bear.
Wtedy wyskoczyl krzakw szary niedzwiedz.
Wtedy krzakw wyskoczyt szary niedzwiedz.
(37) Suddenly there jumped out of the hole a rabbit.
Nagle dziury wyskoczyl krolik.
with stative verbs:
English: There - V - locative - NP
(38) There stands in the corner of the room a file cabinet.
W rogu pokoju stoi szafa na dokumenty.
f/iere-sentences in English and Polish 165
(39) There hangs on the wall a picture of the President.

Na scianie wisi portret Prezydenta.
(40) There lives next door to me a blind woman.
Obok mnie mieszka niewidoma kobieta.
The order of the verb and the subject in English presentational sentences
and in their Polish equivalents can be easily explained by the principle of put-
ting lexical items representing a higher degree ofC.D. after those with a lower
degree of C.D. In these examples verbs of appearance on the scene have a
lower degree of C. D. than the subject NPs denoting those entities that appear
on the scene. Similarly, stative verbs describing the manner of spatial existence
are less communicatively relevant than the NPs following them.
The only important difference between English presentational sentences
and their Polish counterparts can be noticed if we compare these sentences
with other types of sentences in both languages. Thus English presentational
sentences are stylistically marked because they violate the basic word order
subject - verb and the principle that locative adverbials are usually in clause-
peripheral positions, for instance in thematic position as in these sentences
without there:
(36) a. Out of the bushes ran out a grizzly bear.
(37) a. On the wall hangs a picture of the President.
In English the order subject - verb can be changed to follow the degree of
C.D. only for a restricted class of verbs. In Polish, on the other hand, the
communicative order is obligatory and unmarked. The verb can precede the
subject not only in the case of verbs of being and verbs of appearance on the
scene but also in the case of other verbs:
(41) tobie opo wiada I moj dziadek.

*'About you told me my grandpa'.
(42) Ten obraz widzialo wiele osob.

""This picture have seen many people'.
(43) W ogrodzie tanczyly dzieci.
*'In the garden danced children'.
(44) ksiqikf dal jej Janek.
""This book gave her John'.
166 Maria Grzegorek
3 Conclusion
3.1 Existential sentences in English and Polish
a. the order of subject and verb
In both languages the obligatory (contextually unmotivated) change of the
basic order subject - verb to verb - subject distinguishes existential sentences
from other types of sentences in these languages.
b. theme
Bare existential sentences are in both languages themeless. The initial bye in
Polish cannot be treated as theme because being a verb it cannot be used re-
ferentially, and having such general meaning it cannot be contextually bound.
Locative-existential sentences have different themes in English and Polish: in
English they are basically themeless, and only optionally the locative phrase
can be shifted to the thematic position. In Polish, on the other hand, locative
phrases are obligatorily thematic in these sentences. The change of the position
of the locative adverbial in Polish results in a stylistically marked utterance.
3.2 Non-existential ere-constructions in English and their Polish counter-

parts
The Polish counterparts of English non-existential Mere-sentences do not con-
stitute any separate class. Like in other Polish sentences, the word order is
motivated by communicative functions of particular expressions. All these
sentences have themes in Polish. English non-existential there-constructions
constitute an exceptional group among other English sentence types because
the basic word order subject - verb has been violated in them for communica-
tive reasons.
Notes
1. The definition of theme and various types of themes are discussed in some detail in
my paper "Thematization in English" to appear in Studia Anglica Posnaniensia 17.
2. This observation is due to Rando and Napoli (1978). They replace the notion of
cardinality words introduced by Milsark (1977) with the notion of anaphoricity. Ac-
cording to Milsark only NPs with cardinality determiners (as opposed to quantifica-
tion determiners) can occur in existential sentences. Milsark's notion does not ac-
count for definite NPs and generics in fAere-constructions, whereas Rando and
Napoli's notion of anaphoricity does. By anaphoric NP they mean such that the
speaker can safely assume that the hearer can associate not only a unique referent
with this NP but also the proper unique referent.
3. It is important to remember that only the first occurrence of be in the clause wheth-
er 'existential' or passive auxiliary or progressive auxiliary) can trigger fAere-insertion
if the NP is indefinite. The only exceptions are the occurrences of be in modal ex-
pressions such as be about to and be going to (cf. Milsark 1974).
4. Which verbs other than be can occure in existential sentences cannot be determined
on a purely semantic basis. The matter remains a mystery. For instance, there is no
explanation why begin can occur in fAere-constructions and its synonym start is less
acceptable:
{ ??*started }
5. Prepositional phrases of location which are preposable are excluded here and are
treated separately.
6. In fact, very few adjectives can appear in codicil position. According to Milsark
(1974, 1977) only adjectives denoting temporary states can occur in this position.
Gee (1978) stresses the relevance of the opposition between adjectives with and
without a complement because of the following contrasts in acceptability:
T h e r e are some people arrogant.
There are some people arrogant to the point of obnoxiousness.
7. Van Oosten (1978) points out that this scale corresponds to a scale in their syntactic
behaviour, such as embeddability. True existential fAere-constructions are freely em-
beddable whereas presentational sentences are not. This fact is explained in terms of
word order markedness - in true existential sentences the word order is unmarked,
whereas presentationals have a marked word order. Clauses with marked word order,
i.e. clauses in which syntactic changes have been made in order to meet the com-
municative requirements are always restricted in embeddability.
8. In fact both sentences:
a. A book is on the table.
b. The book is on the table.
are locative (cf. DuSkova 1971). The difference is in the way the identity of the refer-
ent of the subject NP is established. In (a) it can be established only with reference
to the predicate, whereas in (b) it has been previously established.
9. Cf. Breivik (1975). Kimball (1973), from whom the example 37 is taken, explains
the difference in acceptability between 37 (a) and 37 (b) in terms of inalienable pos-
session: existential sentences with the inalienably possessed NPs do not have non-
existential counterparts.
10. The possibility of replacing a locative fAere-construction with subject - have - ob-
ject structure differs for different types of locative phrases. Also, there are differences
in the behaviour of pronominal copy which in some cases has to be retained, in others
it cannot be retained, and sometimes it is optional (cf. Erdmann 1978), e.g.:
a. There are footsteps on the stairs.
b. The stairs have footsteps (on them).
a. There are many good theatres in London.
b. London has many good theatres.
c. *London has many good theatres in it.
a. There is snow on the ground.
b. *The ground has snow.
c. The ground has snow on it.
168 Maria Grzegorek
11. Breivik (1975) argues that analogy can account for occurrences of there after the
clause initial locative.
12. The term "lexical function" is used here in the sense in which it is used by such
authors as I. A. Mel'iuk and J. D. Apresjan.
13. Polish makes a distinction between assertion of existence and assertion of nonexist-
ence not only in terms of negation but also by means of changing in present tense the
copula bye to the copula miec 'have'.
14. The term is taken from Firbas 1975.
15. Sentences with bye are also acceptable but less natural.
16. Babby uses the notion of lexical function to explain common features of all verbs
which can replace the verb bye in existential sentences, and he opposes this notion
to the concept of a natural semantic class which is of little use here.
17. The second (existential) reading is marginal, and sometimes even impossible. For in-
stance,' Milsark (1974) gives the following example in which existential interpretation
leads to absurd conclusions:
There are peasants constantly being murdered.
The existential interpretation, according to which constantly being murdered is a de-
scription, claims that the peasants mentioned here can die repeatedly.
References
Aissen, J.
1975 "Presentational f/iere-insertion: a cyclic root transformation", CLS 1 1 : 1 - 1 4 .
Allan, K.
1971 "A note on the source of there in existential sentences", FL 7:1 - 1 8 .
Apresjan, J. D.
1980 Semantyka leksykalna (Wroclaw: Zaklad Narodowy im. Ossolinskich).
Babby, L.
1978 "Lexical functions and syntactic constructions: Russian existential sentences",
Papers from the Parasession on Lexicon, edited by D. Farkas, W. M. Jacob-
sen, and K. Todrys (Chicago: CLS), 2 6 - 3 4 .
Breivik, L. E.
1975 "The use and non-use of 'existential there'' in present-day English", Forum
Linguisticum. Contributions to applied linguistics 2 : 5 8 - 1 0 3 .
1978 "Existential sentences revisited", Papers from the 4th Scandinavian Con-
ference of Linguistics, edited by K. Gregersen (Austin: University of Texas),
235-240.
Cole, P. - J. Sadock, eds.
1977 Syntax and semantics 8: Grammatical relations (New York: Academic Press).
DuSkova, L.
1977 "A note on 'there is' in present-day Enghsh", Philologica Pragensia 5 9 : 9 7 - 1 0 5 .
Erdmann, P.
1978 "rfcere-constructions in English and German", IRAL 1 6 : 1 8 7 - 2 1 1 .
Firbas, J.
1975 "On the thematic and the non-thematic section of the sentence", Style and
text: Studies presented to Nils Erik Enkvist, edited by H. Ringbom (Stock-
holm: Sprkforlaget Skriptor AB), 3 1 7 - 3 3 4 .
Gee, James Paul
1978 "Adjective preposing and there-insertion: a point about syntactic rules and
semantic processes", Studies in language 2 : 1 0 3 - 1 1 1 .
i/zere-sentences in English and Polish 169
Green, Georgia M.
1977 "Do inversions in English change grammatical relations?", Studies in the lin-
guistic sciences 7 : 1 5 7 - 1 8 1 .
Halliday, . A. K.
1967 "Notes on transitivity and theme in English", JL 3 : 3 7 - 8 1 .
Jenkins, Lyle
1975 The English existential (= Linguistische Arbeiten 12) (Tbingen: Niemeyer).
Kimball, John P.
1973 "The grammar of existence", PCLS 9 : 2 6 2 - 2 7 0 .
Kirkwood, H. W.
1977 "Discontinuous noun phrases in existential sentences in English and German",
JL 1 3 : 5 3 - 6 6 .
Milsark, G. L.
1974 Existential sentences in English. Repr. by the Indiana University Linguistics
Club in 1976.
1977 "Toward an explanation of certain peculiarities of the existential construc-
tion in English", Linguistic analysis 3:1 - 2 9 .
Oosten, J. van
1978 "Expletive f/iere-sentences", unpublished MS.
Pullum, G.
1977 "Word order universale and grammatical relations", Syntax and semantics 8:
Grammatical relations, edited by P. Cole and J. Sadock (New York: Academic
Press), 2 4 9 - 2 7 7 .
Rando, Emily - Donna Jo Napoli
1978 "Definiteness in r/iere-sentences", Language 5 4 : 3 0 0 - 3 1 3 .
Ross, J.
1974 "There, there, (there, (there, (there . ..)))", CLS 1 0 : 5 6 9 - 5 8 7 .
Searle, John R.
1971 "What is a speech act?", The philosophy of language, edited by J. R. Searle
(London: Oxford University Press), 3 9 - 5 3 .
EDMUND GUSSMANN
Abstract phonology and contrastive analysis
It is a truism to say that contrastive linguistics is inextricably connected with

general (or theoretical) linguistics. The relation is straightforward and uni-
directional in that general linguistics provides a model in terms of which any
specialized or purpose-oriented description including a contrastive one is
couched. Thus the selection of a specific linguistic model mechanically, so to
speak, determines the boundaries within which the contrastivist may operate
as it provides him with descriptions of individual languages which utilize a
particular theoretical apparatus and not some other one. Were our phonolog-
ical descriptions to be formulated in terms of,'say, autonomous phonemes
and allophonic statements, the contrastive interpretation would be very
different from one developed on the basis of descriptions involving systematic
phonemes and phonological rules of the generative type. The task of over-
riding importance that the contrastivist faces at the outset then is the choice
of descriptions with their underlying linguistic model. We may be more
realistic and go one step further since, in actual practice, the contrastivist,
drawing on a specific model, has to construct the descriptions himself. The
rapidly changing views on the nature of language and its structure do not en-
courage scholars to develop comprehensive descriptions of individual languages
but rather to study specific problems in depth attempting to verify and/or
modify general theoretical models. The contrastivist's proper work, on the
other hand, starts where the descriptivist leaves off, even if, often enough,
the contrastivist has to do the descriptivist's job before he can start with his
own.
The fact that contrastive linguistics builds upon the results of decriptive
(theoretical) studies forces us to recognize the former as an independent
branch of language investigation. This has two important consequences which
have not always been explicitly noted and unnecessary confusion has occa-
sionally arisen as a result. For one thing, contrastive linguistics needs to de-
velop a theoretical framework of its own which would be geared to its ob-
jective and possibilities in addition to and independently of the theoretical
172 Edmund Gussmann
apparatus provided, ultimately, by the linguistic theory. This is by no means

an obvious or straightforward task and although certain clear results must be
noted here, 1 a lot remains to be worked out, particularly in phonology. Be-
low we shall discuss some of the major obstacles besetting the attempts to
develop theoretical contrastive phonology, i.e. basically a set of notions which
could be used for the purpose of contrasting phonological systems.
The other aspect of the independence of contrastive linguistics relates to
the contributions it can make to general linguistics. In the sphere of phono-
logy the relevance of contrastive data has been claimed to refer to transfer
phenomena (Eliasson 1 9 8 2 ) . It stands to reason that errors made by learners
of a foreign language may be indicative of the structure of native phonology.
To take an obvious example, the fact that Polish learners of English or
Swedish tend to devoice word-final obstruents incorrectly in the target
language, e.g. bi[k] for &/[g]or c?a[k] for da[g] 'day' can be taken as providing
unambiguous evidence for the reality of a rule of terminal unvoicing in the
native phonology (or within a different theoretical framework for the
reality of specific phonotactic constraints). Taken a few steps further, evidence
of this sort has been brought to bear on the more subtle theoretical issues
such as the question of admissible phonological abstractness (Eliasson 1978:
219ff). In other words, transfer evidence has been claimed to offer one more
mode of so-called external (or substantive) evidence in phonology, which
strives to justify theoretical positions by pointing to the way speakers appear
to utilize their internalized grammatical (phonological) knowledge in actual
lingustic behaviour. Regrettably, things have turned out to be more complex
than this reasoning might indicate. One problem relates to the general insignif-
icance of the absence of evidence, i.e. the failure of certain errors to arise
does not necessarily mean that some postulated phonological regularity is a
figment of the linguist's imagination. More remarkable, however, is the ap-
pearance of errors not predictable on the basis of the native phonology. To
take again a simple case - there is in Polish a rule of regressive voice assimila-
tion affecting obstruents across word boundaries; in standard Polish this rule
does not apply, even in fast speech, if the second word begins with a vowel.
Thus the final obstruent of las 'wood' remains voiceless in the collocation las
iglasty 'coniferous wood'. This regularity might be used to account for the in-
correct voicing of the final obstruent in the German article das in das Wort as
[daz]; it comes as a surprise to note, however, that Polish learners of German
persist in voicing the final obstruent of das in 'das ist ein. . .' or that Polish
learners of English voice the [f] in e.g. 'if I knew...'. These errors arise here
apparently in defiance not only of the German or English data but also in
defiance of the native phonological regularity. Another case of such erratic
behaviour from Polish has been discussed by Gussmann (to appear); similarly
Abstract phonology and CA 173
Eliasson (1982) offers some intriguing data in the interpretation of Finnish

quantity by Swedish speakers and he concludes by saying: "what emerges
is that external data and external verification are no less complex than purely
internal facts and internal argumentation. The verifying power of transfer
data, like that of other external data, will depend on a number of circum-
stances and assumptions". To formulate Eliasson's conclusions in somewhat
stronger terms we would claim that there can be no direct link between trans-
fer evidence and phonological theory in point of fact, transfer evidence is
something that needs to be accounted for within a broader theory of applied
contrastive linguistics. Applied contrastive phonology, while utilizing the re-
sults of theoretical contrastive studies, will need additional apparatus of its
own, an apparatus which is not deducible from nor reducible to notions found
in theoretical contrastive phonology. 2
The division of contrastive studies into theoretical and applied is, hopeful-
ly, well-established by now (Fisiak 1973), but the occasional failure to grasp
it clearly has brought criticism of theoretical studies for their not providing
an explanatory account of linguistic behaviour. In what follows we shall be
concerned with theoretical contrastive studies which aim at establishing sim-
ilarities and differences between two (or more) phonological systems. This is
not to say that we decry the importance of applied studies - quite conversely,
we believe that such studies can be intelligently developed on the basis of ade-
quate theoretical descriptions with the addition of notions and apparatus
geared to the purpose of any such applied description (foreign language teach-
ing being an obvious though not the only imaginable objective).
Contrastive theoretical phonology sets itself the more modest aims of
specifying areas of convergence and divergence in the sound structure of any
two languages; it certainly does not claim to account for all available sorts of
phonological behaviour. Thus its objectives are different and, one might say,
more limited; clearly, however, the recognition of the limitations of a field
does not undermine its validity. It seems that a characterization of phono-
logical similarities and differences between languages is a legitimate goal to
pursue. Below we shall be concerned with the difficulties that arise in the pro-
cess of pursuing such a goal.
In phonological descriptions the question of admissible abstractness re-
mains a pivotal one. The discussion that started with the appearance (in 1968)
of Kiparsky's "How abstract is phonology?" has enormously sharpened the
awareness of the issue among phonologists and has made them look for new
sources of relevant data and evidence; in addition to language-internal argu-
ments for a specific abstract or concrete solution (the number of such attempts
being too vast to make possible even a sample of references), all sorts of ex-
ternal (substantive) evidence have been resorted to (see Zwicky 1975,Fischer-
11A Edmund Gussmann
Jorgensen 1975; for a critical evaluation see Gussmann 1979). Predictably

enough, none of the opposing views appears to have definitively won the day
and the discussion continues; various models of more concrete phonology
have been proposed (Dressier 1977, Goyvaerts 1978, Hooper 1976, Leben
1977) diverging sometimes quite radically from the standard generative para-
digm of The sound pattern of English (Chomsky - Halle 1968); the paradigm
itself, however, including the endorsement of the need for abstractness in
phonology has been upheld - with modifications and revisions by a number
of researchers (e.g.: Kiparsky 1973, Kenstowicz - Kisseberth 1977 and 1979,
Gussmann 1978a, Zonneveld 1978). For the purpose of this article we will
adopt "the prevailing opinion that the realm of phonology is not exhausted
by a kind of systematic phonetic description" (Eliasson 1982), i.e. we will
assume that phonological derivations entail interaction of rules which do
not have to be directly traceable to surface facts; we will also maintain that
phonological representations are abstract, without pursuing this topic or at-
tempting to provide an answer to the moot question of exactly "how abstract"
they must be.
Within this model of phonological organization all the morphemes of a
language consist of systematic phonemes whose combinations observe the
morpheme structure constraints of the language. The process of phonological
derivation consists in modifying the segmental structure of morphemes de-
pending upon the contexts in which they appear, i.e. depending upon the
phonological and grammatical structure of the neighbouring morphemes. The
result is, of course, the phonetic representation of words and also larger units
(phrases and sentences for stress and intonation). Comparison of the phono-
logical structure could, in principle, cover the inventory of underlying seg-
ments, the posited morpheme structure constraints, phonological rules and
possibly also the inventory of phonetic segments. Given relatively adequate
descriptive analyses of two languages, the comparison of underlying segments
and morpheme structure constraints is basically a very simple exercise of say-
ing which segments and what segment combinations are shared by the two
languages and which of them occur separately in each of them. The concept
of shared segments presupposes naturally that systematic phonemes are viewed
not as abstract, indivisible units, distinct from all other units recognized by
the language, but rather that they are approached as elements identifiable by
their substance (i.e. distinctive features). Were we to view the underlying seg-
ments as merely 'different from everything else' or contrastive, then the only
available basis for comparison would be the number of such elements in each
language. Within a substance-oriented model of phonology, such as generative
phonology, the situation is somewhat better since in addition to the number
of segments one can further specify in phonetic terms the segments them-
selves and assign them to appropriate classes. This is, to all intents and pur-
poses, a marginal improvement over a purely quantitative evaluation for two
major reasons. An inventory of phonetically specified segments, even if ac-
companied by a set of morpheme structure constraints, which we bypass here,
remains nothing more than a list; usually no claims are made as to the internal
structure of the inventory and the segments could be arranged in any way
with no consequences following from a particular arrangement (see the list
of segments in Chomsky Halle 1968:177; for a dissenting voice, see Lass
1976:84). More importantly, however, phonological rules modify the seg-
ments in various ways including the complete elimination of some of them as
well as the introduction of units which do not appear in underlying represen-
tations. In this way the static juxtaposition of phonological segments offers
little more than a glimpse into the phonology of the languages. If our com-
parison were to be restricted to inventories, the result would not only be in-
complete but could also be downright misleading or false. By isolating what
we call shared segments we would be making an implicit and illicit claim that
they are the same (identical) segments. In actual fact, however, a group of
segments may be the same in terms of the features which define them but
they need not be identical in terms of the phonological behaviour. To take a
concrete example both for Polish and for English a group of underlying
dental obstruents /s, z, t, d/ has to be recognized 3 where the segments can
be viewed as the same. In terms of the modifications they undergo in various
phonological contexts, however, they are very different, and thus the initial
comparison of the segments in isolation has to be supplemented by a survey
of the rules that affect them.
Any comparison of phonological rules offers serious difficulties. These

were discussed several years ago (Fisiak 1975 and 1976, Gussmann 1975) but
little progress seems to have been made since in overcoming them. To see the
nature of the difficulties let us consider some of the phonological rules affect-
ing the dental obstruents in English and Polish. Since both languages evince
rules palatalizing dental obstruents, we shall restrict our attention to this pro-
cess. The rules have been discussed and justified elsewhere (Chomsky - Halle
1968 for English, Laskowski 1975, Gussmann 1978a,b for Polish), and al-
though the specific formulations can be disputed, we will adopt them here
since alternative interpretations would not seriously affect the points we will
be trying to make.
In their discussion of the English consonantal processes Chomsky Halle
(1968:230) postulate a rule converting underlying or intermediate /s, z, t, d/
into the phonetic [s, , <S, j] if directly followed by the glide /j/ and an un-
stressed vowel. The rule is formulated as
176 Edmund Gussmann
(1) -sonor] -ant -back

L+cor J L+strid J -voc r-cons
-cons L -stress J
This rule accounts for such alternations of consonants as those illustrated

in (2).
( 2 ) [s - 5] expre[s] - expre[5]ion
face facial
[z - ] sei[z]e sei[Z]ure
revise revision
[t - c] ri[t]e ri[c]ual
part departure
[d J] procee[d] proce[j]ure
grade gradual
The rule also affects derived segments, i.e. those that result from the applica-
tion of other (earlier) rules, in particular velar softening, spirantization, voic-
ing and devoicing. These additional processes (specified in parentheses) com-
bine with palatalization to produce the following sets of surface alternations:
(3) [s - z] preci[s]e preci[z]ion (voicing)

diffuse diffusion (voicing)
[ t - s ] Egyp[t] Egyp[s]ian (spirantization)
part partial (spirantization)
[t - 2] averft] aver[z]ion (voicing and spirantization)
[d - z] divi[d]e divi[i]ion (spirantization)
evade evasion (spirantization)
[d - s] inten[d] inten[s]ion (devoicing and spirantization)
concede concession (devoicing and spirantization)
[k - ] logifk] logi[3]ian (velar softening)
music musician (velar softening)
The rule of palatalization is ordered relatively late among the rules of English
phonology (Chomsky - Halle 1 9 6 8 : 2 4 4 ) , which agrees both with the exist-
ence of a certain amount of variation in its application (e.g. gradual with or
without a palatal reflex) and with the extension of the rule's applicability
across word-boundaries, particularly in rapid speech (e.g.: the[i]e young,
ni\s] e young etc.). Rules which follow palatalization include specifically
deletion of the glide /j/, i.e. of the segment which is crucially involved in the
triggering off of the rule (Chomsky - Halle 1 9 6 8 : 2 3 1 , rule ( 1 2 2 ) ) , as well as
the reduction of unstressed lax vowels. Although details of the rule could, no
doubt, be improved on, the basic pattern is clear: dental obstruents are turned
into palato-alveolars before a following 'j+unstressed vowel'; the glide is sub-

sequently deleted while the vowel is reduced to schwa.
Thus the presence of the glide is crucial to the application of palatalization
and partially also of spirantization. Chomsky and Halle (1968:225227)
provide some discussion intended to demonstrate that the glide itself is not
original but rather derived from underlying /i/. In other words, the rule of
palatalization is fed by the glide forming process. The relation between pala-
talization and glide formation has yet another important aspect, viz. a number
of forms can be found which make the palatalization rule opaque when a
dental obstruent is followed phonetically by the glide without any traces of
palatalization (or spirantization). Consider the following examples.
(4) primordial, invidious, cordial, medium, piteous, Pontiac,
gymnasium, potassium.
Chomsky and Halle (1968:225) claim that these forms for a variety of rea-
sons are prevented from undergoing the glide formation rule and then ob-
viously do not meet the structural description of the rules requiring the glide
(spirantization, palatalization). It should be pointed out, however, that at
least in the case of British English all the forms in (4) may contain a coronal
obstruent followed by the glide [j] phonetically so that cordial, pronounced
[ko:djal] is (or may be) bi-syllabic rather than necessarily tri-syllabic. Thus
even if, following Chomsky and Halle, we exempt the forms in (4) from the
glide formation rule, we still need another rule ordered after palatalization
to turn the underlying /i/, which remains /i/ at the stage when palatalization
operates, into the glide [j] in accordance with what we find on the surface
(at least in some dialects). The counterfeeding relation of this late glide for-
mation rule and the rule of palatalization accounts for the opacity of the latter.
Turning now to the Polish palatalization of dental obstruents we find a
general rule formulated in the following way (Gussmann 1978a:22):
(5) [-syllab] [-back]/ [-back]
i.e. all consonants are palatalized if followed by a non-back segment (a front
vowel or the palatal glide). As stated, the rule also covers non-dentals and
sonorants, being thus more general than its English counterpart, but we will
restrict ourselves only to the four obstruents /s, z, t, d/. The rule also obvious-
ly differs from the English palatalization by being sensitive to the [back]
parameter of the following non-consonant only in English the palatal glide
has to be followed by an unstressed vowel. Further and more significant dif-
ferences include subsequent rules which apply to the output of the palatal-
ization formulated in the most general and hence in a theoretically most satis-
fying fashion as (5).
178 Edmund Gussmann
Let us start with a situation which is reminiscent of that found in English,

i.e. when the [ - b a c k ] palatalizing segment is in fact /j/. In such a case a later
rule, termed j-anterior palatalization in Gussmann 1978b, turns the obstruents
into affricates or palatal spirants. This gives rise to alternations such as those
illustrated below:
(6) [s - ] ] 'scythe' - ko[s] scythe'
przymus 'enforcement' - muszg must'
[ - i\ ma[z~fac 'smear' - ma[z smear'
woz 'cart' - woz? cart'
[ t - t s ] lo[t] 'flight' - le[ts\$ fly'
oplata 'payment, due' ptaca 'salary'
[d-dz]wi[(/]ac 'to be seen' - wi[dz~\% see'
zqdac 'demand' - zqdza lust'
The glide is subsequently deleted so that the rule is made opaque in the sense
that the segment responsible for the phonetic reflexes of the underlying ob-
struents cannot be read off from the surface. The rule deleting the glide is
quite general, and sequences such as *[jj], *[sj] etc. are totally impossible
within the native vocabulary. Also sequences of a plain dental obstruent plus
the glide are absent from native vocabulary. In other words, there are no
other rules applying after palatalization and glide deletion which would create
clusters appearing in the input to the palatalization processes. This can be
readily contrasted with the English situation (the primordial type) where such
a rule does exist and contributes to the opacity of the palatalization process.
Thus this fragment of Polish phonology can be neatly juxtaposed with Eng-
lish: dental obstruents when followed by /j/ undergo palatalization with sub-
sequent deletion of the glide. Differences can be noted in the actual phonetic
exponents of the underlying obstruents and in the fact that English has a glide
deriving rule which stands in counterfeeding relation to palatalization. This
area of substantial correspondence between the two phonologies stands in
sharp contrast to the other area which comprises Polish palatalization before
front vowels. Note that the general formulation of the Polish rule as (5) covers
both palatalization by the glide /j/ and by front vowels.
When palatalized dental obstruents are followed by a front vowel, late
spell-out rules turn them into spirants or affricates, which produces the fol-
lowing phonetic alternations:
(7) [s fi]ro[s]a 'dew' - ro[p]ie (loc. sg.)

bosy 'bare-footed' bosi (nom.pl.masc.pers.)
[z-f>] wa[z\a 'vase' - wa[t]ie (loc. sg.)
woz 'cart' - wozic (id.'vb.)
[ t - t p ] lo[t] 'flight' - le[tf]i 'he flies'

lato 'summer' lecie (loc. sg.)
[&&f.\ra[d]a 'advise' - ra[d?]ic 'advise'
bieda 'poverty' - biedzie (loc. sg.)
If this were the end of the story, our contrastive analysis would merely point
out the different scope of the rules themselves as well as note the existence in
Polish of the additional spell-out rules; jointly this could account for the dif-
ference both in the phonetic reflexes of the palatalization rules as in the pho-
netic distribution of these reflexes. Unfortunately, the situation is much more
complex since the Polish rule belongs to the relatively early ones and thus we
run into the question of rule interaction. For the present discussion we select
only three such rules which crucially bear on the output of the palatalization
process: deletion of vowels (where more than one rule is in fact necessary),
backing of front vowels, and depalatalization.4
Vowel deletion - for our purposes: front vowel deletion makes the
palatalization rule opaque and results in palatalized consonants appearing
phonetically in a non-palatalized environment, i.e. before a consonant or in
word-final position (with voiced obstruents additionally undergoing terminal
unvoicing), e.g.:
(8) [s - p] gio[s] Voice' - glo[fn)y loud'
kosa 'scythe' - kos (id.vb.imperat.)
[z mro[z]u 'frost' (gen.sg.)- mro[fn]y 'frosty'
wiozq carry' wie[?\ wie[f~\
(id. imperat.)
[t-tfp] sia[t]ka 'net' - sie[t(\ (id.)
plotg weave' - plec (id. imperat.)
[A-d$]zgo[d\a 'agreement' - zgo[d}] -+zgo[tp]
(imperat.)
niedzwiaêk'Mttie bear' - niedzwie[d$ ->
niedzwie[tp] 'bear'
The rule backing front vowels is responsible for the fact that in a number of
words palatalized consonants appear before non-front vowels, which further
opaquates the palatalization rule. In the examples below we have the [e o]
and [e - a] alternations after a palatalized consonant where only the phonetic
shape of the first member of each pair conforms to the palatalization rule:
(9) [e o]pa[tpe]rz 'prayer' - pa[tpo]rek (id. dim.)
[fe]lnik 'herbary' - [fo]lo "herb'
[?e]lski'rural' - [o)lo Village'
gar[dfe]l 'throat' - gar[d?o]tko (id. dim.)
180 Edmund Gussmann
[e - a] [fee]/e'body'(loc.sg.) - [ta)to (nom. sg.)

ty[ee]c'grow bald' - iy[ea}t 'he grew bald'
[e]nnik 'hay mattress' - []'hay'
[d?e]to 'work' - [dfa]tac 'act'
In other word, forms exhibiting effects of the rule backing front vowels further
diminish the similarity between the Polish and English processes of dental ob-
struent palatalization. Both the case of front vowel deletion and front vowel
backing might be jointly contrasted with the Polish palatal glide deletion dis-
cussed directly above or with the English palatal glide deletion and lax vowel
reduction, since they all result in the palatalized segment appearing in a pho-
netical context which contradicts that of the palatalization rule. Since, how-
ever, the respective rules perform very different operations and are differently
integrated with the rest of the phonology, a contrastive statement such as this
can have at best impressionistic significance.
Finally the Polish rule of preconsonantal depalatalization affects dental
non-continuants which underwent the palatalization process. This creates a
situation which is paradigmatically important for Polish when, in the same
context, spirants emerge phonetically palatalized while plosives as non-pala-
talized. Here it looks as if the plosives were never subjected to palatalization
in the first place. However, we can speak about opacity both with spirants
and plosives since the palatalized spirants are followed by a consonant due to
the deletion of (front) vowels; with the plosives the rule of depalatalization
seems to be doubly opaquated: once by the vowel deletion rule which re-
moves the palatalizing segment and secondly by the depalatalization rule
which removes the palatal quality from the consonants. Consider the ex-
amples where spirants emerge as palatalized before the adjective-forming suf-
fix -ny [ni] (from /in+i) and before the noun-forming suffix nik [jiik] (from
/Tnik+f/), while plosives as non-palatalized:
(10) gto[s) Voice' gto[<?]ny loud'

gto[$]nik loudspeaker'
las 'wood' lesny 'of the wood'
lesnik 'forestmaster'
przewo[z]u 'transport'(gen.sg.) przewo[?yiy
'transportable'
przewo[f]nik 'carrier'
kotchoz 'collective farm' kotchozny (id. adj.)
kotchoznik
'collective farmer'
lo[t] 'flight' lo[t]ny 'mobile'
lo[t]nik 'pilot'
oplata 'payment' - piatny 'payable'

- platnik *payer'
wo[d]a 'water' - wo[d]ny 'hydrous'
- wo[d]nik 'aquarius'
chtod 'cold' - chtodny 'cool'
- ch lodnik 'cold fruit soup'
This sort of opacity, where the effects of a rule are undone in addition to and
independently of the context of the original rule being eliminated, does not
seem to be embraced by Kiparsky's (1973:79) celebrated definition of rule
opacity. s The specific Polish situation appears to correspond to nothing in
the English palatalization phenomena.
To complete the discussion of opacity involving the Polish rule of palatali-
zation we note the existence of a couple of rules operating after palatalization
which derive front vowels from underlying back ones. We shall not formulate
the rules here but merely illustrate the situation; in the following words we
have phonetically front vowels preceded by non-palatalized dental obstruents:
(11) ko[te\k 'cat' (dim.), [te]raz 'now', [ten]dy 'this way', [fejsc
'father-in-law';
[de^ka 'board', [de]szcz 'rain', [dem]by 'oak trees', ru[de]
'red-haired' (nom. pi.);
[se]n 'dream', [se]rnik 'cheese cake', po[sem]pny 'sombre';
bo[se] 'bare-footed' (nom. pi.);
w0[ze]k 'pram', [ze]rwac 'tear o f f , [zem]by 'teeth',
[ze\szyt 'copy-book', [ze]gar 'clock'.
This sort of opacity is again reminiscent of the situation with glide formation
in English: in both cases rules applying after palatalization derive a segment
which triggers off the rule. The major difference is that in Polish a few rules
deriving the surface [e] can be justified, all of which counterfeed palatalization.
Considerations of Polish and English palatalization phenomena shed some
light on the fundamental problems of contrastive phonology. Segments in
isolation can be readily compared, and we can also juxtapose individual rules
noting similarities and differences in their structural descriptions as well as in
the structural changes they introduce. However, there are very few rules which
operate in a phonological vacuum, and a contrastive analysis that purports to
study its subject matter in any depth must face the question of rule inter-
action. Here the appraisal is bound to be largely impressionistic: we have no
way of formally evaluating the relative importance of a rule, the degree of its
interaction with other rules, or the ensuing degree of phonetic opacity. These
questions, however, must be asked and some answer to them must be given, for
otherwise our analysis will be yet another comparison of apples and cucumbers.
182 Edmund Gussmann
In our discussion of shared rule comparability we have been concerned so

far with processes involving substantial rule interaction, i.e. with cases where
the postulated underlying representations of words were sufficiently abstract
to require several interacting rules for the corresponding phonetic forms of
these words to be derived. It was precisely the existence of rule interaction
which led to the claim that the contrasting of phonological rules is bound to
be partly impressionistic and hence inherently vague since it is not rules them-
selves that must be compared but rather their integration with the rest of the
phonology. Assuming that rules can be regarded as units, it is relations among
units - within one language and across languages - that create the greatest
challenge to contrastive phonology. Quite simply, while we have means (some-
times fairly diverse) at our disposal for comparting units be they segments,
autosegments or separate rules no mechanism seems to be directly available
for comparing relations. Thus we can get reliable and unequivocal informa-
tion about units across languages but this information is to a great extent
either misleading or meaningless. This we have tried to show by considering
how the largely similar underlying dental obstruents differ in terms of the
palatalization rules they undergo in Polish and English, and further how the
partly similar palatalization rules differ in view of their functioning in and
integration with the rest of the phonologies of the individual languages.
It might be suggested that the problems are due to the model of phonology
we have selected and specifically that they would not arise within a more con-
crete or phonetics-oriented approach. As stated earlier, we believe that the
selection of a model for contrastive studies must be done independently of
any contrastive ends; we emphatically reject the idea that some model is to
be preferred solely because of the ease with which it can be adopted for con-
trastive purposes. For the sake of argument, however, let us consider the pos-
sibility of comparing surface phonologies taking as our material again the
dental obstruents /s, z, t, d/. The point we will be trying to establish is that an
analysis along such surface lines runs into exactly the same kind of difficulties
as the more abstract model and that contrasting surface inventories cannot be
taken as a significant aim for contrastive phonology.
After describing our obstruents in terms of the phonetic parameters speci-
fying differences in the place of articulation within the dento-alveolar region
and such features as voicing and aspiration (subsumed possibly under tense-
ness), we would want to look at the (surface) phonological behaviour of these
segments. To make the situation completely uncontroversial, we will consider
the late rules which are characteristic of rapid speech (the allegro style), e.g.
given the phrase next Wednesday there can be no argument that the adjective
can, depending upon the style and tempo of speech, be pronounced with or
without the final dental plosive, i.e. [nekst] or [neks]. In other words, there
can be no doubt that some process of deletion is involved here; it depends on

no abstractness assumptions since it clearly relates two occurring surface
forms. Processes of this type have been extensively studied by Rubach (1977)
and the ensuing remarks draw heavily on his observations and analysis of dental
obstruent deletion in English and Polish (Rubach 1977:90-99, 131-135).
In English dental stops delete, roughly, between two consonantal segments.
Consider the following examples where the deletable segment is italicized.
(12) postpone, lifis, hones/ly, kindly, hanc/bag, boWly, nexf sta-

tion, lefr wheel, finds.6
These uncontroversial, non-abstract deletions are linked with other rules, in
particular they must be ordered after the rules of obstruent degemination and
palatalization. As shown by Rubach (1977:9394), degemination has to pre-
cede dental deletion since otherwise lists would counterfactually end up as
*[lis]. Likewise helped you may emerge as [helpcu] with palatalization but
not as *[helpju] with deletion of the dental obstruent. Rubach stresses that
although these rules are optional, they still require extrinsic ordering. Thus
rule interaction the greatest stumbling block to any straightforward con-
trastive phonology holds not only for deeper regularities but equally strong-
ly for very concrete derivations.
Dental plosives are also optinally deleted in Polish although this happens,
in crude terms, between a coronal obstruent and a consonant, e.g.:
(13) przyjezdny 'visitor', gwiezdny 'of the star', chrzestny
'baptismal', postny 'of the fast', uczestnik 'participant',
astma 'asthma', napastliwy 'aggressive'.
In terms of rule interaction it must be noted that the rule deleting the dental
obstruents has to be followed by one assimilating palatal quality in certain
clusters. This explains why the spirant is non-palatalized in slow speech but
palatalized if the plosive is deleted as in the following examples:
(14) przyje[zdyx\i vs przyje[?n]i
chrze[stji]i chrze[fji]i
The different integration of the dental stop deletion with other rules of Polish
and English phonology clearly wreaks havoc with any notion of similarity (or
difference) among (surface) segments as being the satisfactory starting point
for contrastive analysis. While the segments are undeniably similar as units,
they are equally undeniably involved into different relations as segments of
Polish or English phonology.
Contrastive phonology, if it is to be meaningful, must concentrate primari-
ly on the networks of rule interactions; as we have seen, this is a task for a
184 Edmund Gussmann
phonology which follows an abstract model as well as for a completely sur-

facy one. What cannot be overemphasized in respect to any phonological
comparison is that segments (units) cannot be studied outside relations but,
fundamentally, that relations cannot be reduced to or replaced by units (seg-
ments). Assuming that we have at our disposal a system for comparing seg-
ments, what remains an urgent task for contrastive phonology is developing it
in such a way as to embrace relations. In pursuing this goal we need not worry
too much about whether phonology should be abstract or concrete.
Notes
1. The concepts of congruence and equivalence in contrastive syntax and semantics as

developed by Marton (1968), Krzeszowski (1971), 1974).
2. The same applies in equal measure to all other modes of so-called external evidence as
argued in Gussmann (1979).
3. In English, of course, also // has to be recognized, ( [6] might be viewed as derived
by rule, see Gussmann 1 9 7 8 b : 1 2 5 - 1 2 6 ) , and in this sense the groups are not identical
numerically. Secondly it is not obvious whether underlyingly the English obstruents
should be regarded as dental or as alveolar - they are predominantly alveolar pho-
netically.
4. The rules are not formalized here, see Laskowski (1975) and Gussmann (1978a).
5. It might be argued that there is no opacity involved here at all since un-palatalized seg-
ments [t, d] appear on the surface in a non-palatalizing environment (before a conso-
nant). This would force us to claim that suffices such as -ny, nik (and a few others)
appear phonologically with an initial palatalizing vowel after the spirants /s, / and
without it after the plosives /t, d/, clearly an unacceptable conclusion.
6. The same process, we believe, is responsible for what looks like a change of affricates
into spirants, i.e. [] into [5] and [j] into [f] in examples such as:
lunch, French, bench, branch
orange, change, bulge, challenge.
If we interpret the affricates [2,j] as a sequence 'fricative plus plosive' ([t+S, d+2])
then these simplifications could be regarded as following from the same regularity
whereby a plosive /t, d/ is deleted between a consonant (in the case at hand always a
nasal) and an obstruent (here [5] or []). This is developed in Gussmann (1979), but
see Rubach (1977:105-106) for a different interpretation.
References
Chomsky, . - M. Halle
1968 The sound pattern of English (New York: Harper and Row).
Dressler, W. U.
1977 Grundfragen der Morphonologie (Wien: sterreichische Akademie der Wis-
senschaft).
Eliasson, S.
1978 "Theoretical problems in Scandinavian contrastive phonology", The Nordic
languages and modern linguistics 3, edited by J. Weinstock (Austin, Texas:
The University of Texas Press), 217-243.
1982 "Transfer as evidence for phonological solutions", SAP 14. 185-196.
Fischer-J<rgensen, E.
1975 "Perspectives in phonology", Annual report of the Institute of Phonetics,
University of Copenhagen 9:215-236.
Fisiak, J.
1973 "The Polish-English contrastive project", Papers and studies in contrastive
linguistics 1:5-13.
1975 "The contrastive analysis of phonological systems", Kwartalnik Neofilo-
logiczny 23:119-124.
1976 "Generative phonology and contrastive studies", The Canadian journal of
linguistics 21:171-179.
Goyvaerts, D. L.
1978 Aspects of post-SPE phonology (Ghent: E. Story-Scientia).
Gussmann, .
1975 "How do phonological rules compare?", Papers and studies in contrastive
linguistics 3:113 -124.
1978a Explorations in abstract phonology (Lublin: Uniwersytet im. . Curie-
Skiodowskiej).
1978b Contrastive Polish-English consonantal phonology (Warszawa: Panstwowe
Wydawnictwo Naukowe).
1979 "Abstract phonology and psychological reality", Proceedings of the Ninth
International Congress of phonetic sciences, vol. II, edited by E. Fischer-
J^rgensen, J. Rischel and N. Thorsen (Ktfbenhavn: University of Copen-
hagen), 101-107.
to appear "Contrastive analysis, substantive evidence and the abstractness issue",
Theoretical issues in contrastive phonology, edited by S. Eliasson (Heidel-
berg: Julius Groos Verlag).
Hooper, J. B.
1976 >1 introduction to natural generative phonology (New York: Academic
Press).
Kenstowicz, M. - Ch. Kisseberth
1977 Topics in phonological theory (New York: Academic Press).
1979 Generative phonology (New York: Academic Press).
Kiparsky, P.
1968 How abstract is phonology? (Bloomington, Indiana: Indiana University Lin-
guistics Club).
1973 "Abstractness, opacity and global rules", Three dimensions of linguistic
theory, edited by O. Fujimura (Tokyo: TEC), 5 7 - 8 6 .
Krzeszowski, T.
1971 "Equivalence, congruence and deep structure", Papers in contrastive lin-
guistics, edited by G. Nickel (Cambridge: Cambridge University Press), 3 7 - 4 8 .
1974 Contrastive generative grammar. Theoretical foundations (Lodz: Uniwersytet
Lodzki).
Laskowski, R.
1975 Studia nad morfonologiq wspotczesnego fczyka polskiego [Studies in the
morphonology of modern Polish] (Wroclaw: Ossolineum).
Lass, R.
1976 English phonology and phonological theory (Cambridge: Cambridge Uni-
versity Press).
186 Edmund Gussmann
Leben, W.
1977 "On the interpretive function of phonological rules", Phonologica 1976,
edited by W. U. Dressier and . E. Pfeiffer (Innsbruck: Innsbrucker Beitrge
zur Sprachwissenschaft), 2 1 - 2 7 .
Marton, W.
1968 "Equivalence and congruence in transformational contrastive studies",Studio
Anglica Posnaniensia 1 : 5 3 - 6 2 .
Rubach, J.
1977 Changes of consonants in English and Polish. A generative account (Wroclaw:
Ossolineum).
Zonneveld, W.
1978 A formal theory of exceptions in generative phonology (Lisse: The Peter de
Ridder Press).
Zwicky, A.
1975 "The strategy of generative phonology", Phonologica 1972, edited by
W. U. Dressier and F. V. Mares (Mnchen: W. Fink Verlag), 151 - 1 6 8 .
RAYMOND HICKEY
Towards a contrastive syntax of Irish and English
0 Introduction
In contrasting the syntax of two languages which are typologically as far re-
moved from each other as are Irish and English one is presented with an abun-
dance of conflicting structures. The purpose of the study, of which this article
is a preliminary stage, is to classify these differing structures and also to in-
vestigate the question of interference for the non-native user of either language.
Not all structural differences lead to interference forms, however, and in
many of the cases illustrated below adaption to the varying structures occurs
without the hampering effect of transfer of native language structures.
In the case of this particular study the contrast to be made is non-direc-
tional, assessing what transfer forms can be registered in Irish on the part of
native speakers of English while also viewing the attested interference forms
from Irish in Hiberno-English which have lead to its deviation from Standard
English.
The question of the tertium comparationis arises when comparing struc-
tures of both languages which are suspected of being semantically equivalent. 1
Here the decision procedure used avails primarily of the intuition of the in-
vestigator. Thus with a pair of sentences such as the following
(1) hi s6 ar an dara fear a thdinig isteach,

was he on the second man who came in 2
(2) He was the second man who came in.
a (partially) bilingual speaker can assert their semantic equivalence, a point at
which he is unlikely to be contradicted. In accordance with the model-free
basis of investigation which I have chosen, I will not claim that the deep struc-
ture of both sentences is the same (though I'm sure this is the case) and then
maintain that its realization is different in both languages due to the varying
possibilities of lexico-morphemic combination of each language but simply
claim their equivalence on intuitive grounds. 3 As the differences between
188 Raymond Hickey
Irish and English lie first and foremost in the realm of surface structure it
seems to me to be most fruitful to consider the syntactic realizations with a
view to establishing differences between them. In this study I shall leave aside
the question of positive transfer 4 as it is unproblematical in practice.
Although I have chosen no specific grammatical model for my investigation
it should not be imagined that I will confine myself to a registration of dif-
ferent syntactic constructions in Irish and English. Such diverging syntagmas
are of limited interest when merely given as particular lexical realizations (al-
though this serves the purpose of illustration) but need to be abstracted so
that one can see them in terms of various combinations of syntactic units
which conflict with each other when considered from the standpoint of the
opposing language, but which each serve the purpose of representing a distinct
semantic complex common to both languages.
In contrasting the structures of English and Irish there are some which are
different to those of English and which have within them a unit or units
which form part of a relatively large paradigm. In this paper when dealing
with such structures I will opt on the grounds of concision for illustrating the
particular structure with a single example. A case in point here would be the
following sentence:
(3) Td fdm dul amdrach.
is under-me go tomorrow
intend going tomorrow.'
In each of the many possible cases of such constructions we have the defective
existential verb td followed by a form of one of the many prepositional pro-
nouns and verbal noun. The construction can be abstracted as follows:
V(ftQ + PREPRO + VN
This represents the structure minimally: it may be varied by the paradigmatic
elements of PREPRO and expanded by, for example, the addition of adver-
bials or other complements.
The sections found below each deal with a particular area of the syntax of
Irish and English where divergences are to be encountered. They represent a
selection of such areas only, contain single examples as a rule and do not deal
with detail of lexical realization.
1 Word order in Irish
Apart from the most obvious fact of Irish being a VSO language as opposed
to the SVO character of English5 there are many further instances where Irish
shows a word order which deviates from English. Variation of word order
within Irish itself is greater that in English due to the increased degree of
morphological marking. This allows fronting when emphasis is required. The
first element of Irish word order to be treated, however, is that which obtains
in neutral statement sentences.
1.1 Positional variations with verb complements

In positioning a direct object Irish usually requires that this be placed after the
subject and before further complements. The subject must be placed immedi-
ately after the verb so that it is recognized as fulfilling the function of subject:
(4) Chonaic mi Sedn.
saw I John
saw John.'
In the eventuality of further complements these are then added at the end:
(4) a. Chonaic mi Sedn ag an stdisidn .
saw I John at the station yesterday
saw John at the station yesterday
The direct object in (4 a) can be replaced by a personal pronoun in which case,

however, a displacement to a position after the adverbials occurs: 6
(4) b. Chonaic ag an stdisidn i.
saw I at the station yesterday him
saw him at the station yesterday.'
Such a displacement rule is unknown in English and is often not observed
among Irish learners which has led together with the influence of English syn-
tax on Irish speakers to a weakening of this rule so that when viewed in terms
of the binary distinction, correct or incorrect, it is possible to allow the English
complement sequence. Thus a necessity of Irish syntax is reduced to a col-
location with diminishing frequency in its favour due to undermining from
the syntax of English. 7
When the direct object is a proper or common noun then displacement to
the end of the sentence is not possible:
(4) c. *Chonaic mi ag an stdisidn inni Sedn.
saw John at the station yesterday.'
* Chonaic mi ag an stdisidn inni an gluaistedn.
saw the car at the station yesterday.'
190 Raymond Hickey
1.2 Rhematic fronting

Closely related to this displacement is fronting which can be undertaken in
Irish for the purpose of rhematic emphasis. Let us take a rhematically neutral
sentence such as
(5) Chonaic si an cailin a bhuaigh an duais inniu.

saw he the girl who won the prize today
'He saw the girl who won the prize today.'
Should it be required to emphasize, say, the direct object then this can be
done as follows:
(5) a. An cailin a bhuaigh an duais chonaic si inniu I.

the girl who won the prize saw he today her
The complex direct object can always be placed at the end of the sentence
thus giving a structure which in contrast to (5a) is allowed in English:
(5) b. Chonaic si inniu ian cailin a bhuaigh an duais.

(5) c. He saw her today, the girl who won the prize.
Common to both Irish and English is the restriction that only a complex
direct object, that is one consisting of a noun with a correlating relative clause,
can be removed from the sentence framework. Again in both languages the
space left by the extracted object must be filled by a personal pronoun. Thus
(5) d. *An cailin a bhuaigh an duais chonaic si inniu.
is unacceptable.
Irish goes a step further in allowing fronting of elements qualified by the
subject. A case in point is afforded by the extraction of a genitive attribute
from its normal position after its determiner:
(6) Td mac an mn& a labhair leat marbh.

is son of-the woman who talked with-you dead
'The son of the woman who talked with you is dead.'
(6) a. An bhean a labhair leat td a mac marbh.
the woman who talked with-you is her son dead
Fronting of this kind is not possible in English as the element in the genitive
has a correlating relative clause and so cannot precede its qualifier:
(6) b. * The woman who spoke to you's son is dead.8
When it is a case of the subject itself being fronted then the pattern given
above whereby the place of a functional element in the sentence framework
must be occupied can also be used:
(7) An bheirt bhan atd anseo anois bfonn siad anseo gach bliaiti.
the two woman who-are here now are they here every year
'The two women who are here now are here every year.'
Were it not for siad in (7) it would have the same word order as its English
semantic equivalent; its presence,however, shows that the essential VSO order
of Irish is maintained and furthermore that (7) is a case of rhematic fronting.
2 Unilateral syntactic ambiguity
The constructions to be considered under this rubric are those which are only
ambiguous from the point of view of one of the languages being considered,
so that here we have more than one equivalent each of which is semantically
different. In almost all cases of equivalence we can offer more than one con-
struction on either side which reflects accurately the meaning of the original
construction. This depends on our definition of equivalence. The cases here
are such that a particular distinction is not made which is necessary for a single
interpretation of the construction involved. Consider the following:
(8) An buachaill ar leis an rothar

the boy which-is/was with-him the bicycle
Due to the dependent form of the copula ar which does not, in this case,
formally distinguish between present and past (8) has the following inter-
pretations:
(8) a. The boy who owns the bicycle
(8) b. The boy who owned the bicycle
The same applies when a dependent form of the copula introduces a relative
clause which follows on a verb as in
(9) Ddirt sS gur ndireach an rud .
said he that-is/was shameful the thing it
which because there is no necessity to correlate chronologically the tenses of
the main and subordinate clauses allows of two interpretations:
(9) a. He said that it was a shameful matter.
(9) b. He said that it is a shameful matter.
192 Raymond Hickey
The ambiguity of both (8) and (9), however, is due to a constraint in the mor-
phonemics of Irish which does not allow the mutation of liquids. Should the
elements after ar and gur in (8) and (9) respectively be either vocalic or a
lenitable consonant then the ambiguity is resolved, as a distinction is then
made between past and present. Consider
(8) c. An buachaill arb uaidh [aeisb wan] an rothar
the boy which-is from-him the bicycle
'The boy who wants the bicycle'
(8) d. An buachaill arbh uaidh [aeaav wsei] an rothar
the boy which-was from-him the bicycle
'The boy who wanted the bicycle'
(9) C. Ddirt gur mr [GAJ mo: J] an trua 4.
said he that-is great the pity it
'He said that it is a great pity'
(9) d. Ddirt sdgur mhr [gaj w o n ] an trua 4.
said he that-was great the pity it
'He said that it was a great pity.'
While (9d) is no longer ambiguous with regard to the present indicative it
nonetheless remains so with regard to mood so that it can also be interpreted
as 'He said it would be a great pity.'
There are further instances of unilateral syntactical ambiguity which are
not so easily resolvable. Such an instance is provided by the lack of a pluper-
fect with the verb td in Irish which is noticeable when translating from English.
Thus the following Irish sentence has two interpretations:
(10) Nt raibh ann ach an gaoth.
(10) a. It was only the wind.
(10) b. It had only been the wind.
It is nonetheless possible in Irish to express the pluperfect with other verbs
which then have a finite form of the verb td plus past participle:
(10) c. Bhisd imithe nuair a thdinig a bhean.
was he gone when came his woman
'He had gone when his wife came.'
Quite distinct from tense uncertainty (that is in a contextfree situation) is the
ambiguity with regard to agent which also occurs in relative clauses in Irish.
Here the relative clause connector a does not indicate sentence function
among the units in both main and relative clauses. This is also the case in
English (if we choose not to make the distinction between who and whom)
but it is not critical as the word order in the relative clause indicates the func-
tion of the sentence elements. Compare
(11) Sin an fear a mhol an sagart.
that it the man who praised the priest
with its two possible equivalents
(11) a. This is the man who praised the priest.
(11) b. This is the man who the priest praised,9
The word order of the relative clause in Irish is fixed though there is the pos-
sibility of disambiguizing (11) by substituting ar for a thus obtaining the
meaning of ( l i b ) . Should (11) be negated then it is no longer possible to dis-
ambiguize it and so (1 lc) has the negated meanings of both (1 la) and (1 lb):
(11) c. Sin an fear ndr mhol an sagart.
that is the man who-not praised the priest
3 Unilateral alternative syntagmas
By alternative is meant here that two or more syntagmas can express the same
semantic content and that both do not stand in the relationship of paraphrase
to one another. Such syntagmas exist in both English and Irish for which the
following may serve as examples:
(12) An obair atd sibh a dhSanamh
the work which-is you do 1 0
(12) a. An obair atd dh6anamh agaibh
the work which-is do at-you
(12) b. The work which you (pi.) are doing
Strictly speaking we have here a case of bilateral alternative syntagmas as
(12b) also exists as a sentence without which. However, I choose here not to
regard those syntagmas as true alternatives of which one is merely a reduced
version of the other. Neither do I regard as such those sentences which con-
tain the same elements but in positional variations.
As alternative syntagmas in English we can offer the two translations of
the following Irish sentence:
(13) An obair atthar a dhianamh
the work which-0-are 11 do
(13) a. The work which is being done
(13) b. The work which they are doing
194 Raymond Hickey
4 Passive equivalence in Irish
The above sentences lead to another area of syntax where Irish and English
usage diverge. The English passive has no formal equivalent in Irish. I stress
formal as there are a variety of means by which one can convey the notion of
passivity in Irish. These involve constructions which are peculiar to Irish. Thus
if one begins with a sentence such as
(14) It is said that there is only a poor chance of it.
(where the passive can also be replaced by a finite verb with indefinite agent
They say that .. .) one can most appropriately render this in Irish by using
the autonomous form (the term used traditionally in Irish grammars). This is
a finite verb form which is not subject-marked. Thus the translation of (14) is:
(14) a. Deirtear nach bhfuil ach droch-sheans ann.

(J)-say not is (dependent form) but bad-chance in-it
In those cases where, in English, it is not required or desired to mention the
agent this form is the most convenient Irish equivalent:
(15) Rinneadh an obair.

0-did the work
(15) a. The work was done.
Should it be desirable to mention the agent then a subject-marked verb form

can be used:
(15) b. Rinne an obair.

did I the work
It is possible to express the agent while using the autonomous form. It then
appears in the form of a prepositional pronoun as in the sentence:
(16) Rugadh mac di.

0-bore son for-her
'She gave birth to a son.'
A further approximate equivalent to the passive (with or without specified

agent) is offered by the perfective past of Irish:
(17) The work has been done (by me).
(17) a. an obair dianta (agam).
is the work done (at-me)
Note the specifically Hiberno-English rendering of (17a) which characteristic-

ally has the past participle after the direct object:
(17) b. I have the work done.
Equally common in this connection is a copula form with a relative clause
which affords the degree of emphasis which is usually implied in English pas-
sive sentences with specified agent:
(18) Is12 mise rinne an obair.
is myself who did the work
(18) a. The work was done by me.
5 Concatenative verbs
There are very definite differences in the area of verbal concatenation be-
tween Irish and English which arise out of the dissimilarity of non-finite verb
forms in both languages. In English a concatenated verb can assume one of
three forms:
Infinitive with to : INF+
Infinitive without to : INF
Present participle : PART
as in the sentences
(19) He was told to leave.
(20) She must stay.
(21) They considered emigrating.
In Irish the non-finite forms in (19) (21) are all rendered by the verbal
noun, a verb form which is marked for neither person or number and is used,
among other situations, in that of verb concatenation:
(19) a. Ddradh leis imeacht.

0-said with-him go
(20) a. Caithfidh s( fanacht.

must she wait
(21) b. Smaoinigh siad dul thar sdile.13
thought they go over sea
Although Irish has only one non-finite verb form in such constructions it has
three possible syntagmas with this depending on whether there is (i) single
196 Raymond Hickey
concatenation (above); (ii) double concatenation (22) or (iii) concatenation

with a verb complement (23).
(22) Smaoinigh siad dul ag sndmh.
thought they go swimming
'They thought of going swimming.'
(23) Theip orm an bdd a dhiol.

failed on-me the boat its selling
failed to sell the boat.'
In those cases where a noun is used instead of the present participle in English
the bare verbal noun is used again, this being the equivalent of either a gerund
or as in (25) of a deverbative nominalization:
(24) We heard them talking.
(24) a. Chualamar iad ag caint.
heard-we them talking
(25) We heard the talk of the girls.
(25) a. Chualamar caint na gcailtni.
heard-we talk of the girls
6 Prepositional usage
Among the structural differences between Irish and English which lead to in-
terference that of prepositional usage occupies a prominent place. In most
cases we have prepositional verbs of English with literal translations of the
English prepositions into Irish:
(26) The time is up.
(26) a. * an t-am suas.
is the time up
(26) b. an t-am istigh.
is the time in
But in the case of (26 b) we have a figurative use of istigh which does not exist
in English, the equivalent here being up. The overgeneralization of the range
of up thus gives rise to (26a). The use of istigh is not confined to this particu-
lar phrase but can be used productively to express the notion of completion
or pastness, for example in
(26) c. a saol istigh.

is her life in
'Her time is up/her day is done.'
6.1 Prepositional pronouns

Even the most cursory account of the differences in syntax between Irish and
English could not afford to neglect the significance of prepositional pronouns,
synthetic forms of preposition plus pronoun which play such a vital role in
the system of complements of Irish verbs.
The most straightforward use of prepositional pronouns is where they cor-
respond to the prepositional object of an English verb:
(27) Thug si peann luaidhe dom.
gave he pencil to-me
'He gave the pencil to me.'
Frequently the prepositional pronoun occurs where the object in English is
non-prepositional:
(28) Nioireann si di.
not suits it to-her
'It doesn't suit her.'
Below I offer in tabular form a break-down of the remaining most character-
istic uses of prepositional pronouns:
(i) as the equivalent of the English possessive pronoun
(29) Cailleadh a mhac air.
<Z>-lost his son on-him
'He lost his son.'
(29) a. Briseann si an crof ionam.
breaks it the heart in-me
'It breaks my heart.'
(ii) as directional adverb
(30) Rachaimid ann le chiile.
we-will-go in-it together
'We will go there together.'
(iii) as intensifier
(31) D 'imigh siad leo.
left they with-them
'They left.'
(iv) elliptical use
(32) Cuirfidh mi ceist ar mhamai. Cuir leat, mar sin.
will-put I question on mother, put with-you so
'I'll ask mother. Well do that.'
198 Raymond Hickey
(v) in sentences with the copula

(33) Is beo sldn d<5. (equation)
is alive healthy to-him
'He is alive and well.'
(33) a. Cad is ainm duit? (appellation)
what is name to-you
'What is your name?'
(33) b. Curb as di? (origin)
where is-from to-her
'Where is she from?'
It is necessary to add that not all of the sixteen or so prepositional pronoun
types can be used in each of the above divisions. Many of them contain only
one possibility.
Finally special mention should be made of double occurrences of preposi-
tional pronouns which are common in Irish to express a semantic relation,
frequently that between subject and indirect object of a sentence in English.
It is necessary to grasp the applicational possibilities of such double preposi-
tional pronoun constructions as they are an essential trait of Irish syntax.
Such semantic relations as expressed by them may also occur in sentences
where a prepositional construction is already the equivalent of an English
lexicalised verb. Such is the case with the Irish equivalent to the English verb
owe:
(34) You owe it to him.

(34) a. Td si amuigh aige ort.
is it out at-him on-you
Here the English direct object appears as the subject of the Irish sentence and
both the subject and the indirect object in the English original are found in
prepositional pronoun form in Irish. The subject si is part of a paradigm so
that we can substitute it by another noun, e.g.fiacha, 'debts':
(34) b. Td fiacha amuigh agam air.
are debts out at-me on-him
'He is in my debt.'
7 Polyfunctional morphemes
Among the elements of Irish which present difficulty to the learner are those
which can assume more than one function. This is particularly the case when
English has monofunctional morphological units in its semantic equivalents so

that one is not prepared for the polyfunctionality present in Irish. Consider
the following:
(35) Idir fhir agus mhnd

'Both men and women'
(35) a. Idir medn olche agus breacadh an lae
'Between midnight and the break of day'
The distinction between the two meanings of idir is made morphonemically,

the meaning 'both' requiring lenition of the initial consonant of the following
word and the meaning 'between' not doing so.
But by far the most multifunctional element of Irish is the grapheme a
which can have at least eight different functions. It is important from a con-
trastive point of view that the learner recognize the individual functions and
apply them correctly. The following sentences exemplify them:
(i) vocative particle:

a dhuine uasail
'dear sir'
(ii) particle before numerals:
a haon, a dd, etc.
'one', 'two', etc.
(iii) object of verbal noun:
fion a 01
'to drink wine'
(iv) relative pronoun:
an Id a pdsadh
'the day he was married'
(v) possessive pronoun in the third person 14 :
L: a theach 'his house'
0 : a teach 'her house'
N: a dteach 'their house'
(vi) when concatenating verbs of purpose:
TSigh a chodladh
'Go to sleep'
D'iirigh $ a chaint.
'He rose to speak.'
(vii) a) inclusive general pronoun:
Sin a bhfuil agam.
"That is all that I have.'
200 Raymond Hickey
b) partitive general pronoun:

Sin a bhfuil don scial agam.
'That is all I know of the story.'
(viii) adverbialization of abstract nouns:
A ghSire a labhair si.
a + sharpness which she spoke
'How sharply she spoke.'
In addition to this there are a series of cases where a is found in petrified ex-
pressions, e.g. a chiile 'each other'. In none of the functions given above is a
a source of interference probably as it shows no surface correlation with any
morphological elements 15 of English semantic equivalents thus not giving rise
to negative transfer.
8 Brief characterization of interference types 16
It is necessary in a contrastive syntax as is intended here to distinguish various

interference types. This also applies of course to the basic distinction between
positive and negative transfer. Among negative forms, however, I should like
to introduce a further threefold distinction which I shall label as follows:
(i) pattern interference

(ii) compensatory interference
(iii) neglective interference
By (i) I understand those interference forms which represent the transferring
of a syntactic form from Li to L 2 , where this structure is not permissible. It
is important to stress here that it is a matter of structural patterns being trans-
ferred without the semantic equivalence being affected.
An example of (i) is the use of the definite article in Irish; here this is used
with generalised nominals (as in German, for example):
(36) Is I an Ghaeilge an teanga is deise.
is she the Irish the language most fine
'Irish is the finest language.'
The English equivalent shows no semantic shift but lacks the definite article.
The particular structural pattern of Irish shown here is often to be found in
Hiberno-English:
(36) a. He finds the Irish hard going.
Essentially different from (i) are both (ii) and (iii) which can be characterized
by their effect on the meaning of assumed semantic equivalents. By compen-
satory interference I mean the attempt on the part of a speaker to express a

semantic content for which he can find no ready structure in L 2 . 1 deliberate-
ly say 'ready structure' as all elements of meaning are theoretically expressible
in a second language by means of paraphrase. An example can be readily
furnished from Hiberno-English where we have a syntactical device for ex-
pressing concisely the perfective aspect of Irish, namely the placing of a past
participle after a direct object. Hiberno-English maintains the normal word
order of Standard English for cases where such an aspect is not intended:
(37) Have you read'Ulysses'?

(38) Have you 'Ulysses' read?
Paraphrases of (37) and (38) are:
(37) a. Did you ever read 'Ulysses'?
(38) a. Have you finished reading'Ulysses'?
(37) can then be termed a case of compensatory interference which is pre-

ferred to a paraphrastic expansion such as (37 a). Indeed an interference form
is likely to occur where no given structure (of the same length and of course
similar in syntax) in L 2 corresponds to one in L j , more likely in fact than the
occurrence of a larger structure in L 2 which is semantically equivalent to the
shorter one in L j .
Type (iii), neglective interference, is not always recognized for being what
it is, namely the neutralization in the non-native use of L 2 of a distinction
which exists in L 2 but is not found in L!. Let me give another example from
Hiberno-English, that of clause subordination. Standard English uses temporal
subjunctors in the introduction of temporal subordinate clauses. Such sub-
junctors are not usually used in Irish as the context expresses the notion of
temporal subjunction and/or causality if required. In each case the structure
is that of the additive clause whereby a subordinate clause is introduced by
agus 'and'. Consider the following sentences:
(39) Bhuail leis agus ag teacht antos an bothar.
hit me with-him and he coming up the road
(40) Chuaigh amach agus ag cur bdisti.
went I out and it putting rain
(39) implies a causal connection between the action of the main clause ('meet-
ing') and that of the subordinate clause. This is not so in (40) which merely
states the simultaneity of the actions in both clauses. In Irish the notion of
causality is implied by the meanings of the clauses and not by specifying this
202 R a y m o n d Hickey
with a causative connector. The Hiberno-English equivalents to (39) and (40)

are:
(39) a. I met him and he coming up the road.
(40) a. I went out and it raining.
The syntactic correlation of the non-specific clause connector and with the
non-finite verb forms which correspond to the verbal nouns of the Irish sen-
tences is quite accurate. Here the causal relation implied in (39) is left unspe-
cified in (39 a) thus neglecting the temporal-causal connectors while/when
which could be used (in (40a) in a purely temporal function).
I would like to make a further subdivision in (iii). The cases which I have
just discussed are those of specific neglective interference. General neglective
interference is a term which could be applied to the non-use of syntactical
possibilities present in L2 though not in L t . This gives us a designation for the
phenomenon whereby a limited number of the syntactic possibilities of L 2
are used. I would assert that neglective and compensatory interference go
hand in hand for it is precisely the ignorance of, or unfamiliarity with a large
range of syntagmas which leads to one falling back on those of and trans-
ferring them into L 2 where a certain proportion of them will of necessity not
be permissible.
Notes
1. For a definition of the notion of equivalence, see Krzeszowski 1 9 7 1 : 3 7 - 4 8 .

2. The second line renders what is intended as a word-for-word translation of the Irish
original. This is not entirely satisfactory as in many cases there are elements in Irish
which cannot be rendered sufficiently in English.
3. This is after all a procedure employed constantly in contrastive studies irrespective
of whether it is explicitly stated by the investigator or not.
4. See Carroll 1 9 6 8 : 1 1 4 f . for a discussion of transfer types.
5. The contention of McCawley 1 9 7 0 : 2 8 6 - 2 9 9 , that English has a VSO character refers
to the predicate-initial nature of clauses in an underlying constituent order and not
to surface structure.
6. Under the heading Objektsisolierung' Wigger gives a treatment of this fronting. See
Wigger 1 9 7 2 : 2 6 4 ff.
7. I should add here that I have n o t reached unanimity on this point: one of my in-
formants maintained that pre-adverbial position of the pronominal object was quite
acceptable with particularly long adverbial sequences. It was not possible, however,
to determine an exact correlation between sequence length and degree of acceptabil-
ity. On this subject, see Ahlqvist 1976:171 f.
8. This applies t o Standard English only.
9. The use of ' t h a t ' as relative pronoun is of course very c o m m o n perhaps because of
the merger of the subject and object f o r m s of the relative p r o n o u n : This is the man
that the priest praised.
10. It is difficult to gloss a and in English. For our purposes here we can say that a is
the form used before a verbal noun when a pronominal subject is present with its
connected finite verb, while d is that used when a prepositional pronoun subject is
present. This at least gives the distribution conditions for the two particles in these
syntagmas.
11. The zero symbol here refers to the fact that this relative, finite verb form is not sub-
ject-marked.
12. The copula is is formally a separate verb from the existential verb td. Its syntagmas
differ considerably as well.
13. This sentence is strictly speaking
Smaoinigh siad ar dul thar saile.
but in spoken Irish ar is dropped.
14. The abbreviations L, 0 and stand for (i) lenition (fricativization of the following
consonant), (ii) placing of h before vowels (but no further changes) and (iii) placing
of a homoorganic nasal or equivalent voiced consonant before following consonant.
In each case the added consonant is the only one pronounced, so that, for example,
a dteach, 'their house' [3<Jae:x].
15. That is those elements which perform similar functions. Thus we do not have con-
fusion of Irish a with English a.
16. This is by no means intended to be an original treatment of interference typology
but merely a set of remarks on the subject which are particularly relevant to the at-
tested interference forms, especially those present in Hiberno-English.
References
Ahlqvist, A.
1976 "On the position of pronouns in Irish", iigse 1 6 : 1 7 1 - 1 7 6 .
Carroll, J. B.
1968 "Contrastive linguistics and interference theory", Report of the 19th. annual
round table meeting on linguistics and language studies, edited by J. E. Alatis
(Washington, D.C.: Georgetown University Press), 1 1 3 - 1 2 2 .
Krzeszowski, T. P.
1971 "Equivalence, congruence and deep structure", Papers in contrastive lingu-
istics, edited by G. Nickel (Cambridge: Cambridge University Press), 3 7 - 4 8 .
McCawley, J. D.
1970 "English as a VSO language", Language 4 6 : 2 8 6 - 2 9 9 .
MacMathuna, L.
1980 "I dtreo anailis chodarsnach Gaeilge-Bearla", Teagasc na Gaeilge 1 : 7 2 - 9 1 .
Wigger, A.
1972 "Grammatik und Sprachverwendung in der Satzordnung des Neuirischen",
Festschrift fr Wilhelm Giese, edited by . Haarmann and . Studemund
(Hamburg: Buske), 251 - 2 8 2 .
GEORGE . HORN
Constraints on transformations: evidence from

contrastive analysis
A contrastive analysis is one in which the similarities and differences between

two (or more) langages at a particular level are explicated in the context of a
chosen theoretical framework. (One can also use data which exemplify the
similarities and differences between two (or among more than two) languages
to assist in choosing from among competing frameworks.)
In this paper, I will present a (partial) contrastive analysis of English and
Polish 'W/J'-questions in the context of the framework proposed in Horn 1979
(93108). My primary goal is to show that although Polish and English dis-
play substantially different surface behavior, they are nevertheless quite
similar at a more abstract level, and consequently, significant generalizations
can be formulated which apply to both languages (and language in general).
I will briefly discuss Chomsky's competing framework as formulated in
Chomsky 1977:86, for example, and show that the Polish data cannot be ac-
counted for in his analysis as it stands.
The paper is organized into three sections. In section 1, the proposed ana-
lysis is introduced and discussed; in section 2, it is extended to the relevant
Polish data; and in section 3, these data will be briefly discussed in the con-
text of Chomsky's current framework.
1 The proposal
Essentially, I will argue that 'wfc-movement' is unbounded, applies non-cycli-

cally,and is constrained by two island constraints and an interpretive strategy.
The island constraints are universal, while the interpretive strategy may only
apply in configurational languages like English. I use the term 'vv/z-movement'
to refer to the process which applies in sentences containing wVz-words to as-
sociate these words in COMP with some other position in the structure. In
Chomsky's analysis, this is done by moving the -word to COMP, leaving a
206 George . Horn
trace which is later associated with that wh-word. An equivalent analysis,

which has been suggested on several occasions in the literature, involves base-
generating w/i-words in COMP position in structures that contain empty
nodes, or traces, and binding these latter to the -word. This is the view that
I take, and consequently, the rule of wA-movement is replaced by a rule that
I shall refer to as [e]-anaphora. It is this rule, I will argue, that has the proper-
ties noted above.
Before proceeding to the proposed analysis, I will say a few words about
the framework of which it is a part. This framework is discussed at greater
length in Horn 1979:171-225 and 1983:297-328. It is organized roughly
as shown in the following diagram:
(1) Lexical component
Word-formation rules
I -> Lexical items (and entries)
Lexical transformations
Syntactic component
PS rules
Lexical insertion rules
Surface structures
[e]-anaphora
Pronominal anaphora
Functional component
Operations on functional (Complex) functional
structures structures
The rules of the syntactic component produce surface structures. The surface
structure level is the only level relevant for interpretation.
The lexical entries of verbs and verblike constituents contain, among other
information, functional structures, in which NP terms are identified according
to their grammatical relations and are assigned semantic relations. A typical
entry is that for the verb see, shown below:
(2) SEE
+V
SEE NP! NP2

a
The functional structure of see appears in the final line of (2).
The terms NPX and NP2 represent the syntactically defined grammatical
relations subject and direct object respectively. For English, the subject is the
circled NP in the configuration shown in (3 a), and the direct object is the
circled NP in the configuration shown in (3 b):
CA evidence for constraints on transformations 207
(3) a. S[(NW]
b. V p[ V(g) ]
Grammatical relations are defined on syntactic surface structures. In languages
like English, they are defined configurationally, in terms of tree structures, as
shown above. In highly inflected languages like Polish, they are defined mor-
phologically, in terms of casemarking. The particular definitions of gram-
matical relations can vary from language to language. The function of gram-
matical relations is to identify NPs in surface structures in order that their in-
dices can be inserted into the correct positions in functional structures. All
languages must have some mechanism for accomplishing this task, and there
need not be (and probably is not) a universal set of grammatical relations.
Returning to the functional structure of see, the terms and represent
semantic relations. Semantic relations are distinct from grammatical relations,
and are assigned by verbs and verblike constituents to the NP terms that appear
in their functional structures. There is not a one to one correlation between
semantic relations and grammatical relations, and the particular semantic rela-
tion that a particular syntactically or morphologically defined term bears de-
pends upon the individual verb. It therefore must be listed in the lexical entry
as shown in (2).
The organizational innovation in the model is the functional component.
For any given syntactic surface structure, a (complex) functional structure is
produced by combining the functional structures of the verbs and verblike
constituents that appear in that syntactic structure. Various operations on
functional structures apply to produce the (complex) functional structure in
its final form. Complex functional structures must conform to two well-
formedness conditions, only one of which is relevant to the present discussion.
It is formulated as follows:
(4) Well-formedness Condition I (WFC I): In a well-formed

functional structure, all NP terms must be indexed and as-
signed a semantic relation. Terms prefixed to functional
structures must bind terms in argument position.
The (complex) functional structures produced by the various operations pro-
vide the basis for the interpretation of the syntactic structures from which
they are derived. If more than one wellformed functional structure can be
derived from a given surface structure, the sentence that that surface struc-
ture underlies is ambiguous. The number of readings for any sentence corre-
sponds to the number of well-formed functional structures that can be derived
from its surface structure. If two surface structures have equivalent functional
structures, the sentences that they underlie are synonymous, and if no well-
208 George . Horn
formed functional structure can be derived from a given surface structure, the
sentence that that surface structure underlies is ungrammatical.
Two sample derivations, one from English and one from Polish, are shown
below. Consider the following examples:
(5) a. John saw Bill
b. Jan widziat Marka
John (nom) saw Mark (acc)
The surface structure of (5 a) is shown in (6):
(6) ^ S
NP
John saw Bill

In structure (6), John is identified as NPj and is indexed as, say, i. Bill is iden-
tified as NP 2 and is indexed as, say,/. The functional structure of example (5 a)
is produced from the functional structure of see, which is the only verblike
constituent in structure (6). To derive this functional structure, we begin with
(7), below:
(7) SEE NPj NP 2
The indices of John and Bill, the NP terms in the syntactic structure, are in-
serted into the proper positions in (7) to produce the following:
(8) SEE NPj[i] NP2IJ]
At this point, semantic relations are assigned to the indexed NP terms to pro-
duce (9):
(9) SEE n p j [i] np 2 ]

oc
The structure in (9) is the functional structure of (5 a), in its final form, and
as such, it provides the basis for the interpretation of this example.
Now consider the Polish example in (5 b). Its surface structure is shown
below:
(10) /S^
NP
1
V nom Vr NP
1
V acc
Jan widziat Marka

(I have not included a VP node in this structure. Whether or not Polish sur-
face structures contain VP nodes is not relevant to the present discussion.)
In (10), Jan is identified as NP n o m and can be indexed as i, and Marka is
identified as NP acc and can be indexed as/. The lexical entry for widziec 'see'
is shown in (11):
(11) WIDZIEC
+v
WIDZIEC NP n NP a
(Note that widziec and see have essentially the same type of functional struc-
ture. The difference between the two languages at this level lies in the way
that grammatical relations are defined in each, as mentioned above. As we
will see below, there are important consequences of this distinction.)
The functional structure of example (5 b) is derived from the functional
structure of widziec, beginning with the structure shown in (12):
(12) WIDZIEC N P n o m NP a c c
The indices of Jan and Marka are inserted into the proper positions in (12) to
produce (13), below:
(13) WIDZIEC NPnom[i]NPacc[j]
Semantic relations are assigned to the indexed NP terms to produce (14),

which is the functional structure of this example. 1
(14) WIDZIEC NPnom[i]NPacc[j]

With this background let us look at w/z-movement in English. Consider the

examples in (15), below:
(15) a. Who did John see

b. Who did John believe that Bill saw
c. About what did John read
The surface structures of these examples are shown in (16):

210 George . Horn
about WHAT
The [e]-anaphora rule applies to bind the empty node in each structure to the
constituent in COMP. This rule is formulated as shown in (17):
(17) [e]-anaphora
w j i ] w 2 [e]
1 2 => 1 w 2 [Xi]
The only condition that must be placed on this rule is that Wt and W2 must
be of the same category.
In the structures shown in (16), assume that the terms are indexed as fol-
lows: in a, WHO is indexed as i, John (NP,) is indexed as /'; in b, WHO is in-
dexed as /', John is indexed as / , and Bill (NPX of see) is indexed as k \ in c,
WHAT is indexed as i, about WHAT is indexed a s / , and John is indexed as k.
The [e]-anaphora rule applies to produce the structures shown below:

(18) a. /S-
WHOj Johrij believe that Billk saw [x,]

c.
about WHATj John k read
The next step in the interpretation of these examples is to derive a (complex)

functional structure for each that corresponds to its syntactic structure. For
such examples, the constituent in COMP is prefixed to the functional struc-
ture of the relevant verb, as shown below:
(19) ([. .WH. .]) [PREDNPj . . .]
Thus, the functional structures of the examples in (15) are derived from the
following:
(20) a. (NP[WH]) [SEENPi NP 2 ]
b. ( n p [WH]) [BELIEVE NP t [SEE NPJ NP 2 ]]
c. (p p [. .WH..]) [READ NPj PP]
212 George . Horn
Insertion of indices and assignment of semantic relations to indexed NP terms

produces the following:
(21) a. ( N P [wh 0 i ]) [SEE N P [ j ] N P [x,]]

b. ( NP [who,D [BELIEVE N P [j] [SEE N P [k] N p 2 [x,]]]
a
c. (pp [ABOUT NP [what,]]) [READ N P [k] PP [Xj]]
7 a
(In c, y represents the semantic relation that the preposition about assigns to
its object.)
These structures are well-formed since all of the NP terms are indexed and
assigned semantic relations. The NP constituents prefixed to structures a and b
are assigned semantic relations by virtue of the fact that they bind NP terms
in argument position which themselves bear semantic relations. Pairs of the
form NPj/xj are, in fact, functional level units. In structure c, the PP about
what, /, binds a term in argument position as required by the WFC.2
There is no need to constrain whconstituent insertion into COMP. To see
this, first consider the examples in (22):
(22) a. *Who did John see Bill
b. *Who did John believe that Bill saw Fred
c. *About what did John read about politics
Examples like (22) have syntactic structures like the following which contain
no empty node:
(23) a.
COMP
WHO
about WHAT John read about politics

Suppose, that in a, WHO is indexed as i, John is indexed as /', and Bill is in-
dexed as k\ that in structure b, WHO is indexed as i, John is indexed as j, Bill
is indexed as k, and Fred is indexed as m; and that in structure c, WHAT is
indexed as i, about what is indexed as /', John is indexed as k, politics is indexed
as m, and about politics is indexed as n. The [e]-anaphora rule cannot apply
to any of these structures.
The functional structures of (22) must be derived from the structures in
(20). The application of index insertion and assignment of semantic relations
to (20) produces the following:
(24) a. ( N P [ w h O i ] ) [SEENPl[j]NP2[k]]

b. ( NP [whOi]) [BELIEVE N P l [ j ] [SEE N P l [ k ] N p 2 [ m ] ] ]
oc a
c. (p p [ABOUT N p[whati]])[READ N P1l [k]
> 7
PP n [ABOUT N P [ m ] ] ]
"
None of these functional structures are well-formed. Structures a and b con-

tain an indexed NP term, w/ioj in each case, that bears no semantic relation
and does not bind a term in argument position that bears a semantic relation.
All three structures contain prefixed terms that do not bind a term in argu-
ment position. Consequently, all three structures violate Well-formedness
Condition I. No well-formed functional structures can be derived from the
214 George . Horn
syntactic structures shown in (23), and this accounts for the ungrammaticality
of the examples in (22). There is no need for a condition on lexical insertion
to guarantee, for example, that there must be an empty node in any structure
that contains a -form in COMP.
Examples like the following are also ungrammatical:
(25) a. *Which city did John go
b. *Which closet did John put the clothes
c. * About what did John see
These examples have surface structures like the following:
(26) a. ^S
which closet John put the clothes [e]

c. /S,
about what Jo
Each of these structures contains a constituent in COMP that does not occur
in the VP with the relevant verb. In these structures, suppose that the various
terms are indexed as follows: in a, which city is indexed as i and John is in-
dexed as /; in b, which closet is indexed as i, John is indexed as /, and the
clothes is indexed as k\ and in c, what is indexed as /', about what is indexed
as /, and John is indexed as k. The [e]-anaphora rule cannot apply in any of
these structures. In structure a, there is no empty node, and in structures b
and c, the empty nodes are not of the same syntactic category as the con-
stituents in COMP.
The functional structures of these examples must be derived from the fol-
lowing:
(27) a. (Wh-NP) [GO NP t ]
b. (Wh-NP) [PUT NP t NP2 PP]
c. (PP) [SEE NPX NP 2 ]
Index insertion and assignment of semantic relations apply to produce the fol-
lowing from (27):
(28) a. ( N P . [ w h , . . ] ) [ G O N P l [ j ] ]

b. ( N P .[wh-.. .]) [ P U T N P l [ j ] N p 2 [ k ] ]

c. (ppjfABOUT Np[whatj]]) [SEE N P j [k]]
7
None of the functional structures in (28) are well-formed. The ones in a and b
contain indexed NP terms that are not assigned a semantic relation, and the
one in c contains a prefixed PP term that does not bind a term in argument
position. Thus none of the three structures satisfies WFC I. 3 So we see that
there is no need of a condition on the insertion of /-words, or phrases con-
taining them, into COMP position to guarantee, for example, that these con-
stituents can only be inserted into the COMP of a structure that contains a
verb with the proper subcategorization features. In fact, no conditions need
be placed on either the rule that inserts whwords (and constituents contain-
ing them) into COMP or on the optionality of lexical insertion.
Furthermore, the [e]-anaphora rule itself is optional and unbounded, and
no command or precedence conditions need be placed on it. If this rule had
not applied to any of the syntactic structures in (16), for example, the func-
tional structures derived from these (on the basis of the structures in (20))
would be the following:
29. a. ( N p[wh 0 i ]) [SEE N P l ] ]
a
b. ( N p[wh 0 i ]) [BELIEVE N P l [ j ] [SEE N P l [ k ] ] ]
a a
c. ( p p [ABOUT Np[whati]]) [ R E A D N P l [ k ] ]
' y oc
216 George . Horn
These structures are not well-formed. All three contain a prefixed term that
does not bind a term in argument position, and a and b contain an indexed
NP term that is not assigned a semantic relation. These structures thus do not
form the basis of a possible interpretation of the examples in (15). However,
this presents no problem for the analysis since the well-formed functional
structures in (21), above, can be derived for these examples. The rule, then,
need not be obligatory.
It is easy to see that the [e]-anaphora rule need not be bounded. The only
information required for interpretation is the surface structure position of the
wA-word (in COMP) and the position of the single empty node which is bound
to it. Consider the following example:
(30) Who did John believe that Bill thought that Sally told Frank
that Fred hit
The following functional structure can be derived for this example:
(31) (Npfwho,]) [BELIEVE NPl[j] [THINK Npjk]

It is well-formed and provides the basis for the proper interpretation of ex-
ample (30).
Turning to the third point, consider the following structures, in which a
-word has been inserted into the COMP of an embedded S.
NP VP
V NP S
COMP s
John told [e] who Frank hit Bill
COMP S
S V NP
COMP S
who John kissing Mary annoy [e]
In structure a, the wh-word neither commands nor precedes the empty node,
and in structure b, it does not command the empty node. Now, let us suppose
that the [e]-anaphora rule, in the absence of any further conditions, can apply
to bind the empty node to the w/i-word in both cases, ignoring for the mo-
ment the fact that in (17), above, it is formulated as a unidirectional rule in
which NPj must precede [e], The functional structures derived from these
syntactic structures will then be the following:
(34) a. T E L L N p [ i ] N P [Xj] [( N P [who,]) [HIT N P l [ k ] N p 2 [m]]]
a a
i = John; k = Frank; m = Bill
b. ANNOY NpJCNptwhOi])
[KISS N P [j] N p 2 [k]]] Np 2 [ X i ]
i = John; k = Mary
Each of these functional structures contains a substructure that is not well-

formed. These substructures are shown below:
(35) a. . . . [( N p[wh 0 j ]) [HITNPl[k]Np2[m]]]...

b. . . . [ ( N p [ w h O j ] ) [ K I S S N P l [ j ] N p2[k]]] ...

Each structure contains a -word that is not assigned a semantic relation, and
to which no term in the structure that bears a semantic relation is bound. If
we require that the Well-formedness Condition apply to all proper substructures
of a complex functional structure, then the functional structures in (34) are
not well-formed and cannot form the basis for the interpretation of the syntac-
tic structures in (33). This is a quite reasonable way to view the application of
the Well-formedness Condition, and adopting it means that no conditions relat-
ing to command and/or precedence need be placed on the [e]-anaphora rule.
218 George . Horn
The [e]-anaphora rule is the only major operation that applies to syntactic
structures, and because it is unbounded and unfettered by command/preced-
ence conditions, there is no need for a syntactic cycle. (On this point, see
Horn 1979:125-144 and Horn 1983, sections 2.1 and 2.2.).
Because the [e]-anaphora rule in the proposed framework is the analog of
w/z-movement in earlier analyses, we can conclude that 'w/i-movement' has
the properties noted at the beginning of this section: it is an unbounded, non-
cyclic process.
The [e]-anaphora analysis extends, with no major modifications, to data
involving topicalization and AP deletion, to provide account of all unbounded
syntactic processes. This is discussed in Horn 1983, section 2.2, and here I will
only offer three examples as illustrations. These are shown in (36):
(36) a. John we saw yesterday

b. It was John that we saw yesterday
c. John is easy to please
The syntactic structures of these examples are the following:
NP VP
V NP ADV
John we saw [e] yesterday
It was John we saw [e] yesterday

The [e]-anaphora rule applies to all three structures to bind the empty node
to John. Suppose that in structure a, John is indexed as /', we is indexed as / ;
in structure b, John is indexed as i, we is indexed as / ; and in structure c, John
is indexed as /'. The complex functional structures derived from these syntac-
tic structures are shown in (38):
(38) a. ( N P [i]) [SEE N ? 1 [j] 2 [ ]]

b. [BE][( N P [i]) [SEENPjtjjNpJXi]]]
Oc
c. [BE[EASY[PLEASE NP, N p, [*i]]] N P [ i ] ]

These functional structures are well-formed and provide the basis for the inter-
pretations of the examples in (36). 4 (In structure (38c), the unindexed NP t
term is interpreted as indefinite by a convention proposed in Horn 1983,
section 2.2.
We see then that an analogous generalization to Chomsky's W/J-movement
generalization can be made in the proposed analysis.
The ungrammatically of examples like the following suggests that the
[e]-anaphora rule must be constrained:
(39) a. *Who did you see the man who hit

b. *Who did that the boss fired annoy Mary
c. *Whose did you see employer
d. *Who did John destroy a book about
e. * About whom did John destroy a book
f. *By which Greek authors does he have books
g. *Who was it John that hit
h. *Who did these books John give to
i. *Who did you wonder who saw
j. *Which sonatas is that violin easy (for us) to play on
220 George . Horn
The following constraints on the [e]-anaphora rule will account for the un-
grammaticality of the examples in (39):
(40) a. The Noun Phrase Constraint: The [e]-anaphora rule can-
not apply if either term is properly contained in an NP.
b. The Single Gap Constraint: The [e]-anaphora rule cannot

apply to the following configuration:
. . . NP; . . . NPj . . . g [ . . . [e] . . . [e] . . .]
In addition to the two constraints, the following 'processing strategy' pertains

to the application of the [e]-anaphora rule:
(40) c. The NP/Gap Order Condition: The following sequence

is always interpreted by relating the leftmost NP to the
rightmost empty node, and the rightmost NP to the left-
most empty node:
. . . N P j . . . NPj . . . [e] . . . [e] . . .
The Noun Phrase Constraint (NPC) was first proposed by Horn 1974:1143.
It is discussed in Horn 1975:348, Bach - Horn 1976:279-284,Horn 1979:
9 5 - 1 0 1 , and Horn 1983, section 3.1. The Single Gap Constraint (SGC) has
appeared in various formulations, and the NP/Gap Order Condition (NPGOC)
is discussed by Bach, who attributes it to Arlene Berman, and by Chomsky,
who credits it to Bordelois. The formulation of the NPGOC in (40c) differs
slightly from Bach's formulation.
I will not repeat the arguments justifying this set of constraints for English
and supporting it over other possible sets. Rather, I will demonstrate how the
constraints predict the ungrammatically of examples like (39).
Consider examples (3 9 a - f ) . The surface structures of these all contain empty
nodes which are properly contained in an NP. The structure of (39 d) is shown
below:
(41)
COMP
in destroy a book about Np[ e ]

Suppose that who is indexed as i, John as /, and a book about [e] as k. The
NPC prevents the [e]-anaphora rule from applying to this structure.
The following functional structure is derived from (41), by the application
of index insertion and semantic relations assignment:
(42) ( N P [ w h 0 i ] ) [DESTROY NFl [j]Np2[k]]

This functional structure is not well-formed. It contains an indexed N P term

which is not assigned a semantic relation, and furthermore, the prefixed N P
does not bind a term in argument position. Consequently, ( 4 2 ) does not form
the basis for a possible interpretation of example ( 3 9 d). Because no well-
formed functional structure can be derived from the surface structure o f this
example, it cannot be interpreted and is ungrammatical. It is easy to see that
this situation will arise whenever there is a constituent in COMP in the syntac-
tic structure and the [e]-anaphora rule is blocked, and the ungrammatically
o f ( 3 9 a - c ) and ( e - f ) is accounted for in the same way.
Examples (39g, h) and ( i ) have surface structures in which there is a clause,
S, which contains two empty nodes. The structure o f ( 3 9 i ) is shown below:
(43) ^
COMP S
V S
who you wonder who [ e ] saw [ e ]
The SGC blocks the application of the [e]-anaphora rule to this structure. As-
sume that the leftmost who is indexed as i, you is indexed as /, and the right-
most who is indexed as k. The following functional structure is derived from
this surface structure:
(44) ( N P [ i ] ) [WONDER NPl [j] [(NPM) [SEE]]]

a
Structure ( 4 4 ) is not well-formed. It contains two indexed N P terms that do
not bear semantic relations. Because no well-formed functional structure can
be derived from the syntactic structure in (43), example ( 3 9 i ) cannot be inter-
preted and is ungrammatical. The same type o f explanation applies to the un-
grammatical examples in ( 3 9 g ) and (h).
222 George . Horn
Finally consider example (39j), repeated below:
(39) j. * Which sonatas is that violin easy to play on
Such examples contrast with examples like the following, which are structural-
ly identical and differ only in the surface order of the NPs that occur in them:
(45) Which violin are the sonatas easy to play on
The surface structures of these examples are shown in (46 a) and (b) respec-
tively:
(46) a.
which sonatas violin BE easy to play [e] on [e]
b. X
COMP
NP NP VP
which violin sonatas BE easy to play [e] on [e]
In these structures, there is a clause S which contains two gaps. However, only
one of the antecedent NPs (which sonatas in structure a and which violin in
structure b) is located outside of that S. Therefore the [e]-anaphora rule can
apply in both structures to bind the empty nodes to the NPs which sonatas
and violin in structure a and which violin and sonatas in structure b. In ac-
cordance with the NPGOC, the leftmost [e] in structure a is bound to the
rightmost NP, violin, and the leftmost [e] in structure b is bound to the right-
most NP, sonatas. Then the rightmost [e] in both structures is bound to the
leftmost NP.
Assuming that the NPs are indexed as shown, the application of the [e]-
anaphora rule to the structures in (46) produces (47):
wft-sonatasj violinj BE easy to play [xj] on [ x j

b. .S.
w/i-violinj sonatasj BE easy to play [xj] on [Xj]
The following functional structures are derived from these surface structures:
(48) a. ( N P [i]) [[BE[EASY[PLAY N P , 2[;][

Np[î]]]]] NPj [j]]
b. ( N p[i])[[BE[EASY[PLAYNP, 2[;][

Np[ x i]]]]] NP[ Lj]]
7
These structures are well-formed and thus provide the basis for interpretations
of examples (39j) and (45) respectively. However, the interpretation based on
the structure in (48a) is nonsensical. That violin is interpreted as the object of
play and which sonatas is interpreted as the object of the preposition on. Ex-
224 George . Horn
ample (39j) is unacceptable for the same reason that the following examples
are unacceptable:
(49) a. *That violin is easy to play on the sonatas
b. *It is easy to play that violin on the sonatas
In contrast, the interpretation based on the functional structure in (48b), for
example (45), is acceptable. Which violin is interpreted as the object of the
preposition on and the sonatas is interpreted as the object of the verb play.
This interpretation is analogous to the interpretation of the following, accept-
able, examples:
(50) a. The sonatas are easy to play on that violin

b. It is easy to play the sonatas on that violin
The NPC, SGC, and NPGOC suffice to constrain the [e]-anaphora rule and
provide an account of island phenomena within the context of the proposed
framework. No notion of cyclicity is required to account for these facts, and
the cycle can thus be eliminated in this framework.
The true test of any analysis, of course, lies in how easily it can be extend-
ed to account for additional relevant data. The obvious question to ask at this
point is whether the proposal outlined in this section will account for island
phenomena in other languages. In the next section, I will look at an analogous
range of data in Polish, which superficially pattern quite differently from the
English data, and I will show that the same generalizations apply to both
languages.
2 Extension to Polish
The Polish analogs of wh-words appear in sentence-initial (COMP) position, as

shown below:
(51) a. KogoJan widzial
Who (acc) John (nom) saw
'Who did John see'
b. kim Jan pisal
About whom John wrote
'About whom did John write'
The surface structures of these examples are shown below:
(52) a.
COMP
NP
1>r
acc NPnom
1,1
Kogo Jan widziai [e]

b. S.
COMP
kim Jan pisat [e]

As mentioned in the first section, grammatical relations in Polish are defined
morphologically, in terms of case-marking. In structure (52a) kogo is identi-
fied as NP acc in COMP and can be indexed as i, and Jan is identified as N P n o m
and can be indexed as /. In structure b, the PP kim in COMP can be indexed
as i and Jan, N P n o m , can be indexed as /. The [e]-anaphora rule can now apply
to bind the empty node in each structure to the constituent in COMP. Before
the application of the [e]-anaphora rule, the empty nodes, of course, have no
case-marking. After the empty node in structure (52a) is bound to the NP in
COMP, however, it can be identified as NP a c c . The following structures are
produced by applying the [e]-anaphora rule:
(53)
kogoi Jan} widziai [xj]

S _
COMP S
PPj NP,
kim Janj pisat

The complex functional structures of the examples in (51) are derived from
the following (as discussed above for English):
226 George . Horn
(54) a. ( N p[+WH]) [WIDZIEC NP n o m NP a c c ]

b. (pp[. ,+WH. .]) [PISAC NPnom PP]
Insertion of indices and assignment of semantic relations to indexed NP terms

produces the following:
(55) a. ( N P a c c [i]) [WIDZIEC p n o m [j] P acc [ x i]]

b. (PPJONP^JIC]]) [P/^CNpnom[j]pp[Xi]]
7 Of
k = kim
These functional structures are well-formed and thus provide the basis for the
interpretations of these examples.
As in English, if the [e]-anaphora rule is blocked for some reason, or does
not apply, to a given syntactic structure like the above, then the functional
structure ultimately derived from it will not be well-formed. To see this, con-
sider the following:
(56) a. *Kogo Jan widziat Marka

who (acc) John (nom) saw Mark (acc)
'Who did John see Mark'
b. *0 kirn Jan widziai
about whom John saw
'About whom did John see'
c. *Kto Jan widziat
who (nom) John (nom) saw
'Who did John see'
The surface structures of these examples are the following:
(57) a.
COMP s
NPacc NP,nom
Kogo Jan widziai Marka

CA evidence for constraints o n transformations 227
b. S
COMP s
PP NP,nom V NP
kim Jan widziai [e]
c. S.
NP
Kto Jan widziai [e]
Structure (57a) contains no empty node and structure b contains an empty

node of a different category from the constituent in COMP. The [e]-anaphora
rule cannot apply to either structure. In structure c, the [e]-anaphora rule can
apply to bind the empty node to kto, which, however, is 'incorrectly' case-
marked as N P n o m rather than NP a c c (as in 51a)). Suppose that in structure a,
kogo is indexed as i, Jan is indexed as /, and Marka is indexed as A:; in struc-
ture b, kim is indexed as i, kim is indexed as / , and Jan is indexed as k\and
in structure c, kto is indexed as i, and Jan is indexed as /. The complex func-
tional structures of the examples in (56) are derived from the following:
(58) a. ( N P [ + W H ] ) [WIDZIEC N P N O M N P A C C ]
b. (pp[. ,+WH. .]) [WIDZIEC HPnom NP a c c ]
c. ( N P [ + W H ] ) [WIDZIECHPnomnom N P A C C ]
The application of index insertion and assignment of semantic relations pro-

duces the following:
( 5 9 ) a. ( N p a c c [ i ] ) [ W I D Z I E C N P n o m ] NP a c c [k]]

b. r D D.ro WD m n \wiDziEc mo [k]]
None of these functional structures are well-formed. Structures a and c con-

tain an indexed NP term that does not bear a semantic relation, and struc-
ture b contains a prefixed PP term that does not bind any term in argument
position. 5 Consequently, they do not provide the basis for interpretations of
228 George . Horn
the examples in (56). Because no wellformed functional structures can be

derived for these examples, they cannot be interpreted, and this accounts for
their unacceptability. No conditions need be placed on the rule that inserts
'w/j'-words (and phrases containing them) into COMP or on the optionality of
lexical insertion to account for these facts. It need not, for example, be stipu-
lated that 'w/i'-words can only be inserted into COMP position of structures
that contain empty nodes; that they can only be inserted into the COMPs of
structures which contain a verb with the proper subcategorization features;
or that they must be properly case-marked.
The rule of [e]-anaphora, then, applies in the derivation of both Polish and
English sentences to bind an empty node in the structure to a constituent in
COMP. It is unbounded, non-cyclic, and no command/precedence conditions
on it are necessary. In Polish, as in English, this account of 'wh' data extends
to topicalization data. I now turn my attention to island phenomena in Polish.
The pattern of grammaticality of the relevant Polish data is quite different
from that of the English data. To see this, first consider the following examples:
(60) a. Jaki numer wykrciles

Which number (acc) (you) dialed
'Which number did you dial'
b. Ktorych studentow oblales dzisiaj
Which students (acc) (you) failed today
'Which students did you fail today'
c. Jakie gazety najczsciej czytasz
Which newspapers (acc) (you) read most often
Which newspapers do you read most often'
(61) a. Jaki wykreciles numer
Which (acc) (you) dialed number (acc)
'Which number did you dial'
b. Ktorych oblales studentow dzisiaj
Which (acc) (you) failed students (acc) today
'Which students did you fail today'
c. Jakie najczsciej czytasz gazety

Which (acc) most often (you) read newspapers (acc)
'Which newspapers do you read most often'
The examples in (60) are structurally analogous to their English translational
counterparts, and present no problems for the analysis. The syntactic struc-
ture of (60a) is shown below:
(62)
COMP
NP
I
1- acg
jaki numer (ty) wykr$ cites
(I assume that the pronominal subject ty 'you' is present and is later deleted.)
Suppose that jaki numer, NP acc in COMP, is indexed as i. The [e]-anaphora
rule can apply as above to bind the empty node to this constituent. The fol-
lowing well-formed functional structure can be derived for example (60a):
(63) ( N P a r c [i]) [ W m ? C / C N p n o m [ j ] Npa,jxi]]

a
i = jaki numer,) = ty
In the same way, functional structures can be derived for the other examples
in (60).
The examples in (61) are also grammatical, and are variants of (60). Their
surface structures, however, contain a constituent in COMP that must be re-
lated to an empty node which is properly contained in an NP elsewhere in the
structure. The syntactic structure of (61a), shown below, illustrates this:
(64)
COMP
I
ADJ
Jaki (ty) wykr$cites [e] numer
The grammatically of such examples demonstrates that the [e]-anaphora rule

can apply into an NP in apparent violation of the NPC. If we assume that it
applies to bind [e] to jaki, the following well-formed functional structure can
be derived for example (61a):
(65) ( A a c c [i]) [WYKR^CICNPnomD]NPacc[A[Xi]k]]

i = jaki", j = ty, k = numer
The functional structure in (65) is equivalent to the functional structure in
(63) since both contain indices that represent the same lexical items, and both
contain the same predicate (wykr^cic).
230 George . Horn
If the adjectival form jaki in COMP does not bind the empty node in
structure (65), the functional structure derived from it will not be well-formed.
Such a functional structure is shown in (66):
(66) ( Aacc [i]) [WYKR^CIC N p n o m [j] NPacc[k]]

i = jaki', j = ty \ k = [e] numer
In (66), the prefixed term A [i] does not bind a term in argument position,
and this structure does not meet the requirements of the Well-formedness
Condition as formulated in (4).
We might conclude from these data that the NPC is not a universal con-
straint, and does not apply to block the application of the [e]-anaphora rule
in Polish. If this were the case, it would suggest that the explanatory power of
the proposed set of constraints is not great.
It is not true, however, that the [e]-anaphora rule can apply freely into
NPs in Polish structures. The following examples, like their English counter-
parts, are ungrammatical:6
(67) a. *0 kim on zniszczyt ksiqzk
About whom (he) destroyed book (acc)
'About whom did he destroy a book'
b. *0 kim Jan podari artykul
About whom John tore up article (acc)
'About whom did John tear up an article'
c. *0 czym Jan przegapil film
About what John missed film (acc)
'About what did John miss a film'
d. *Z jakim dzieckiem Jan spotkal kobietf
With which child John met woman (acc)
'With which child did John meet the woman'
e. *Z jakiego miasta lubil ludzi
From which town (he) liked people (acc)
'From which town did he like the people'
f. *W ktorym pokoju podobaly mu si meble
In which room pleased him (dat) REFLEX furniture
(nom)
'In which room did he like the furniture'
The corresponding statements, in which the PPs are located in the NP with
the head noun, are, of course, grammatical:
(68) a. On zniszczyi ksiqzk$ oNixonie

'He destroyed a book about Nixon'
b. Jan podart artykul tym polityku
'John tore up an article about that politician'
c. Jan przegapil film Poznaniu
'John missed a film about Poznan'
d. Jan spotkal kobietq malym dzieckiem
'John met a woman with a small child'
e. Lubit ludzi Krakowa
'He liked people from Cracow'
f. Podobaly mu si meble w sypialni
'He liked the furniture in the bedroom'
The ungrammatical examples in (67) have surface structures like the following,
for (67 a) and (e):
(69) a.
okim Jan zniszczyi ksiqzkg pp[e]
Z jakiego miasta lubil ludzi pp[e]
In such structures, the NPC correctly blocks the [e]-anaphora rule, preventing
it from binding the PP [e] term to the PP in COMP, and the functional struc-
tures that can be derived from these surface structures, shown below, are not
well-formed because they contain a prefixed term that does not bind a term
in argument position:
(70) a. ( P P [ i ] ) [ZNISZCZYC N p n o m [j] N P a c c [ k ] ]

i = kim; j = Jan\ k = ksiqzk$ [e]
232 George . Horn
b . (pp[i]) [LUBIC N P n o m [j] NPacct^] ]

i = jakiego miasta\ j = on\ k = ludzi [e]
These functional structures cannot provide the basis for an interpretation of
examples (67a) and (e), thus accounting for their ungrammaticality. The un-
grammaticality of the other examples in (67) is accounted for similarly.
In contrast, empty PP nodes that are not dominated by NP can be bound
to PPs in COMP, as illustrated below:
(71) a. OkimJanpisal
'About whom did John write'
b. kim Jan napisat ksiqzk$
'About whom did John write a book'
c. Na jaki stoi Jan wskoczyt
Onto which table did John jump'
d. W jakim miescie spotkat Bill Jana
'In which town did Bill meet John'
These examples have syntactic structures like those shown below for (71a):
PP: NP,nom V PP
kim, Jan napisat [e]

The [e]-anaphora rule can apply in such structures, without violating theNPC,
to bind the empty PP node to the PP in COMP, and well-formed functional
structures like the following can be derived:
(73) ( P P i [ 0 N P [ j ] ] ) [ ^ / M d N p ^ f k J p p I X i ] ]
y a
Consequently, examples like (71) are grammatical.
Examples like the following, which contain an NP in the highest COMP,
show that it is not possible for the [e]-anaphora rule to apply to bind an empty
node within a (complex) NP to a term outside of that NP:
(74) *Jaki plaszcz weszia jakas dziewczyna ktora miala na sobie
Which coat came in a girl who wore
'Which coat did a girl who wore come in'
This example has a syntactic structure like that shown in (75):

(75) g[ c \jaki ptaszcz] s[weszta N?[jakas dziewczyna ktora
miata [e] na sobie\\\
As above, the NPC correctly blocks the application of the [e]-anaphora rule,
and the only functional structure that can be derived for this example is the
following:
(76) ( N P [i]) [WEJSCNPnomD] ]

a
i = jaki ptaszcz', j = jakas dziewczyna ktora miata [e] na
sobie
This structure is not well-formed and this accounts for the ungrammaticality
of (74).
So we see that examples like (67) and (74) support the NPC analysis, while
examples like (61) argue against it. The examples in which the NPC can be
violated differ from those in which it holds in the following way: In examples
like (61), the constituent in COMP is inflected in such a way that it can be
associated with the head noun of the NP constituent that contains the empty
node. For examples (67) and (74), this is not the case.
Adjectival forms like jaki are marked morphologically to agree with the
head noun that they are related to in gender, number, and case. In contrast,
PPs as units are not marked according to whether they occur in an NP, as in
(77), below, or S (or VP, depending upon the analysis), as in (78):
(77) $[Jan zniszczyi np[ksiqzk$ pp[o Nixonie]]]

(78) s [Jan napisat P P [o Nixonie]]
The PP in (77) is identical in form to the PP in (78). Similarly, NPs are not
marked according to whether they occur in a relative clause, or in some em-
bedded S that is not an NP complement. The above data can be accounted for
by assuming that the NPC is valid in Polish, but can be overridden just in case
there are morphological markers present in the surface structure that can
function to associate a constituent in COMP with an empty node elsewhere in
the structure.
In (61a), for example, the form of jaki indicates that it is the accusative
singular masculine form, and it can thus be associated with numer, which is
an accusative, singular, masculine noun. In (61b), the form of ktorych indi-
cates that it is the accusative plural masculine animate form, and it can thus
be associated with studentow, which is a masculine animate noun marked for
accusative case (plural).
234 George . Horn
In contrast, the PP kirn in (67 a) is not marked to agree with the noun
ksiqzk, and no morphological marker is present in the surface structure that
can serve to associate the two constituents. The situation is the same for the
other examples in (67).
This proposal is further supported by the following data. Consider example
(79), below:
(79) Jan zniszczyl zdfecie Marii

'John destroyed a picture of Mary'
The following are possible echo questions corresponding to (79):
(80) a. Jan zniszczyl zdjcie kogo
'John destroyed a picture of whom'
b. Jan zniszczyl czyje zdj$cie

'John destroyed a picture of whom'
Example (80 a) contains the form kogo, which is an interrogative noun marked
for genitive case. Example (80b) contains the form czyje, which is adjectival
in form and is marked as neuter, accusative case, singular, to agree with
zdj^cie, which is a neuter noun marked accusative singular. The proposed ana-
lysis correctly predicts that (81), below, is ungrammatical, while (82) is gram-
matical:
(81) *Kogo Jan zniszczyl zdjqcie
(82) Czyje Jan zniszczy I zdjgcie
The syntactic structures of these examples are the following:
(83) a.
COMP S
NP,gen
Kogo Jan zniszczyl [e] zdjcie
COMP S
czyje Jan zniszczyl [e] zdjecie

Kogo is not morphologically marked in such a way that it can be associated

with zdjgcie, while czyje, in structure b, is so marked. Therefore, in struc-
ture a, the [e]-anaphora rule cannot apply (because of the NPC) to bind the
empty node in the NP acc constituent to kogo in COMP, while in structure b,
it can apply to bind the empty node in NP acc to the adjectival form in COMP.
Consequently, the functional structure derived for example (81), shown in
(84a), is not well-formed, while that derived for example (82), shown in
(84b), is well-formed:
(84) a. ( N P [ i ] ) [ZNISZCZYCvPnomM NPacc[k] ]

i = kogo , j = Jan; k = zdj^cie
b. ( A [ i ] ) [ZNISZCZYCNPnom[j]
NP a c c [ a [ X i ] k ] ]
i = czyje; j = Jan; k = zdjqcie
This accounts for the contrast beween (81) and (82).
The proposed analysis correctly predicts that uninfected forms cannot be
related to empty nodes in NPs. One such form is the adverb bardzo Very',
which can occur in NPs, APs, and adverbial phrases, as illustrated below:
(85) a. Widziatem bardzo wysokiego m$zczyzn$
(I) saw very tall man
saw a very tall man'
b. On widzial bardzo ladny dom
He saw very nice house
'He saw a very nice house'
c. Studenci w mojej grupie sq bardzo wysocy
Students in my group are very tall
"The students in my group are very tall'
d. Czut bardzo niepewnie
(He) felt REFLEX very uncertainly (adv.)
'He felt very uncertain'
These examples have syntactic structures like those shown below:
(86) a. Widziatem wp[bardzo wysokiego mzczyzn]
b. On widzial [bardzo ladny dom]
c. Studenci w mojej grupie sq A p [bardzo wysocy]
d. Czut si Advp[bardzo niepewnie]
The proposed analysis correctly predicts that the topicalized sentences in a
and b, below, are ungrammatical because the NPC prevents the [e]-anaphora
236 George . Horn
rule from binding the empty node in the NP to the adverb bardzo in COMP.
These contrast with the acceptable c and d examples, to whose surface struc-
tures the [e]-anaphora rule can apply to bind the empty node to the adverb
bardzo:
(87) a. *Bardzo widzialem wysokiego m$zczyzn

Very (I) saw tall man
b. *Bardzo on widzial ladny dom
Very he saw nice house
c. ?Bardzo studenci w mojej grupie sq wysocy
Very students in my group are tall
d. Bardzo czul si niepewnie
Very (he) felt uncertainly
The surface structures of these examples are the following: 7
(88) a.
COMP
Adv NP n o m V
bardzo (ja) widzialem [e] wysokiego mgzczyzne
b. S.
COMP
Adv NP n o m V
bardzo on widzial [e] ladny dom
c. S.
bardzo studenci w . .. sq [e] wysocy

d. s
COMP S
bardzo on czut si [e]niepewnie
The following functional structures can be derived from examples (87ad):
(89) a. ( A d v [i]) [ WIDZIE C N P [j ] [k] ]

i = bardzo; j = ja; k = [e] wysokiego m?zczyzn
0. ( A dv[i]) [^WAdv[Xi]k]] N P n m]
i = bardzo; j = studenci w mojej grupie; k = wysocy
d. (AdvU)
VAOVL'J/ [CZUC
N P n o m [ uj
i>ir j ] Advp[Adv[Xi]
rtu k] ]
nom
i = bardzo', j =on\ k = niepewnie
Structures a and b are not well-formed, accounting for the ungrammatically

of examples (87a) and (b), while structures c and d are well-formed, and con-
sequently, examples (87c) and (d) are grammatical.
We see then, that the NPC is valid for Polish as well as English, with the
condition discussed here. The analysis accounts for a wide range of data. I
now turn to data that relate to the SGC.
The following examples are apparent counterexamples to the SGC,suggest-
ing that it does not hold for Polish:
(90) a. Kto komu dat prezent
who (to) whom gave present
'Who gave a present to whom'
b. Kto czego nie rozumie
who what not understand
'Who doesn't understand what'
c. Kto komu co przyrzekat
who (to) whom what promised
'Who promised what to whom'
238 George . Horn
Such examples have surface structures like the following:

(91) a. s [c[kto komu] s[[t] dal [e]prezent]}
b. s [c [kto czego] s [[e] nie rozumie [e]]]
c
- s [c[kto komu co]s[[e]przyrzekal [e] [e]]]
The [e]-anaphora rule must apply in each structure to bind the empty nodes
to the constituents in COMP in order that a well-formed functional structure
can be derived for each of the examples in (90). It might be argued that only
one 'w/z'-word is in COMP and that the others are merely scrambled to the
front of the clause S. However, if all 'wh'-words are not in S-initial position,
the sentences can only be interpreted as echo questions. Thus the following is
analogous to English echo questions like that used to translate it:
(92) to dal komu prezent
'Who gave whom a present'
It would be a coincidence that an obligatory scrambling rule applies in ''-
questions that has the same effect as w/z-movement, so such an alternative is
not plausible.
The following examples, in contrast to the ones in (90), indicate that the
SGC applies in some cases:
(93) a. Bill chciat s [zeby Janowi dano prezent]
Bill wanted that John be given present
'Bill wanted John to be given a present'
b. Jaki prezent Bill chcial s [zeby Janowi dano [e]]
'Which present did Bill want John to be given'
c. Komu Bill chcial s [zeby [e] dano prezent]
'To whom did Bill want a present to be given'
d. *Jaki prezent komu Bill chcial s [zeby [e] dano [e]]
'Which present to whom Bill wanted to be given'
(94) a. Bill mowil ze Jan dal prezent Adamowi
'Bill said that John gave a present to Adam'
b. Jaki prezent Bill mowil Jan dal Adamowi
'Which present did Bill say that John gave to Adam'
c. Komu Bill mwil Jan dal prezent
'To whom did Bill say that John gave a present'
d. *Jaki prezent komu Bill mwil Jan dal
'Which present to whom did Bill say that John gave'
C A evidence for constraints on transformations 239
The a examples are statements corresponding to the 'wA '-questions in b and c.

The b and c examples show that the [e]-anaphora rule can apply to bind a
single empty node in the embedded sentence to the word in COMP, and the
d examples show that two empty nodes in the subordinate clauses, S, cannot
be bound to the words in COMP in such structures.
The following examples are also ungrammatical:
(95) a. *Kogo kto powiedziat zeJanekpobii

Whom who said that John beat
b. *Kogo Janek wie kto spotkai

Whom does John know who met
These examples have surface structures like those shown in ( 9 6 ) :
(96) a. s[c[kogokto]s[[e]powiedziats[zes[ranekpobit[(;]]]]]
b s[dkogo}s[Janekwies[c[kto]s[[Q]spotkat [e]]]]]
Examples ( 9 3 d ) , (94d), and ( 9 5 ) all have surface structures which contain

clauses, S, which themselves contain two gaps, and thus the [e]-anaphora rule
cannot apply. Consequently, no well-formed functional structures can be de-
rived for these examples, and this accounts for their ungrammatically.
Four structure types in which the SGC might be expected to block the ap-
plication of the [e]-anaphora rule can be isolated. These are shown below:
( 9 7 ) a. s [ c [ X Y ] s [ - -[e]...[e]...]]
b- s [ c [ X Y ] st- [ e ] s [COMP s [ . . . [e] . . . ] ] ] ]
c. s [ c [ X Y ] s [ - s[COMPs[...[e]...[e]...]]]]
d. s[c[X]s[.-.s[c[Y]s[-..[e]...[e],..]]]]
Examples (93 d) and ( 9 4 d ) have structures of the type ( 9 7 c ) . Example (95a)

has a structure of type (97b), and example (95 b) has a structure of type(97d).
The grammatical examples in (90), which are problematic for the SGC ana-
lysis, have structures of type (97 a). T o understand the distinction between
these and the ungrammatical examples, it is necessary to look again at case-
marking in Polish.
All nouns are casemarked according to their syntactic function with respect
to the verb that they occur with. Casemarking plays the same role in Polish
that syntactic configuration plays in English. Consequently, word order in
Polish is relatively free. Thus the examples in b-e below are variants of a (as
previously observed):
240 George . Horn
(98) a. Jan widzial ksiqik

'John saw the book'
b. Jan ksiqzk? widzial
c. Ksiqzk$ Jan widzial
d. Ksiqzke widzial Jan
e. Widzial ksiqzk Jan
Even though the basic syntactic configuration of (98a) has been destroyed in
b through e, the casemarkings on the nouns serve to identify their syntactic
function and associate them properly with their verb. (This is analogous to
the function of casemarkings to associate modifiers and head nouns as dis-
cussed above.)
The grammatical examples in (90) consist of a single clause, and even
though the w/i-words have been moved out of the clause (S) to COMP of S,
they are no more difficult to process than the examples in (98). Consequently,
in these cases, the SGC can be overridden.
Structures of type (97 b, c) and (d) are more complex and contain more
than a single clause. Hence examples like (93d), (94d) and (95) are not ana-
logous to examples like those in (98). In these cases, the SGC functions in
Polish, as in English, to limit the set of surface structures that can be inter-
preted.
To summarize, Polish and English compare and contrast in the following

ways in the context of the proposed framework.
Verbs and verblike constituents in both languages (and in all languages)
have functional structures, specified in their lexical entries, of the form shown
below:
(99) PRED NPX NP y . . .

The term PRED stands for 'predicate'. NPX and NP y represent terms defined
according to their grammatical relation to the predicate, and a and represent
semantic relations. The same operations on functional structures (such as in-
dex insertion and semantic relations assignment) apply in both languages to
produce complex functional structures from syntactic surface structures.
Well-formed functional structures in both languages must satisfy the criteria
of the Well-formedness Condition in (4), and only well-formed functional
structures can form the basis for the interpretations of the examples from
whose syntactic surface structures they are derived. If no well-formed func-
tional structure can be derived from the syntactic structure of a given sen-
tence, then that sentence is ungrammatical.
The most significant difference between English and Polish at this level is
that grammatical relations, to include the syntactic associations between con-
stituents such as head nouns and their modifiers, as we have seen, are con-
figurationally indicated in English and morphologically indicated in Polish
(while configuration plays only a marginal role.). As a consequence of this
distinction, surface word order in Polish is relatively freer than it is in English.
The overall mechanics of w/2-movement is the same in both English and
Polish. In both languages, 'w/i'-words (and constituents containing them) oc-
cur in COMP (in S-initial position), and the [e]-anaphora rule operates to bind
empty nodes to these ''-words (and constituents). In both languages, this
rule is unbounded and applies non-cyclically, and it is constrained by the
same island conditions: the Noun Phrase Constraint (NPC) and the Single Gap
Constraint (SGC). 8
The quite different pattern of grammaticality observed in the Polish data is
another consequence of the 'morphological/configurational' distinction be-
tween the two languages. In those situations in which constituent associations
are not indicated by a surface structure morpheme, the constraints cannot be
violated. Thus the NPC prevents PPs in COMP, which are not marked in a spe-
cial way when they are related to the head nouns of NPs, from binding empty
PP nodes located in NPs, as illustrated by the ungrammaticality of examples
like (67). In complex sentences like (93 d), (94d) and (95), casemarking is not
sufficient to associate the constituents in COMP with empty nodes elsewhere
in the structure, so the SGC prevents the empty nodes from being bound to
the constituents in COMP. Examples like these are ungrammatical in both
English and Polish. However, because English lacks casemarking, the condi-
tions necessary for overriding the NPC and SGC (to relate adjectives in COMP
to head nouns, for example) never occur.
Another way in which Polish and English differ is that in Polish surface
structures more than one constituent can occur in COMP, while this is not
possible in English surface structures. These surface differences tend to obscure
underlying, and more fundamental, similarities between the two languages.
In the proposed framework, the effect of casemarking on the applicability
of the [e]-anaphora rule is to be expected. The [e]-anaphora rule functions to
relate a constituent A to some constituent in constructions of the form
shown in (100):
Ii.:",: ':::!!
The constituent A can be any of several constituent types, as we have seen,
such as NP, PP, Adj, or Adv, and can be a verb (PRED) in structures like
(101 a) and (b) below, or the head noun of an NP in structures like (101 c), for
example:
242 George . Horn
(101) a. Npjwfco] did John believe s [that Bill liked N p[e]]

A
b. pp[about whom] did John want v p [ t 0 write pp[e]]
A
c. Aaki] Jan wykr$cit Np[[e] numer]
A
Because casemarking is one device for providing additional information about
the grammatical function of a constituent (that is, more information about
which constituents it is related to, and the nature of the relationship) the [e]-
anaphora rule can apply to Polish structures in situations where it cannot ap-
ply to analogous English structures.
It might not, in fact, be totally accurate to think of [e] nodes in Polish as
fixed, structurally definable constituents in phrasemarkers, as they are in
English, but rather to consider them to be potential word or phrase 'slots' in
internally unordered strings of category NP, AP, or S. Because grammatical
function can be determined by morphological markers alone, words (that is,
lexical items of category N, A , . . . ) as well as phrases (NP, AP,. ..) can be re-
lated in configurations like (100) without recourse to structural considerations.
The way in which empty nodes are thought of, however, does not affect the
present discussion, and I will not pursue it here. The important point that
I have attempted to make is that in the context of the proposed framework,
a unified account of so-called island phenomena can be formulated for Polish
and English.
3 Chomsky's analysis
I do not intend to argue against Chomsky's framework in this paper, and will
not discuss it in detail. For a more extensive treatment of it as it applies to
English, see the Chomsky papers listed in the bibliography. Rather, I wish to
point out that, in its present version, his analysis is unable to account for the
Polish data discussed in the preceding section.
Chomsky attempts to account for island phenomena by introducing the
notion of 'bounding' node and formulating the Subjacency Condition in terms
of bounding nodes as shown below:
(102) The Subjacency Condition: No transformation can apply to
move a constituent over more than one bounding node.
(His other constraints, the Nominative Island Constraint and the Opacity Con-
straint, also play a role in accounting for island phenomena, but I will not dis-
cuss them here.)
His rule of w/i-movement applies to move a wh-word, or constituent con-
taining a w/i-word from its original position in the structure to the COMP of
the S in which it is located, and then from COMP to COMP until it arrives at
its surface position (usually in the COMP of the highest S in the structure).
Because of the nature of the contraints, w/z-movement must apply cyclically.
In surface structures, traces, or empty nodes from which constituents have
been moved, must be bound by these constituents. An additional condition
on binding is that the antecedent constituent must c-command its trace.
The following example illustrates the mechanics of w/j-movement in this
framework:
(103) Who did John tell Mary that Bill hit
Example (103) has the following surface structure:
(104) s l c j1 w h o ] s [ J o h n tell Mary g [c2[ s [Bill hit [i]]]]]

v z^ '
The rule of wft-movement has applied to move who out of the embedded S
into C 2 , and then from C 2 to C j , its surface position. None of Chomsky's
conditions are violated, and the example is grammatical.
Now consider the following:
(105) a. *Who did John destroy a book about

b. *Who did John wonder who saw
These examples must be derived from structures like the following:
(106) a. s [COMP s [John destroy NP[a book about WHO]]]
b. [COMP s [John wonder s[COMP 2

sfWHO, saw WH0 2 ]]]]
In structure (106a), movement of WHO out of NP and S to the COMP of S
violates the Subjacency Condition, and is therefore disallowed. Consequently,
the ungrammatical example (105a) cannot be derived.
In structure (106 b), suppose first that WHO! is moved to COMP2 and
then to COMP of the highest S, leaving a trace in its original position and in
COMP 2 . Now, if WH0 2 is moved to COMP 2 , a structure will be created in
which the trace of WHOi in COMP2 no longer c-commands the trace of
WHOi in the original position of WHO]. If, instead, WH0 2 is first moved to
COMP 2 , then subsequent movement of WHOi to COMP2 will create a struc-
ture in which WH0 2 no longer c-commands its trace in the embedded S. The
244 George . Horn
Subjacency Condition prevents the movement of either wA-word from its

original position directly to the leftmost COMP. This accounts for the ungram-
matically of example (105b).
Now consider the following Polish examples:
(107) a. Jaki wykr$cites numer

b. Ktorych oblales studentow dzisiaj
The grammaticality of these examples can be accounted for by saying that

either NP or S is not a bounding node in Polish, and deriving them from struc-
tures like the following by applying w/i-movement as indicated:
(108) a. [COMP s [wykr^ciles NP[jaki numer]]]

b. g[COMP s[oblales N p[ktorych studentow] dzisiaj]]
However, if this is done, then the Subjacency Condition will not prevent ex-
traction from the structure shown in (109) to block the derivation of the un-
grammatical example in (110):
(109) s [COMP s [Jan spotkal Nr[kobiet jakim dzieckiem]]]

(110) *Z jakim dzieckiem Jan spotkal kobiet$
Presumably, if either NP or S is not a bounding node in structures like (108),

then the same applies to structures like (109), which are identical except for
lexical material. Consequently, in Chomsky's analysis, it is not possible to
distinguish examples like (107) from ones like (108).
One might argue that Chomsky's analysis can be modified such that the
Subjacency Condition can be 'overridden' just in case the constituent to be
moved is morphologically marked in such a way that it can be associated with
its original position in the structure. However, in Chomsky's analysis as it
stands, the Subjacency Condition is a structural condition on rule application,
and not a condition on any sort of anaphora, or binding, rule. There is no in-
trinsic connection between the subjacency principle and the factors involved
in the determination of syntactic associations (such as casemarking and syn-
tactic configuration). Even if such a modification is made, it will constitute a
mere ad hoc addition to the analysis.
I now turn to examples like (90), one of which is repeated below:
(90) a. Kto komu dal prezent
The grammaticality of such examples suggests either that the wh -movement
rule itself is fundamentally different in Polish than in English, or that the
c-command requirement is not valid for Polish.
KVz-movement, for example, might be analyzed as attaching -words to

COMP in such a way that they are sisters at the surface structure level, as
shown below:
(111) s i c P r o ] [komu]\s[[i\dal [t]prezent]]
In structure (111), each 'w'-word in COMP c-commands its trace. No other

constraints are violated, and example (90a) is therefore grammatical.
Alternatively, wft-movement in Polish can be analyzed as applying as it
does in English, in which case the structure of (90a) is the following:
( 2 ) s [ e t c k o m u ] s[[t] dal [t]prezent]]
Here, I assume that kto is first moved to COMP, and that komu is subsequent-
ly moved and adjoined to the COMP node containing kto. In this structure, as
in analogous structures for English examples, kto does not c-command its
trace in S. Thus, in this alternative, the c-command requirement cannot be
maintained for Polish.
A final alternative is the one briefly mentioned in section 2, in which only
the first 'w/i'-word is located in COMP and the remaining ones are merely
scrambled (obligatorily) to the initial position of the clause, S. Under this al-
ternative, the c-command condition can be retained.
None of these alternatives is desirable. Under the first and third, a universal
rule of w/?-movement cannot be maintained with the concomitant loss of any
explanation for the observed similarities between Polish and English (and many
other languages). Under the second alternative, the universality of the c-com-
mand condition on anaphoric binding cannot be maintained, and yet this is
precisely the sort of condition that, if required by one language, would be ex-
pected to be universally required.
Furthermore, there is no way in Chomsky's analysis to distinguish ungram-
matical examples like (93 d), (94d), and (95) from grammatical examples like
(93 b and c) and (94b and c). Examples (93d), (94d), and (95) are repeated
below, along with their syntactic structures:
(93) d. *Jaki prezent komu Bill chciat zeby dano
s tc \jaki prezent komu] s[Bill chciat [c[zebv] s[t dano t]]]]
(94) d. *Jaki prezent komu Bill mwil Jan dal

s\jaki prezent komu] s[Billm0wil <3 [COMP s \Jan dal 11]]]]
(95) a. *Kogo kto powiedzial ze Janek pobil

s [c [fcogo kto]^sJt powiedzial [cj/e] s[Janek pobilj]]]]
246 George . Horn
b. *Kogo Janek wie kto spotkul

s [ c s [ J a n e k wie s [ c j f r e h p spotkalj]]]]
To see this more clearly, suppose first that either S or S is a bounding node,
but not both. In (93 d), on the lower cycle, both jaki prezent and komu can
be moved to the rightmost COMP, and then on the highest cycle, both can be
moved to the leftmost COMP. The same movement can apply to the structure
of (94d). In the structure of (95a), on the first cycle, kogo can be moved to
the rightmost COMP. On the next cycle, both kogo and kto can be moved to
the leftmost COMP. In the structure of (95b), kto and kogo can be moved to
the rightmost COMP on the first cycle, and on the next cycle, kogo can be
moved to the leftmost COMP. In no case is the Subjacency Condition violated,
and, as we have seen in the previous case with examples like (90), more than
one 'wh'-word can occur in a single COMP. Thus, ungrammatical examples like
these can be derived under these assumptions.
If both S and S are bounding nodes, the derivation of these examples is
correctly blocked by the Subjacency Condition, and at the same time, exam-
ples like (90) can be derived (if we make the undesirable assumptions concern-
ing c-command or the \-movement rule that were discussed above). How-
ever, in this case, w/j-movement will be blocked in the structures of examples
like (93b) and (c), and (94b) and (c), which are repeated below:
(93) b. Jaki prezent Bill chciai zeby dano Janowi
s [ c \Jakiprezent] s[Bill chcial s [ c t f f i y j s[dano Janowi jt]]]]]
c. Komu Bill chciat zeby dano prezent

s [c[komu] schciat [c[zeby\s[dano\t]prezent]]]]
(94) b. Jaki prezent Bill mwil Jan dal Adamowi

s [c[jakiprezent] s[Billmwil g [COMPs[/AM dal Adamowi]]]]
c. Komu Bill mwil Jan dal prezent

s [c [komu]
v s mwil [COMP s [Jan dal prezent [t]]]]]
^
Consequently, the analysis incorrectly predicts that examples like these are
ungrammatical.
Examples like the following demonstrate that COMPs of embedded sen-
tences can be filled by more than one VA'-word:
(113) a. Jan wiedzial kto komu dal prezent
John knew who whom gave present
'John knew who gave a present to whom'
b . Jan / ciekawy } fao iaki prezent dal Billowi

[ zastanawiat sif I
'John wondered who gave which present to Bill'
Thus the ungrammaticality of examples like (93 d), (94 d), and (95) is not due
to the fact that more than one constituent cannot be moved into a COMP of
an embedded sentence.
So we see that the Subjacency Condition/bounding node analysis makes
the wrong predictions for Polish, and some additional condition or mechanism
is needed to account for these data. Because Chomsky's analysis does not
formally relate the mechanisms for determining grammatical function (and, in
general, the associations between constituents in syntactic structures) to con-
ditions on movement rules, there is no straightforward way in which his anal-
ysis can be modified to account for the relevant data.
To conclude, I have constructed a contrastive analysis of Polish and English
w/i-questions, primarily to show that in spite of superficial differences, the
two languages share certain fundamental characteristics and properties. Of
secondary importance here was the demonstration that Chomsky's analysis
does not account for the Polish data. The methodology of contrastive analysis,
of course, has other than purely theoretical applications, and its importance
as a tool for linguistic research cannot be overestimated.
Notes
1. It is important to note that, because grammatical relations are defined in terms of case-
marking, syntactic configuration, and hence, word order, are not relevant in identify-
ing NPs so that their indices can be inserted into the correct positions in functional
structures. The N P n o m and N P a c c terms can be identified as Jan and Marka in the fol-
lowing, which are variants of (5 b):
i. a. Widziat Jan Marka
b. Marka Jan widziat
c. Jan Marka widziat
and so forth.
Assuming, as above, that Jan is indexed as i and Marka as /, the functional structure
shown in (14), repeated below, can be derived for these examples:
(14) W/DZ/ifC'NP n o m [] NP a c c l

Consequently, word order in Polish is relatively freer than it is in English.
2. Example (15c) and the following are synonymous:
i. What did John read about
The functional structure of i is shown below:
ii. ( N P [ i ] [READ NPj 1 pp[ABOUT N P l x i l l l

248 George . Horn
This structure and (21 c) are equivalent because they contain the same predicate, read,
and the same NP terms, John and what, which are assigned the same semantic relations,
and 7, respectively.
3. If the [e]-anaphora rule binds the empty node in structure (26c) to Np(what] in
COMP, the prefixed PP term in the functional structure derived from this syntactic
structure will still not bind a constituent in argument position, and the functional
structure will not be well-formed.
4. The [e]-anaphora rule can apply to bind empty nodes to antecedents that are not lo-
cated in COMP, in structures like (37 c). However, if the rule applies to bind the empty
node to we in (37a), or to either we or it in (37 b), then the functional structures de-
rived from the resulting surface structures, shown below, will not be well-formed:
L (NPI'1) [SEE NPj [j] NP2I x jll

ii. a. [BE[( N P [i])[SEE N P l [j]NP 2 [Xjl]]]

i = John; j = we
b. [ B E [ ( N p [ i ] ) [ S E E N P l [ j ] N p 2 [ x k ] ] ] ]

i = John; / = we; k = it
All three functional structures contain a prefixed NP term that is not assigned a se-
mantic relation and does not bind a term in argument position.
5. If the empty node in structure (57b) is bound to the NP, kim, in COMP, the func-
tional structure produced will still not be well-formed, as it will contain the prefixed
PP term, kim, i, which does not bind a term in argument position.
A more complete explanation for the ungrammatically of example (56 c) is, per-
haps, called for. The application of the [e]-anaphora rule to structure (57 c) produces
the following:
-S.
nom
The functional structure of widziec contains a single N P n o m term and a single NP a c c

term. In structure i, kto is identified as the term in COMP and its index, i, is inserted
into the NP term that is prefixed to the functional structure of widziei, as shown in
(59c). At this point, either the index of Jan, which is an N P n o m constituent, or the
bound variable, Xj, which is also an N P n o m constituent by virtue of the fact that it is
bound to an N P n o m constituent in COMP, but not both, can be inserted into the
NPnom s ^ o t ' n t l l e functional structure of widziei. Because there is no NP a c c constit-
uent in structure i, no index can be inserted into the N P a c c slot in the functional
structure of wiedzied. As a result, either the functional structure in (59c), which is
not well-formed, or the following functional structure, can be derived from syntactic
structure i:
(NP nom 1[i])

" WIDZIEC 1
> r l [Xi]]
111
a nom
This structure is, technically speaking, well-formed. However, in order for a functional
structure to provide the basis for a full interpretation of a given syntactic structure, the
functional structure must contain the indices of all the indexed constituents in the
surface structure. Moreover, widziec requires an N P a c c term in its functional structure

for a complete interpretation. Consequently, the functional structure in ii cannot pro-
vide the basis for a complete interpretation of structure i, and this accounts for the
ungrammatically of example (56 c), since the only other functional structure that can
be derived for it is the one in (59c), which is not well-formed.
6. Some speakers accept (67 a, b) and (c). For a possible explanation of this, see Horn
1979:153-165.
7. Actually, the adverb bardzo need not be located in COMP in the structures in (88). Its
location in the syntactic structures is not relevant to the present discussion.
8. The fact that the [e]-anaphora rule is unbounded does not imply that it can always ap-
ply unboundedly. In both Polish and English, there are verbs into whose complements
this rule cannot apply. Two English examples are the following:
i. a. *Who did John know that Bill saw
b. *Who did John wonder Bill saw
These contrast with the examples in ii.
ii. a. John knew who Bill saw
b. John wondered who Bill saw
Any analysis of wA-movement must somehow account for these data. Following
Chomsky and others, we might incorporate the notion of 'bridge' conditions into the
analysis, formulated in such a way as to constrain the [ej-anaphora rule. This would
not require any other major modifications to the analysis, and I will not discuss the
mechanics of doing it. I will add, however, that another difference between English and
Polish is that in Polish, there are fewer 'bridge'verbs, and consequently.it appears on the
surface that the process is bounded. Examples like (93b) and (c) and (94b) and (c) do
not represent as widespread a pattern in Polish as their counterparts do in English.
References
Bach, Emmon
1977 "Comments on the paper by Chomsky", Formal syntax, edited by Peter
Culicover, Thomas Wasow and Adrian Akmajian (New York: Academic
Press), 133-155.
Bach, Emmon - George M. Horn
1976 "Remarks on conditions on transformations", Linguistic inquiry 7:265:299.
Bresnan, Joan
1978 "A realistic transformational grammar", Linguistic theory and psychological
reality, edited by Morris Halle, Joan Bresnan and George Miller (Cambridge,
Mass.: M.I.T. Press), 1 - 5 9 .
Chomsky, Noam
1973 "Conditions on transformations", A Festschrift for Morris Halle, edited by
Stephen Anderson and Paul Kiparsky (New York: Holt, Rinehart and
Winston), 2 3 2 - 2 8 6 .
1976 "Conditions on rules of grammar", Linguistic analysis 2:202-352.
1977 "On wft-movement", Formal syntax, edited by Peter Culicover, Thomas
Wasow and Adrian Akmajian (New York: Academic Press), 7 7 - 1 3 2 .
1978 On binding (Cambridge, Mass.: M.I.T. Press).
Culicover, Peter - Thomas Wasow - Adrian Akmajian, eds.
1977 Formal syntax (New York: Academic Press).
250 George . Horn
Fisiak, Jacek - Maria Lipiriska-Grzegorek - Tadeusz Zabrocki

1978 An introductory English-Polish contrastive grammar (Warszawa: Paristwowe
Wydawnictwo Naukowe).
Horn, George M.
1974 The Noun Phrase Constraint (Bloomington, Indiana: Indiana University Lin-
guistics Club).
1979 A lexical-interpretive approach to some problems in syntax (Bloomington,
Ind.: Indiana University Linguistics Club).
1983 Lexical-functional grammar (= Trends in linguistics. Studies and monographs
21) (Berlin: Mouton).
Horn, George M. - Tadeusz Zabrocki
1978 "Island constraints in Polish", Paper presented at the 12th International
English-Polish Contrastive Conference, Tien, Poland.
EWA JAWORSKA
On the structure of adverbial subordinate

constructions in English and Polish*
In this paper I will be concerned with a suggestion made within X-bar theory,
notably by_ Jackendoff (1977) and Emonds (1976), that PP (P') 1 has the
string S as one of its expansions. This structure, they claim, can be as-
signed to adverbial subordinate clauses, with the traditionally understood sub-
ordinating conjunctions generated under the node and the clauses under the
S node.
In what is to follow, I will first survey the analysis at issue, as presented in
Jackendoff 1977 and Emonds 1976. Then, I will discuss a criticism of the ana-
lysis made by Hendrick (1976). Next, I will re-examine some data and tenta-
tively suggest that an alternative approach to English adverbial subordinate
clauses, proposed in passing by Hendrick, should be assumed. Finally, I will
consider some adverbial subordinate clauses in Polish and show that the
Jackendoff-Emonds proposal is not applicable there. I will tentatively con-
clude that a generalization made within X-bar theory, involving the expansion
of PP mentioned above, is questionable as far as English is concerned and, as
far as the Polish facts show, it cannot be claimed to be universal.
1 Theoretical preliminaries
Jackendoff (1973) argued that prepositional phrases have a richer syntax than
had been traditionally assumed. His most recent suggestion (Jackendoff 1977:
81, 82) is that (1.1) is an appropriate rule for expanding PP (P').
(N"')-(P"')
(i.i)P'-P-, (S)
This rule, he claims, generates the PP's in the following sentences:

252 Ewa Jaworska
(1.2) John was standing [ P '[ P outside]]
(1.3) Mary is [p'[pin] [ N -"the house]]
(1.4) Bill ran [p'[pup] [ P " ( t o y o u r bedroom]]
(1.5) Sam sent a letter [ P [ P t o ] [ N "-Bill] [ P " ' i n New York]]
(1.6) Tom left [ P '[ P before] ["the ball was over]]
The rule for P' in (1.1) is rather similar to the rule for V' in (1.7):
(1.7) V' -> V - (N'") - (P'") - ( 5 )
Comparing (1.1) and (1.7), it can be noticed that prepositions and verbs may
be followed by the same constituents in the same relative order. These paral-
lels in the expansions of P' and V' lead Jackendoff to the assumption within
the feature analysis of X-bar theory that both verbs and prepositions have the
anc
same basic set of syntactic features j ^ Q ^ p J * that their phrase structure
rules can be collapsed into one general rule. Thus, in place of (1.1) and (1.7)
Jackendorff (1977:82) introduces (1.8):
(1.8) X
+ Obj X-(N'")I-(P",)a-(S)3
. + Comp.
Condition: If X = P, not (3 and (1 or 2)).
Such cross-categorial generalizations provide evidence for the X-bar theory as-
sumption that syntactic categories should be analyzed as feature complexes. 2
The crucial point about the generalization involving (1.1) and (1.7) that I
will be concerned with here is, as mentioned in the introduction, the occur-
rence of S in the expansion of P' as well as V'. While it is quite uncontroversial
that an S can be a V' complement, the claim that it can also be a P' comple-
ment is a novel development in the treatment of prepositional phrases.
As Jackendoff points out, the idea that he develops goes back to Klima
(1965) who "claims that subordinating conjunctions can also be analyzed as
prepositions which take an S complement [ . . .] This provides the simplest
description of the relations between the prepositions in { ^ e J o r e J the ball

1 after
I before) '
and the 'conjunctions' in j ^ ^ J the ball is over: before and after, like
many verbs, nouns and adjectives allow either an object or a subordinate

clause." (Jackendoff 1977:79)
Adverbial subordinate constructions 253
It follows from this quotation that the occurrence of an S in the expansion

of P' provides for an additional generalization within X-bar theory, namely, that
the four major syntactic categories, V', ', N' and A', are expanded in a sim-
ilar way. This generalization is broader than the other one. It is obvious that
if the narrower generalization does not hold, the broader one does not hold
either. I will argue that the assumption about the expansion of P' as S is
problematic and that it is questionable whether the generalizations mentioned
above can be maintained.
Jackendoff, quoting Klima, illustrates the point with before and after only.
In terms of traditional grammar (e.g. Quirk et al. 1972), they can function as
both prepositions and subordinating conjunctions, depending on whether
they immediately precede a noun phrase or a clause. There are other words,
such as since, until and despite, which can also be used in these two ways.
( 1 . 9 ) - (1.11) illustrate:
(1.9) a. John has been living in Canada since the end of the war.
b. John has been living in Canada since the war ended.
(1.10) a. I will stay here until the beginning of the film,
b. I will stay here until the film begins.
(1.11) a. Peter had a swim despite the storm.

b. Peter had a swim despite that the storm was very heavy.
There is also a class of traditionally understood subordinating conjunctions

which cannot be immediately followed by a noun phrase, and thus do not
function as ordinary prepositions. These are, for example, because, i f , unless
and although in (1.12) - (1.15) . 3
(1.12) a. John will come because you have asked him to.
b. * John will come because you.
(1.13) a. John will come if you ask him to.
b. *John will come if you.
(1.14) a. John will come unless you ask him not to.
b. *John will come unless you.
(1.15) a. John will come although you haven't asked him.
b. *John will come although you.
Because can be immediately followed by the preposition of and a noun phrase.

Thus, parallel to (1.12) b, we have (1.16).
(1.16) John will come because of you.

254 Ewa Jaworska
This is not the case with the other items. Clearly, then, there are at least three
distinct classes of traditionally understood subordinating conjunctions: those
which can be immediately followed by a noun phrase (e.g. before) and thus
function also as prepositions, those which can be followed by a noun phrase
with the preposition of intervening (e.g. because) and those which can never
be immediately followed by a noun phrase (e.g. i f ) . It is the first class of sub-
ordinating conjunctions that has been the main motivation for Klima and,
consequently, for Jackendoff to assign them to the category P. Emonds(1976:
175) explicitly assigns the subordinating conjunctions of the two other classes
to this category too. Within the assumptions presented here, this seems to be
a reasonable thing to do and perhaps Jackendoff would agree on the matter.
If he did not, the claim would not be a very interesting one as it would con-
cern only a small group of traditionally understood subordinating conjunc-
tions which would be analyzed as prepositions and the remainder would be
analyzed as something else. In this way, a possible generalization about sub-
ordinating conjunctions would be missed. It is preferable for a theory to ana-
lyze all subordinating conjunctions in a uniform way. Whether this is a simple
task to perform is not clear. For the purpose of my discussion, I will concen-
trate on the first two classes of subordinating conjunctions mentioned at the
beginning of this paragraph.
While Jackendoff (1977) gives only one reason for regarding adverbial sub-
ordinate constructions as PP's, namely that certain subordinating conjunc-
tions appear elsewhere as ordinary prepositions followed by a noun phrase,
Emonds (1976:173-175) provides a number of arguments for such an ana-
lysis. The first observation he makes is that adverbial subordinate clauses, like
ordinary PP's, can appear in focus position in cleft sentences:
(1.17) It was to John that I spoke.

(1.18) a. It was after the president had finished that the disorder
began.
b. It was because John left that Mary cried.
The second observation is that prepositions expressing spatial or temporal

direction or location can be preceded by the particle right, as illustrated in
(1.19) and (1.20).
(1.19) Mary is planning to go right to London.
(1.20) John arrived right before the last speech.
The particle right can also precede before when it introduces a clause:
(1.21) John arrived right before the last speech began.
Consequently, before in (1.21) should, according to Emonds, be regarded as

the head of a PP.
The third argument for the position that adverbial subordinate construc-
tions are PP's given by Emonds (1976:176178) concerns the operation of a
structure preserving PP lowering rule. This rule relates the following two sen-
tences:
(1.22) a. John believed that he was right until the end of the
lecture.
b. John believed until the end of the lecture that he was
right.
It is clear that in (1.22)b the until phrase has been moved to the position pre-
ceding the embedded clause. Since the embedded clause is a strictly subcate-
gorized argument of the verb believe, it orginates as a daughter of V'. The ad-
verbial PP originates as a sister of V'and so, when it precedes the S in(1.22)b,
there is a clear indication that movement of the PP has taken place. The struc-
ture to which the PP lowering applies is presented in (1.23). 4
V' PP
John V PP S NP
believed that he was right until the end of the lecture
Assuming that adverbial subordinate clauses are also PP's, the relation between
the two sentences in (1.24) can be explained by claiming that the PP lowering
rule has moved the adverbial subordinate construction, a PP, into V' and thus
changed the order of the last two constituents.
(1.24) a. John believed that he was right until the lecture ended,
b. John believed until the lecture ended that he was right.
These three pieces of evidence, then, argue quite strongly that the tradition-
ally understood subordinating conjunctions are heads of prepositional phrases.
It does not, of course, follow that the underlying structure of adverbial sub-
ordinate constructions is S, as Emonds (1976:172) recognizes.
There are, then, two distinct questions to be asked. Firstly: are subordinat-
ing conjunctions in fact prepositions? Secondly: if they are prepositions, are
they followed by S or some other constituent in underlying structure?
256 Ewa Jaworska
2 Hendrick's criticism of the PP analysis of adverbial subordinate constructions
Hendrick(1976) suggests that subordinating conjunctions are not prepositions

but complementizers and hence that adverbial subordinate constructions
should be analyzed as S's and not as PP's. At the end of his critique, he briefly
mentions an alternative possibility that some constructions of this kind should
be analyzed as consisting of a preposition followed by a complex NP, hence
as PP's.
Although it is not my aim here to defend the - S analysis, I will show
that Hendrick's criticisms have no force at all. Moreover, the facts which
Emonds is concerned with argue against his analysis. In the light of Emonds'
arguments, the S analysis seems superior to the proposed COMP - S ana-
lysis. However, there is an objection to the former analysis. This can be avoided
if we take up Hendrick's suggestion to employ the - NP structure for English
adverbial subordinate constructions.
Hendrick presents three arguments against JackendofPs - S structure.
The first one concerns the behaviour of adverbial subordinate clauses with
respect to extraction. The second argument involves a parallel between com-
plementizers and subordinating conjunctions with respect to a rule moving
sentence adverbs. The third argument concerns the distributional fact that
generally no surface complementizer is possible after a subordinating con-
junction.
The first of the above arguments is based on the following assumptions:
1. sisters of V' are syntactic islands; 2. daughters of V' are not syntactic is-
lands; 3. there is a structure preserving PP lowering rule which can move an
adverbial PP into V' with the result that it ceases to be an island. Hendrick
claims that if adverbial subordinate constructions are PP's, originating as sisters
of V', it should be possible for them to be lowered into V' and hence for ex-
traction to be possible too. This prediction, he notices, is false because even
if the PP lowering rule would have operated, extraction from an adverbial sub-
ordinate clause gives an ungrammatical result (Hendrick 1976:117):
(2.1) a. *What is John tired because he went to late?

b. *Who won't we be satisfied until our demands are met
by?
c. *What did John quit school when he realized?5
In contrast extraction from an ordinary PP is possible, due to the restructur-

ing presented in (2.3).
(2.2) Which party did John talk to Sally at?

(2.3) a
talked
to Sally at which party
HTj-movement applies to (2.3)b giving the question in (2.2). The lowering rule
can operate on (2.3)a because there is an available PP node in V'.
Hendrick concludes that adverbial subordinate constructions cannot be
PP's but are S's, sisters of V'. A close look at the data, however, suggests that
the situation is not as straightforward as Hendrick would like it to be.
Let us first consider sentences containing three ordinary prepositional
phrases following the verb. On Hendrick's assumptions, it should be impos-
sible to extract from the third PP. Such a PP will always be a sister of V' as
there are only two PP nodes generated in V'. The sentences in (2.4) and (2.5)
are Hendrick's (1976:117). He marks the questions in b as ungrammatical.
This is somewhat surprising since these questions, as well as the other two, are
grammatical for all my English informants.
(2.4) a. John sent a package to New York by registered mail
for his friend.
b. (*)Who did John send a package to New York by re-
gistered mail for?
258 EwaJaworska
(2.5) a. We talk about mathematics in the reading room on

Tuesdays.
b. (*)What days do we talk about mathematics in the
reading room on?
(2.6) a. John talked to Bill about Mary at Harry's birthday party,
b. Which party did John talk to Bill about Mary at?
(2.7) a. Mike talked to his aunt from France on Monday at the
'Poznaii' Hotel,
b. ?Which hotel did Mike talk to his aunt from France on
Monday at?
If the wh-questions in (2.4)b - (2.7)b, and especially the last one, are dubi-
ous, this is probably due to perceptual factors. (2.7)b contains an extra PP,
from France, embedded in an NP. This extra phrase extends the 'distance' be-
tween the w/z-phrase and the preposition associated with it at the end of the
sentence. A structurally similar but shorter sentence (2.6)b, containing simple
NP's in the prepositional phrases, seems fully acceptable. Since there are only
two PP nodes in V', there is no possibility for a third PP to become a daughter
of V'. Therefore, we must conclude that the third PP remains its sister at the
point at which wft-movement applies.
Another piece of evidence supporting the criticism of this aspect of
Hendrick's account comes from questions like the following:
(2.8) What tone of voice did John say that he had met a spy in?
(2.9) Which day did John hear that Martin owns a Fiat on?
The final prepositional phrases in (2.8) and (2.9) cannot be said to have been
lowered into V' even though there are two available PP nodes for them there;
the base rules devised for V' by Emonds (1976:175) and Jackendoff (1977:
71) require that the that-clause is generated at the end of the string in V', fol-
lowing the PP nodes.6 Therefore, since the in- and o-phrases in (2.8) and
(2.9) respectively follow the embedded clauses, they must be directly domi-
nated by V". Again, then, we are dealing with extraction from a PP that is a
sister of V'.
Facts like the above undermine Hendrick's claim that prepositional phrases
which are sisters of V' are syntactic islands. I will now show that another of
Hendrick's assumptions, that derived daughters of V' are not islands, is
dubious too.
One set of data supporting this criticism involves a contrast between the a
and b questions in (2.10) - (2.12).
(2.10) a. What tone of voice did John say that he had met a
spy in?
b. ?What tone of voice did John say in that he had met
a spy?
(2.11) a. Which day did John hear that Martin owns a Fiat on?
b. ?Which day did John hear on that Martin owns a Fiat?
(2.12) a. Which party did John announce that he was married at?
b. ?Which party did John announce at that he was married?
The questions in (2.10)a and (2.1 l)a are repeated examples (2.8) and (2.9).
We have just seen that these sentences are good, contrary to the predictions
of Hendrick's assumptions. His assumptions also predict that the b examples
in (2.10) - (2.12), involving extraction from a lowered PP, should be gram-
matical. Clearly, this prediction does not hold either. It should be stressed
that while it is difficult to extract from a PP which has been lowered into V',
extraction from a PP which originates in V' as a strictly subcategorized argu-
ment is perfectly natural. (2.13) (2.15) illustrate:
(2.13) Who did John hear from that Mary lives in Sweden?
(2.14) Who did Bill argue with that Carter was insane?
(2.15) Who did Sally shout at that she wasn't going to do the
washing-up?
It seems, then, that Hendrick's first argument against the PP analysis of ad-
verbial subordinate constructions is unacceptable. The assumptions it is based
on and the predictions they lead to are disconfirmed by the facts about ex-
traction from PP's illustrated in (2.4) - (2.15).
It should be noted finally that the formation of questions like the one in
(2.2) above does not necessarily involve the derivation argued for by Hendrick
and presented in (2.3) since the crucial PP does not precede a strictly subcate-
gorized argument.
Hendrick's second argument against the position that adverbial subordinate
constructions are PP's of the structure - S is that there is a parallel between
complementizers and subordinating conjunctions with respect to a rule mov-
ing sentence adverbs. He refers here to a rule called Sentence Adverb Fronting
(SAF), discussed by Wexler and Culicover(1976). This rule is responsible for
relating the following sentences:
(2.16) a. John said that he would arrive tomorrow, hopefully,

b. John said that, hopefully, he would arrive tomorrow.
260 Ewa Jaworska
Hendrick notices that an adverb can occur between that and the following
clause as well as between a subordinating conjunction and the following clause.
(2.17) John will be in Paris next year because, hopefully, he'll get
a Fulbright.
He claims that an adverb cannot immediately precede an S if the complemen-
tizer that is not present on the surface, as illustrated below:
(2.18) *John said, hopefully, he would arrive tomorrow.
Thus, he concludes, if we assume that subordinating conjunctions are com-
plementizers, the position of sentence adverbs in sentences like (2.16)b and
(2.17), and their impossibility in sentences like (2.18) can be explained in a
simple way by saying that they always have to follow an overt complementizer
in a complement clause. Under the S analysis of adverbial subordinate
constructions this statement would have to involve two categories, a comple-
mentizer and a preposition. Therefore, since the former description is less
complex than the latter one, the COMP - S analysis of adverbial subordinate
constructions should be preferred.
It is not clear from what Hendrick says, however, what kind of constraint
he is assuming. One possibility is to say that SAF is an ordinary movement
rule and that it is a constraint on the rule stipulating that it can only apply in
an embedded clause when there is a complementizer. This, however, would
not block the derivation of (2.18). Following Chomsky and Lasnik (1977),
Hendrick would probably assume that movement rules precede deletion rules
and thus that (2.18) had an underlying complementizer which was deleted
after the application of SAF. Hence, the condition stated on the rule itself
would fail to prevent the generation of sentences like (2.18).
Another possibility, then, is to say that the condition is a surface filter.
For Chomsky and Lasnik (1977), surface filters operate on the output of
deletion. A simple filter like (2.19) would rule out sentences like (2.18), as
Hendrick's grammaticality judgements require.
(2.19) *[coMpe] Adv
Yet another possibility is to say that SAF is not an ordinary syntactic but a
stylistic rule. For Chomsky and Lasnik (1977), stylistic rules operate after
deletion and filters. The condition could then be stated on the application of
the rule.
Whichever possibility Hendrick would adopt, the constraint on SAF within
his COMP S analysis would be simpler than within Jackendoff's S ana-
lysis. For Jackendoff, if the constraint had the form of a surface filter, an ad-
tional condition indicating that a sentence adverb may follow either a com-
plementizer or a preposition would be required. The filter would, then, be

something like (2.20):
(2.20) *[coMpe] Adv

unless [coMP e ] preceded by P.
A similar complication would arise for Jackendoff if the constraint imposed
on SAF were regarded not as an ordinary syntactic but as a stylistic rule. In
either case, then, Jackendoff s analysis would necessitate a more complicated
condition.
This would be an advantage for Hendrick's position, only if he were right
about the facts. Crucial for him is sentence (2.18), with no overt complemen-
tizer. According to his grammaticality judgements, this sentence is not pos-
sible. However, all my informants agree that (2.18) as well as (2.21) and(2.22)
below are grammatical.
(2.21) Mary thought, probably, Bill would phone her up.
(2.22) Fred believed, eventually, he'd get a grant.

What may be problematic with these sentences is that out of context they are
a little odd for perceptual reasons: in the absence of the complementizer it is
unclear whether the adverb should be associated with the higher or the lower
clause. (2.23) provides a context due to which the sentence in question is un-
ambiguous.
(2.23) .: What did John say when you last saw him?
B.: He said, hopefully, he'd be back with us next year.
Thus, the rule of SAF does not provide any evidence one way or another as
far as the S analysis of adverbial subordinate constructions is concerned.
This conclusion follows from our re-examination of some crucial data, mis-
interpreted by Hendrick. The oddity of the data can be ascribed not to their
ungrammatically but, apparently, to perceptual factors.
The third and the last of Hendrick's arguments has to do with the fact that,
in general, subordinating conjunctions are not followed by complementizers.
He suggests that the reason why there is no complementizer following a con-
junction is that conjunctions are themselves complementizers. Therefore, we
do not get strings like *because that, *before that, etc. An analysis assuming a
S structure of adverbial subordinate constructions has to account for the
absence of a complementizer after certain subordinating conjunctions. An ac-
count, however, can be easily provided.
Firstly, one might assume with Jackendoff (1977:79) that the comple-
mentizer that in adverbial subordinate clauses is obligatorily deleted in modern
English. 7
262 Ewa Jaworska
Secondly, one might claim that prepositions are followed by a bare S, not
an S. Bresnan (1979) proposes that this is true of certain verbs. Assuming that
prepositions often behave like verbs, subordinating conjunctions, regarded as
prepositions within X-bar theory, could be said to be followed by a bare S in
the underlying structure and thus a complementizer would never appear in
such structures.
It is clear, then, that the absence of the complementizer that in adverbial
subordinate constructions does not constitute a serious problem for the as-
sumption that subordinating conjunctions are prepositions followed by sen-
tences. Furthermore, as Hendrick himself notices, there are some cases where
subordinating conjunctions, which can also function as ordinary prepositions,
are followed by a /-clause:
(2.24) a. John went for a walk despite the rain.

b. John went for a walk despite that it was raining.
(2.25) a. The problem lies in Mary's arrival.
b. The problem lies in that Mary is arriving tomorrow.
In traditional grammar, despite that and in that are regarded as complex sub-
ordinating conjunctions(cf. Quirk et al. 1972). Clearly, they do not constitute
a problem for the Jackendoff analysis. In fact, he does not discuss such cases.
If he did, he would probably modify the claim about obligatory deletion of
that by saying that it is obligatory with some prepositions, like before, and
blocked with others, like despite. Again, this would make the class of pre-
positions look more similar to the class of verbs. Certain verbs, like want,
normally require deletion of a complementizer:
(2.26) a. *John wanted for Mary to go out with him.

b. John wanted Mary to go out with him.
Alternatively, following Bresnan (1979), we could say that, like verbs, pre-
positions can be grouped into those which take a bare S and those which take
an S as their complements.
Hendrick suggests two distinct analyses for despite that and in that. On the
first one, they are complex complementizers. On the second, strings like de-
spite that S and in that S are analyzed as a preposition followed by a complex
NP containing an empty head noun and an S. He suggests that there is no
direct motivation for the empty NP node in English but quotes an example
from Dutch where a pronoun following a subordinating conjunction appears
when the complement clause has been extraposed out of the NP.
It seems to me that the - NP analysis of adverbial subordinate construc-
tions has more to it than Hendrick seems to be aware of. I will tentatively
propose that while the S analysis cannot satisfactorily explain certain

facts about extraction and about the structure of some adverbial subordinate
clauses in English, the - NP analysis can.
Concluding this section, we can say that it is clear that Hendrick's criticisms
of the S analysis of adverbial subordinate constructions cannot be ac-
cepted. This is so because his assumptions about extraction from PP's, sisters
and daughters of V' are untenable; the data he considers are more complex
than his rather idiosyncratic judgements indicate; and, finally, where he is un-
doubtedly right about the facts, they pose no real problem for the proponents
of the - S analysis. Furthermore, Hendrick completely ignores Emonds'
three arguments for the PP analysis: the fact that prepositional phrases, not
sentences (Emonds 1976:133),can occupy the focus position in cleft sentences;
the occurrence of the intensifer right before prepositions; and the fact that ad-
verbial subordinate constructions, generated as sisters of V'by phrase structure
rules, can, in some circumstances, appear as daughters of V'. If these construc-
tions were S's and not PP's, there would be no simple explanation of these facts.
3 Extraction and the structure of adverbial subordinate constructions
Facts about extraction become relevant when the structure of a construction

is to be determined. As we have seen, Hendrick's attempt to use this kind of
evidence to explain the islandhood of adverbial subordinate clauses fails
through his inadequate attention to data. For him, extraction from sisters of
V' is not possible. Adverbial subordinate clauses generated as S's directly un-
der V" are always islands since they cannot undergo Emonds' lowering rule.
We have seen, however, that Hendrick's general assumption about the island-
hood of sisters of V' is wrong.
Given Jackendoffs analysis of adverbial subordinate constructions, the
only principle that might be invoked to block sentences like (3.1) is Choms-
ky's subjacency condition. 8 In its most recent formulation (Chomsky 1980),
the condition states that certain rules, in particular movement rules, cannot
operate across the boundary of more than one bounding node.
(3.1) * What did Bill arrive before started?
In (3.2), successive cyclic application of w/z-movement, in accordance with
Chomsky's assumptions about the operation of the rule, is presented.
(3.2) [g what [ s did Bill arrive [pp[pbefore][s e[g e started]]]]]
The implications of subjacency are dependent on the choice of the bounding
nodes. On the second application of wA-movement in (3.2), the w/i-word
264 Ewa Jaworska
crosses three main boundaries. At least two of them must be bounding nodes
in order for subjacency to account for (3.1). Baltin (1978) argues that PP, NP
and S are all bounding nodes. Clearly, the wh-word in (3.2) crosses two of
them, S and PP on the second cycle. This, then, is how the islandhood of ad-
verbial subordinate clauses might be explained. However, this combination of
bounding nodes is problematic. Although it can correctly block sentences like
(3.3)a in the way presented schematically in (3.3)b, it cannot block sentences
like (3.4)a, noticed by George Horn (personal communication), as illustrated
in (3.4)b.
(3.3) a. *Who did you destroy a book about?

b. [ who [sdid you destroy [upa book [ P P about e]]]]
(3.4) a. * About who did you destroy a book?

b. [about who [sdid you destroy [ N P a book e]]]
While in (3.3) the w/j-word crosses two bounding nodes, NP and PP, on its
way to the front of the sentence, the wh -phrase in (3.4) crosses only one
bounding node, NP. To block (3.4)a, S and NP must be bounding nodes.
However, if PP is also a bounding node, as it is for Baltin, (3.5)a is blocked, as
illustrated in (3.5)b.
(3.5) a. Who did you talk to?

b. [s who [ s d i d y o u talk [ppto e]]]
If PP were not a bounding node and S and NP were, sentences like (3.3)a,
(3.4)a and (3.5)a would be correctly predicted by the subjacency condition
on w/i-movement. With S and NP as bounding nodes, however, subjacency
would not be able to explain the islandhood of adverbial subordinate clauses
exemplified in (3.1) and (3.2).
The above considerations suggest that subjacency cannot account for the
islandhood of adverbial subordinate clauses. 9
In fact, it is worth noting that Hendrick cannot invoke subjacency either.
Consider the following:
(3.6) a. Bill arrived [ [coMpbefore] [ s John hit Mary]]

b. *Who did Bill arrive before John hit?
c . [ s w h o [ s did Bill arrive [ [ c o M P e before] [ s John hit e]]]]
(3.6)a represents the structure which Hendrick assumes for adverbial subordi-
nate constructions. Unless S and S are both bounding nodes, there is no way
to prevent the derivation of (3.6)b. 1 0 Obviously, these two nodes cannot be
both bounding nodes, as it would never be possible to generate sentences like
(3.7)a.
(3.7) a. Who did you think that John hit?

b. [g who [ s d i d y o u think [s[ C O MP e that] [ s John hit e]]]]
We can conclude, then, that there is no obvious way on Jackendoff's assump-
tions on the one hand and Hendrick's assumptions on the other to account
for the islandhood of adverbial subordinate clauses. This casts doubt on both
these analyses.
A possible account of the islandhood of adverbial subordinate clauses is
suggested by Hendrick's brief proposal that certain constructions of this type
could be analyzed as a preposition followed by a complex noun phrase. (3.8)
illustrates the idea. 11
(3.8) ^ ^ P P ^ ^
NP
COMP S
that
For Hendrick, this structure is relevant only in the cases of adverbial subordi-
nate clauses where that is overtly present. I would like to propose, rather ten-
tatively, that all English adverbial subordinate clauses involve a complex NP
in the underlying structure. This proposal would treat all adverbial subordi-
nate constructions as PP's and would thus be compatible with the three argu-
ments given by Emonds. It would also, however, affect Jackendoff's generali-
zation concerning verbs and prepositions in that it would be only verbs but
not prepositions which can be followed by sentential complements. Thus, sen-
tences like (3.9)a and (3.10)a would, respectively, have structures presented
schematically in (3.9)b and (3.10)b.
(3.9) a. John went out despite that it was raining.

b. John went out [pp [pdespite] [^p [NP e ] [s i* was
raining]]]
(3.10) a. John went out before it started to rain.
b. John went out[ P P [pbefore][ N p[ N pe][gitstarted to
rain]]]
The advantage of assuming this analysis of adverbial subordinate constructions
rests in the fact that extraction from these clauses can be regarded as a viola-
tion of the complex NP constraint first formulated by Ross (1967:70). The
266 Ewa Jaworska
exact nature of the constraint is not important here: it can be regarded either
as a consequence of Chomsky's subjacency condition or of Horn's (1974,
1979) NP constraint. What is important here is that complex NP's are syntac-
tic islands and thus that the extraction facts can be explained if adverbial sub-
ordinate constructions involve a complex NP.
As mentioned earlier, Hendrick sees no real motivation for the empty head
of the complex NP in English. It seems, however, that some motivation can be
provided. Consider the following pairs of sentences:
(3.11) a. John arrived on time despite that the train was late,
b. John arrived on time despite the fact that the train was
late.
(3.12) a. John didn't come because his car had broken down,
b. John didn't come because of the fact that his car had
broken down.
(3.13) a. Bill arrived before John hit Mary.

b. Bill arrived before the time at which John hit Mary.
Sentences (3.1 l)a(3.13)a contain the controversial adverbial subordinate

constructions. They are synonymous with the b sentences of (3.11)(3.13).
The latter can plausibly be assigned the structure in (3.8) with a non-empty
head NP. As regards the a sentences, they can be derived from an underlying
structure identical to that of the b sentences via deletion of the lexical heads
the fact or the time.12 The deletion of the head noun in (3.11)(3.13) is
optional. Alternatively, one might claim that the underlying structure of the
a sentences differs only slightly from that of the b sentences in that the head
NP in the former is an empty node. No deletion would then be necessary to
derive (3.1 l)a-(3.13)a.
If the above proposal for an analysis of adverbial subordinate structures in
English is correct, Emonds' and Jackendoffs claims about the similarity be-
tween verbs and prepositions in that they both take sentential complements
are not true for English. The - NP analysis has the advantage over the
S analysis in that it explains why extraction from adverbial subordinate
clauses is not possible. Furthermore, it offers a way to account for the syno-
nymy of certain types of sentence pairs, like those in (3.11)(3.13). Whether
these two arguments are strong enough to make us definitely reject the other
analyses of adverbial subordinate constructions in English is not entirely clear.
Tentatively, however, I will assume that (3.8) is an appropriate structure for
at least some English constructions of this kind.
4 Adverbial subordinate constructions in Polish
As we have seen, the initial motivation for regarding the English subordinating
conjunctions as prepositions is the fact that they can be followed by a sen-
tence as well as by a noun phrase:
(4.1) a. John arrived before midnight,

b. John arrived after midnight.
(4.2) a. John arrived before Jack left,
b. John arrived after Jack left.
As is illustrated below, the equivalent Polish prepositions przed 'before' and

po 'after' can only be followed by a noun phrase, not by a sentence.
(4.3) a. Jan przyjechai przedMariq.

John arrived before Mary(ins)
'John arrived before Mary.'
b. Jan przyjechai po Marii.

John arrived after Mary(loc)
'John arrived after Mary.'
(4.4) a. *Jan przyjechai przed Jerzy wyjechal.

John arrived before George left
'John arrived before George left.'
b. *Jan przyjechai po Jerzy wyjechal.
John arrived after George left
'John arrived after George left.'
(4.5) a. Jan przyjechai przed wyjazdem Jerzego.
John arrived before departure(ins) George's
'John arrived before George's departure.'
b. Jan przyjechai po wyjezdzie Jerzego.
John arrived after departure(loc) George's
'John arrived after George's departure.'
(4.6) a. Jan przyjechai przed tym, jak Jerzy wyjechal.
John arrived before this(ins) how George left
'John arrived before George left.'
b. Jan przyjechai po tym, jak Jerzy wyjechal.
John arrived after this(loc) how George left
'John arrived after George left.'
268 Ewa Jaworska
As is evident from (4.3), przed takes an object NP in the instrumental case

and po takes an NP in the locative case. If a subordinate clause immediately
follows either of the prepositions, as in (4.4), the whole sentence is ungram-
matical. The adverbial subordinate clause has to be introduced by the demon-
strative pronoun ten 'this' (masc.) in the appropriate case, 13 followed by the
complementizer jak 'how', 14 as in (4.6). It seems plausible, then, that (4.6)
should be assigned an underlying structure like (4.7), along the lines suggested
by Hendrick (1976) and presented in (3.8).
(4.7)
Jan przyjechat
jak Jerzy wyjechat

Similar to przed and po is mimo 'despite'. On the surface, mimo can be fol-
lowed either by a simple or a complex NP:
(4.8) Jan byi w dobrym humorze mimo uwagAnny.
John was in good humour despite remarks(gen) Ann's
'John was in a good mood despite Ann's remarks.'
(4.9) Piotr poszedt na spacer mimo tego, ze nie miai parasola.
Peter went on walk despite this(gen) that not (he)had
umbrella
'Peter went for a walk despite that he hadn't got an um-
brella.'
Mimo takes an NP in the genitive case. 15 Again, we notice that when it is fol-
lowed by an adverbial clause, the clause is preceded by a demonstrative. Un-
like with przed and po, where the demonstrative is obligatory, the demon-
strative after mimo is optional. (4.10) and (4.11) illustrate.
(4.10) *Jan przyjechat | ^ ^ J jak Jerzy wyjechat.
(4.11) Piotr poszedt na spacer mimo, ze nie miai parasola.

Therefore, we can say that mimo is quite like the English despite in that it is
optionally followed by an NP preceding an embedded clause. In the case of
mimo, the NP is a form of a demonstrative pronoun; in the case of despite it
is the fact. The non-occurrence of the demonstrative in Polish, like of the fact
in English, can be explained in two ways: either the demonstrative head of
the complex NP is present in the underlying structure and then optionally
deleted or the head is generated as an empty node in some circumstances.
Which analysis is correct is not important here. The important thing is that
with prepositions like przed, po and mimo followed by adverbial subordinate
clauses the underlying structure is clearly - NP and not - S.
Another set of Polish adverbial subordinate constructions is presented in
(4.12M4.14).
(4.12) Jan bdzie w Paryzu w przyszlym roku dlatego, ze
dostanie stypendium.
John will be in Paris in next year for-this that (he) will get
grant.
'John will be in Paris next year because hell get a grant.'
(4.13) Piotr poznat Ann przedtem, jak kupil samochod.
Peter met Ann before-this how (he) bought car.
'Peter met Ann before he bought the car.'
(4.14) Piotr poznat Ann potem, jak kupit samochod.
Peter met Ann after-this how (he)bought car
'Peter met Ann after he bought the car.'
Extending Jackendoff s analysis, dlatego, przedtem and potem might be as-
signed to the category P, followed by ze- and /afc-clauses within a PP. How-
ever, dlatego, przedtem and potem can never be followed by a noun phrase.
Therefore, treating them as prepositions in the above sentences would be
rather dubious.
Following Hendrick's proposal of the COMP - S analysis of adverbial sub-
ordinate constructions, one might claim that dlatego ze, przedtem jak and
potem jak are complex complementizers. We have seen, however, that the idea
of a complex complementizer is not very plausible for English. In Polish,
there are two more plausible analyses for the adverbial subordinate construc-
tions in (4.12)-(4.14).
The first possible analysis is based on the observation that dlatego, przedtem
and potem are each composed of two elements: a preposition, dla 'for', przed
'before' and po 'after', and a form of the demonstrative pronoun ten. In the
case of dlatego, ten appears in its genitive form tego. This is not surprising
since dla is normally followed by a noun phrase in the genitive case:
270 Ewa Jaworska
(4.15) Herbata jest dla Aliny.

tea is for Alina(gen)
'The tea is for Alina.'
In the cases of prezdtem and potem, ten appears in a phonologically modified

form of tym. We have seen that przed takes a noun phrase in the instrumental
case and po takes a noun phrase in the locative case.
The ze following dlatego in (4.12) is clearly a complementizer, a tradition-
ally understood subordinating conjunction, as it introduces subordinate clauses.
This is shown by (4.9) and sentences like the following:
(4.16) Jan powiedziat, ze przyjdzie.

John said that (h.e)will come
'John said that he'd come.'
The jak following przedtem and potem in (4.13) and (4.14) is also a comple-
mentizer (cf. note 14).
Thus, using evidence from case marking of ten, we could say that dlatego
ze S, przedtem jak S and potem jak S derive from a PP containing the prepo-
sition dla, przed and po, respectively, followed by a complex NP with ten as
the head noun. This analysis is essentially similar to the one proposed for sen-
tences like (4.6). (4.17) illustrates the underlying structure of (4.12).
(4.17) S
NP V"
V' PP PP
Jan b$dzie w Paryzu w przyszfym

/ \ NP
roku
dla S
ten COMP S
ze Jan dostanie
stypendium
After ten in (4.17) has been assigned the genitive case marking, the preceding
preposition is adjoined to it, which results in dlatego. This readjustment opera-
tion is independently motivated by w/i-questions like (4.18).
(4.18) ktrym Maria rozmawiaia mfzczyzrtq?

with which Mary talked man
'Which man did Mary talk to?'
Assuming that (4.18) derives directly from an underlying structure like(4.19),
we would have to say that a non-constituent ktrym is fronted by wh -move-
ment.
(4.19)
Maria rozmawiaia
ktrym m^zczyznq
Since it is generally accepted that only constituents can be moved (Schwartz

1972), the preposition 'with' in (4.19) must first form a constituent with
ktrym 'which'. The result of the operation of this readjustment rule is some-
thing like (4.20).
(4.20)
Maria rozmawiaia
Det
Det mqzczyznq
ktrym
Thus, to claim that dlatego derives from dla and tego as well as przedtem and
pot em from przed and tym, and po and tym, respectively, does not seem un-
reasonable.
272 Ewa Jaworska
Under the second possible analysis of the adverbial subordinate clauses in

(4.12>(4.14), dlatego, przedtem and potem are simply adverbs and the ad-
verbial subordinate constructions there are complex adverbial phrases. (4.21)
illustrates the idea for (4.12).
(4.21)
Jan bfdzie w Paryzu w przysziymroku AdvP ^ S ^
dlatego COMP S
ze Jan dostanie
stypendium
Shortly, I will argue that this is preferable to the one illustrated in (4.17).
A rather different type of adverbial subordinate clause is illustrated in the
following:
(4.22) Jan nie przyjdzie, bo ztamai nog.
John not will come because (he)broke leg
'John won't come because he's broken a leg.'
(4.23) Maria by la tu, zanim poznala Paw la.
Mary was here before (she)met Paul
'Mary was here before she met Paul.'
(4.24) Przetlumacz ten tekst, odkqd zaznaczylem.
(you)translate this text from-where (I)marked
'Translate this text from where I've made a mark.'
(4.25) Przeczytaj ten list, dokqd cipokazalem.
(you)read this letter to-where you (I)showed
'Read this letter up to where I've shown you.'
(4.26) Piotr zadzwoni, skqd bgdzie mgt.
Peter will phone from-where (he)will be can
'Peter will phone from where he'll be able to.'
Bo 'because', zanim 'before', odkqd 'from where', dokqd 'to where' and skqd
'from where' function here as subordinating conjunctions. They can never,
however, be followed by an NP like ordinary prepositions and therefore it
would be quite strange to claim that (4.22)-(4.26) conform to Jackendoff's

analysis of adverbial subordinate constructions.
There is strong evidence, on the other hand, that these subordinating con-
junctions are in COMP. While bo and zanim in (4.22) and (4.23) introduce
only adverbial subordinate clauses, odkqd, dokqd and skqd appear in wh-ques-
tions, as illustrated in (4.27)-(4.29).
(4.27) Odkqd mam przetlumaczyc ten tekst?
from-where (I)have translate this text
'Where do I have to translate this text from?'
(4.28) Dokqd mog$ przeczytac ten list?
to-where (I)can read this letter
'How far can I read this letter?'
(4.29) SkqdPiotr zadzwoni?
from-where Peter will phone
'Where will Peter phone from?'
It seems reasonable, then, to assume that the words in question in (4.24)
(4.26) are wA-words. In fact, they appear to be wh -words in COMP in free
relative clauses.16 Notice that they can appear with the suffix -kolwiek, which
is characteristic of free relatives.
(4.30) a. Przetlumacz ten tekst odkqdkolwiek chcesz.
(you)translate this text from-wherever (you)want
'Translate this text from wherever you like.'
b. Przeczytaj ten list dokqdkolwiek chcesz.
(you)read this letter to-wherever (you)want
'Read this letter to wherever you like.'
c. Piotr zadzwoni skqdkolwiek bgdzie mgt.
Peter will phone from-wherever (he)will be can
'Peter will phone from wherever he'll be able to.'
There are three further arguments in support of this analysis of(4.22)-(4.26).
Firstly, there is a general agreement that all subordinate clauses in Polish, un-
like in English, must have a non-empty COMP in surface structure. (4.31)
(4.34) illustrate the point.
(4.31) John married the girl he had met on the train.
(4.32) a. Jan poslubil dziewczynf, ktorq poznai wpociqgu.
John married girl which (he)met in train
'John married the girl who he had met on the train.'
b. *Jan poslubil dziewczyn, poznai w pociqgu.
274 Ewa Jaworska
(4.33) John thinks he is clever.

(4.34) a. Jan myi, ze jest zdolny.
John thinks that (he)is clever
'John thinks that he is clever.'
b. *Jan mysli, jest zdolny.
We can conclude, then, that the subordinating conjunctions in (4.22)(4.26)
can be regarded as occupying the COMP position since there is no other ob-
vious candidate for this position in these sentences.
Secondly, the first and second person past tense inflections can be moved
away from the verb and attached to the words under consideration just as
they can to other, more obvious complementizers and w/j-words in COMP.
(4.35)(4.38) illustrate.
(4.35) a. Jan wiedziat, ze ty przysz edles.
John knew that you came
'John knew that you'd come.'
b. Jan wiedziat, zes ty przyszedl.
(4.36) a. Kiedy widzieliscie Jerzego?

when (you/pl.)saw George
'When did you see George?'
b. Kiedyscie widzieli Jerzego?
(4.37) a. Maria byla tu, zanim jq poznales.

Mary was here before her (you)met
'Mary was here before you met her.'
b. Maria byta tu, zanims jqpoznal.
(4.38) a. Przeczytalem ten list, dokqd mi pokazales.
(I)read this letter to-where me (you)showed
read the letter up to where you'd shown me.'
b. Przeczytalem ten list, dokqds mipokazal.
As Bob Borsley has pointed out (personal communication), mobile inflections
cannot appear outside the first S that dominates the verb with which they are
associated. Thus, for example, an inflection can be attached to a relative pro-
noun in COMP but not to the head noun, which is outside the S. (4.39) il-
lustrates.
(4.39) a. Mfzczyzna, ktorego widziales, wyjechal do Francji.
man which (you)saw went to France
'The man you saw has gone to France.'
b. M^zczyzna, ktoregos widzial, wyjechat do Francji.
c. *M$zczyznas, ktorego widzial, wyjechal do Francji.
This fact argues for an analysis of subordinating conjunctions under which

they are within an S, i.e. in COMP. If subordinating conjunctions were treated
as prepositions, mobile inflections would have to be said to be crossing a
clause boundary when they appear with bo, zanim, etc. Moreover, another
restriction on the movement of mobile inflections would be violated, namely,
that they cannot be attached to prepositons. (4.40) illustrates.
(4.40) a. Do Paryza pojechales.
to Paris (you) went
'You went to Paris.'
b. Do Paryzas pojechat.
c. *Dos Paryza pojechat.
Thus, the occurence of mobile inflections with subordinating conjunctions

provides evidence that they should be treated as complementizers rather than
prepositions.
We can also use the facts about mobile inflections to justify the rejection
of the idea of a complex COMP for dlatego ze, przedtem jak and potem jak,
suggested in connection with sentences (4.12)-(4.14). The inflections can
only be attached to the second element, not to the first.
(4.41) a. Bytes tarn dlatego, zes dostai stypendium.
(you)were there for-this that(you) got grant
'You were there because you'd got a grant.'
b. *Byles tarn dlategos, ze dostai stypendium.
(4.42) a. Poznales przedtem, jaks kupit samochod.17
(you)met Ann before-this how(you) bought car
'You met Ann before you bought the car.'
b. *Poznales Ann$ przedtems, jak kupit samochod.
(4.43) a. Poznales Ann potem, jaks kupit samochod.
(you)met Ann after-this how(you) bought car
'You met Ann after you bought the car.'
b. *Poznales potems, jak kupit samochod.
The third argument in favour of regarding the subordinating conjunctions in
(4.22)-(4.26) as occupying the COMP position is that they can be preceded
by various adverbs. Parallel to (4.22)-(4.26), we have (4.44)-(4.48).
(4.44) Jan nie przyjdzie dlatego, bo lama I nog$.
John not will come for-this because(he)broke leg
(4.45) Maria byla tu przedtem, zanim poznala Pawla.
Mary was here before-this before (she)met Paul
276 Ewa Jawoiska
(4.46) Przetiumacz ten tekst odtqd, odkqd zaznaczylem.

(you)translate this text from-there from-where (I)marked
(4.47) Przeczytaj ten list dotqd, dokqd cipokazatem.
(you)read this letter to-there to-where you (I)showed
(4.48) Piotr zadzwoni stqd, skqd bdzie mgt.
Peter will phone from-there from-where (he)will be can
The above sentences show a great similarity to those in (4.12)(4.14). There-
fore, they should all be analyzed in a uniform fashion.
We can now return to the question of whether the analysis in (4.17) or
(4.21) is appropriate for certain adverbial subordinate constructions in Polish.
If we tried to extend the analysis in (4.17) to odtqd and stqd 'from there', and
dotqd 'to there' we notice that while -tego of dlatego and -tern of przedtem
and potem can appear on their own elsewhere, -tqd of odtqd, stqd and dotqd
cannot. Therefore, it seems preferable to postulate that odtqd, stqd and dotqd,
and, consequently, dlatego, przedtem and potem are not derived forms. Thus,
the analysis in (4.21) is preferable for both (4.12)(4.14) and (4.44)-(4.48).
This analysis provides a uniform account for two types of adverbial subordi-
nate clauses: those introduced by an ordinary complementizer and those in-
troduced by a w/i-word in COMP.
It is easy to see that the adverbial heads in (4.44)-(4.48) are optional.
Potem in (4.14) is also optional, as (4.49) shows.
(4.49) Piotr poznat Anne, jak kupil samochod.
Dlatego in (4.12), unlike in (4.22), is obligatory, as (4.50) illustrates.
(4.50) *Jan b$dzie w Paryzu w przyszfym roku, ze dostanie
stypendium.
Przedtem in (4.13) is obligatory, as its absence results in the reading under
which the order of the events described in the original sentence is reversed.
Przedtem is optional, however, if the complementizer zanim, not jak, intro-
duces the embedded clause.
Generalizing the description, we can say that the adverbial head is obliga-
tory unless it precedes some specialized complementizer or w/i-word associ-
ated only with this adverb. If the complementizer is ze or jak, associated with
dlatego and potem, respectively, in adverbial subordinate constructions but
also appearing elsewhere introducing other kinds of subordinate clause, the
adverb is obligatory. Whether, in the cases where the adverb is optional, the
AdvP node always dominates some lexical material in the underlying structure
which can later be deleted or whether it is sometimes generated as an empty
node is not important for the present discussion.
Conclusion
In the above discussion I have been concerned with adverbial subordinate con-
structions. According to Jackendoff (1977) and Emonds (1976), the subor-
dinating conjunction belongs to the category and is followed by an S in the
underlying structure. This assumption makes the base rule of a PP parallel to
the base rule of a VP, which provides some motivation for X-bar theory: verbs
and prepositions can be referred to in terms of the same basic features [+Obj]
and [+Comp], and this kind of generalization across various syntactic cate-
gories is exactly what the theory is supposed to capture. If an S were not a
possible complement of a PP, the generalization about the two categories and
the subsequent simplification would be precluded. Consequently, a broader
generalization within which all the four major syntactic categories, NP, VP,
PP and AP, can take a final S in their expansions, would not be possible. The
initial motivation for treating subordinating conjunctions as prepositions was
that some of them, like before and after, are ordinary prepositions elsewhere.
Hendrick (1976) tries to show that, for various reasons, an adverbial sub-
ordinate construction cannot be treated as a preposition followed by an S.
His main alternative to the S analysis is that the subordinating conjunc-
tion is a complementizer, followed by an S. As I have shown, his arguments
against Jackendoff (and Emonds) and for his own position are untenable. He
makes an important observation, however, that extraction from adverbial sub-
ordinate clauses, unlike from other types of subordinate clauses, is impossible.
As we have seen, there is no obvious way to account for this with either a
- S or COMP - S analysis. In this respect, the COMP S analysis is as de-
fective as the - S analysis.
A re-examination of the English data within yet another analysis, men-
tioned but not explored by Hendrick, has lead us to an explanation of the
islandhood of certain adverbial subordinate clauses. This analysis treats an
adverbial subordinate construction as a PP consisting of a preposition (the
subordinating conjunction) followed by a complex NP. As is generally ac-
cepted, complex NP's are islands. A tentative claim is, then, that adverbial sub-
ordinate constructions are, indeed, prepositional phrases, which is consistent
with Emonds' arguments and Jackendoff s assumptions, but that their internal
structure is - NP, which is inconsistent with the X-bar theory claim. Thus,
the facts about extraction seem to argue against the cross-categorial generali-
zation and remove one piece of motivation for X-bar theory.
Certainly, more data have to be considered, especially sentences containing
subordinating conjunctions like i f , although, etc., which I have neglected in
the present paper. They seem to constitute a problem for the analysis advanced
here. Regarded as prepositions, such conjunctions could not be subcategorized
278 Ewa Jaworska
just for an NP because of the impossibility of, for example, *although John.
They would have to be subcategorized for a specific type of an NP, namely,
a complex NP with an empty head. Obviously, this is not a desirable solution.
It may be the case, then, that they should, indeed, be regarded as comple-
mentizers. Evidence more plausible than that presented by Hendrick would
have to be provided. It also remains to be seen whether the argument against
the - S analysis of English adverbial subordinate constructions, favouring
the - NP analysis, is the only one available and whether it is strong enough
to constitute a serious problem for X-bar theory where it makes a generaliza-
tion concerning verbs and prepositions.
Polish adverbial subordinate constructions do not provide any motivation
at all for the analysis advocated by Jackendoff. No ordinary preposition in
Polish can be immediately followed by a clause. Prepositions in Polish always
have to be followed by an NP, either simple or complex. Unlike irx English,
then, in Polish, the traditionally understood subordinating conjunctions and
prepositions are two distinct classes of words, not overlapping with each other.
As far as Polish subordinating conjunctions are concerned, good reasons have
been given above for regarding them as complementizers in an S embedded in
a complex NP or AdvP. Jackendoff s analysis of adverbial subordinate struc-
tures, even if it is adequate for English, is not applicable in Polish and cannot
be claimed to be universal. A further examination of Polish data should reveal
if it is true of all Polish subordinating conjunctions that they are comple-
mentizers.
Finally, it must be noted that other cross-categorial generalizations should
be subjected to similar scrutiny. The current state of X-bar theory requires
that more elementary research is done before the most basic assumptions can
be regarded as securely established. Along with English, other languages should
be investigated to secure the viability of X-bar theory generalizations.
Notes
* I would like to express my sincere thanks to Bob Borsley, who encouraged me to take
up the topic and was patient enough to give me a lot of guidance and enlightening com-
ments as the work on the paper progressed. He has also provided advice on the English.
He is not responsible for the faults of this paper. I am also grateful to my informants
for the English data, especially Karelia Trabold-Szkoda and Dick Weist.
1 . 1 will use the X-bar notation where it is necessary for the clarity of the presentation
of the material. Otherwise, I will use the traditional notation.
2. The apparently simple condition in (1.8) handles the fact that, unlike with verbs,
with prepositions, there cannot be any constituent beween the head and the com-
plement sentence. Jackendoff (1977:82) admits that in the cases of collapsing other
categories, where their individual expansions differ to a larger extent, the conditions
stated in the negative form may become so elaborate that the generalizations intended
to be captured by the phrase structure rules may look implausible.
3, Hendrick (1976:117), in his discussion of Jackendoff 1977, which I will consider in

the next section, implies that when too would be dominated by in Jackendoff's
framework. This, however, is not Jackendoff's position. He explicitly says (1977:75)
that wAew-clauses are S's generated immediately under V".
4, For simplicity, I will include lexical items in their surface forms.
5. As noted earlier (see note 3), Hendrick takes when to be dominated by in Jacken-
doff's approach.
6 . Unlike Jackendoff, Emonds does not give a full expansion of V' (his VP). The part of
the phrase structure rule for V' he provides, however, is sufficient to establish the
order of the final elements in the phrase:
(i) VP - . . . V . . . (PP) (S)
The final part of Jackendoff's expansion of V' allows the following possibilities:
(ii) V ' - . . . V . . . ( P P ( jP^}))
7, Chomsky and Lasnik (1977) would say that the that is deleted by a rule of free
deletion in COMP. This rule has been criticized by Pullum and Postal (1979).
8, The arguments outlined here owe much to discussion with Bob Borsley.
9, Subjacency is a rather dubious constraint anyway. For critical discussion see Bresnan
1976, Maling 1978, Allen 1980, Borsley 1981 and Horn - Borsley (1981).
10 Actually, the derivation of (3.6)b might be blocked by the opacity condition of
Chomsky (1980). The condition says that if a is in the domain of the subject of
( = S or NP) then cannot be free in . In (3.6)b, the trace in the original position
is in the domain of the subject of the embedded clause and, thus, should not be free
within the lower S. It is, however, since the trace in COMP does not c-command it.
It seems dubious, on the other hand, whether the opacity condition is a viable means
of blocking (3.6)b since, as Bob Borsley pointed out to me, it should also block good
sentences like (3.7)a.
11. (3.8) is a slightly modified version of the diagram given by Hendrick (1976:119).
12, The deletion of the head NP in (3.1 l)b and (3.12)b would trigger the deletion of the
complementizer. It is not quite clear how exactly to handle the of in (3.12).
Geis (1970) suggests that sentences like (3.13)b constitute an intermediate struc-
ture from which (3.13)a is derived through deletion of the time at which.
13. The instrumental and locative forms of ten are both tym:
(i) Jan rozmawiat tym ioinierzem.
John talked with this soldier
(ii) Jan rozmawiat tym zotnierzu.
John talked about this soldier
14. See Borsley 1981 for arguments that in certain circumstances jak should be regarded
as a complementizer rather than as a vWi-word.
15, Mimo takes also noun phrases in the accusative case:
(i) a. mimo wszystko
despite everything(acc)
b. *mimo wszystkiego(gen)
The demonstrative pronoun ten following mimo as the head of a complex NP may
appear either in the accusative or in the genitive form (cf., for example, Urbaiiczak
1966:274 -275 and Pisarek 1978:56).
16, For discussion of Polish free relatives and evidence that the -words are ordinary
wft-words in COMP see Borsley (this volume).
17. Speakers vary in the realization of this phenomenon with jak: jakzei and jakei are
the alternatives. Jaks has been chosen here for the sake of simplicity.
280 Ewa Jaworska
References
Allen, C.
1980 "Movement and deletion in Old English", Linguistic inquiry 11:261-323.
Baltin, .
1978 " as a bounding node", Proceedings from the Eighth Annual Meeting of
the North Eastern Linguistic Society, edited by M. J. Stein (Amherst, Mass.:
University of Massachusetts), 3 3 - 4 0 .
Borsley, R. D.
this volume "Free relatives in Polish and English".
1981 "WA-movement and unbounded deletion in Polish equatives", Journal of
linguistics 1 7 : 2 7 1 - 2 8 8 .
Bresnan, J. W.
1976 "Evidence for a theory of unbounded transformations", Linguistic analysis
2:353-393.
1979 Theory of complementation in English syntax (New York: Garland).
Chomsky, N.
1980 "On binding", Linguistic inquiry 1 1 : 1 - 4 6 .
Chomsky, . - H. Lasnik
1977 "Filters and control", Linguistic inquiry 8 : 4 2 5 - 5 0 4 .
Emonds, J. E.
1976 A transformational approach to English syntax (New York: Academic Press).
Geis, . L.
1970 "Time prepositions as underlying verbs", Papers from the Sixth Regional
Meeting of the Chicago Linguistic Society, edited by M.A.Campbell,
J. Lindholm, A. Davison et al. (Chicago, Illinois: University of Chicago),
235-249.
Hendrick, R.
1976 "Prepositions and the X' theory", UCLA papers in syntax 7 : 9 5 - 1 2 2 .
Horn, G. M.
1974 The noun phrase constraint (Ph. D. dissertation, University of Massachusetts,
Amherst). Reproduced by the Indiana University Linguistics Club, 1977.
1979 A lexical interpretive approach to some problems in syntax (Bloomington:
Horn, G. M. - R. D. Borsley
1981 "On O n binding'", Linguistics 1 9 - 1 1 / 1 2 : 1133-1164.
Jackendoff, R. S.
1973 "The base rules for prepositional phrases", A Festschrift for Morris Halle,
edited by S. Anderson and P. Kiparsky (New York: Holt, Rinehart and
Winston), 3 4 5 - 3 5 6 .
1977 X syntax: A study of phrase structure (= Linguistic inquiry monograph, 2)
(Cambridge, Mass.: M.I.T. Press).
Klima, . S.
1965 Studies in diachronic syntax (Ph. D. dissertation, Harvard University).
Maling, J.
1978 "An asymmetry with respect to w/i-islands", Linguistic inquiry 9 : 7 5 - 8 9 .
Pisarek, W.
1978 Stownik fezyka niby-polskiego, czyli bt^dy fezykowe w prasie [A dictionary
of semi-Polish or linguistic errors in the press] (Wroclaw: Zak tad Narodowy
im. Ossoliriskich).
Pullum, G. . - P. M. Postal
1979 "On an inadequate defense of 'trace t h e o r y ' " , Linguistic inquiry 10:689
706.
Quirk, R. - S. Greenbaum - G. Leech - J. Svartvik

1972 A grammar of contemporary English (London: Longman).
Ross, J. R.
1967 Constraints on variables in syntax (Ph. D. dissertation, .I.T.). Reproduced
by the Indiana University Linguistics Club, 1968.
Schwartz, A.
1972 "Constraints on movement transformations", Journal of linguistics 8 : 3 5 - 8 5 .
Urbariczyk, S., ed.
1966 Polszczyzna pi^kna i poprawna [Polish beautiful and correct] (Wroctaw:
Zaktad Narodowy im. Ossolinskich).
Wexler, K. - P. Culicover
1976 "Formal properties of language acquisition", unpublished manuscript.
ANDRZEJ KOPCZYNSKI
Problems of quality in conference interpreting
Conference interpreting, as is well known, is a type of oral translation prac-

ticed at international meetings to aid participants to understand each other.
Two main varieties are distinguished:
consecutive interpreting (CI), in which the message is first spoken and
then translated; in CI the interpreter does not normally use electronic
equipment but frequently takes notes to help him remember the input
message; he is usually visible;
simultaneous interpreting (SI), in which the message is translated while it
is being spoken; in SI the translation is transmitted through electronic
equipment and the interpreter is usually not visible to the audience.
There are other subdivisions, e.g., Van Hoff 1962 subdivides CI into
integral (integrale), when the text is translated in extenso without ab-
breviations,
abbreviated (abregee), when the text is translated in a summary form,
interrupted, when the text is translated paragraph by paragraph, sentence
by sentence, etc. also called bilateral interpreting,
uninterrupted, when the whole text is spoken without interruptions and
then translated.
Van Hoff says that the most difficult and accomplished type of CI is integral
and uninterrupted but the most common is abbreviated and interrupted.
SI is subdivided into
an extempore translation from a written text (traduction vue, translation
at sight),
whispering simultaneous (le chuchotage), when the interpreter sits near the
hearer and whispers the text into his ear,
simultaneous proper, with the use of electronic equipment.
It is claimed that while SI is more widespread because of its efficiency and
speed, in CI logical organization of the input text is rendered more exactly
than in SI because the interpreter translates longer chunks of the text.
It is also claimed that the main difficulty in SI is the division of attention
connected with simultaneous listening and translating while in CI it is the
284 Andrzej Kopczyriski
overloading of memory in connection with the bigger chunks of input text to

be translated.
There is not very much research done on conference interpreting, especial-
ly from the linguistic point of view (but see e.g. Goldman-Eisler 1968;Barik
1969, 1973, 1975; Gerver 1969, 1971, 1976; Minyar-Beloruchev 1969;
Chernov 1978; Shirayev 1979).
In the present paper I wish to address myself to quality in conference in-
terpreting suggesting certain methods in approaching the problem. I shall try
to define an idealized translational competence against which departures in
actual performance can be discussed. The focus is primarily linguistic, i.e., we
are interested in linguistic operations of the interpreter, whether in SI or CI,
on the input text in Lj () as evidenced by the output text in L 2 (T 2 ). Na-
turally, a purely linguistic approach to the problem is not sufficient and there
is a need of placing the linguistic act of interpreting within a broader context
of communication. We shall touch on this problem too.
Finally, when there are departures or deviations from model translation
there is a need to determine the borderline between correct and erroneous
translation. And this is the main task of this paper.
In the communicative act of translation the translator is a black box that is
not accessible to empirical study. The only elements that are subject to obser-
vation are the input and output texts, the translator's behaviour, his strategies
of tackling the task of translation, etc. One way of approaching this problem
is to hypothesize a model of translational competence of an ideal translator.
We shall assume that the following elements add up to translational com-
petence:
a) perfect knowledge of grammars of Li and L 2 including the knowledge of
rules at all levels: semantic, syntactic, morphological, and phonological;
the knowledge of the totality of lexicons of L j and L 2 including the ability
to use all terminological fields and styles:
b) possession of an unfailing transcoding mechanism, which consists of:
b l ) a decoding mechanism which enables the translator to perceive and
interpret all of the texts of Li and L 2 ,
b2) a transcoding mechanism which enables the translator to convert all
texts in Li OYs) into equivalent texts in L 2 (T 2 's), and vice versa,
b3) an encoding mechanism which enables the translator to generate equi-
valent texts in L 2 , and vice versa (this mechanism is a corollary of Bj
and B2 and, it seems, does not have to be mentioned as a separate
entity).
We shall also assume that the above statements about translational competence
are common to both written and oral translators. The differences that exist
are determined by the channel of translation and the situational context.
Quality in conference interpreting 285
The choice of the channel, either written or spoken, imposes specific con-
straints on the situation of the act of translation and, consequently, on the
requirements on the translator as well as the form of the message.
Here are some of the situational differences in interpreting as opposed to
written translation: the same context of situation for the speaker, interpreter
and the hearers; transient character of the message; transcoding spoken message.
As a result of these differences there are specific requirements on the inter-
preter, which can be summarized as follows: perfect command of the spoken
mode of Lj and L 2 ; ease of spoken expressions; good memory and/or skill of
note-taking (CI); special ability to divide attention between listening and
speaking (SI); resistance to psychological pressure resulting from the transient
character of the message and the presence of the speaker and the hearers;
ability of quick response to necessary adjustments in interpreting.
From our discussion of translational competence it follows that an ideal
interpreter/translator has an ability to transcode TVs into equivalent T 2 's,
which also means that he can unequivocally tell which texts or sentences in
L j and L 2 are equivalent. It seems that the problem of equivalence is of
crucial importance when we discuss quality in interpretation and in analyzing
the interpreter's strategies.
I submit that four concepts are important when we talk about interlan-
guage synonimity: equivalence, congruence, correspondence and communica-
tive appropriateness. Let us discuss each of them briefly,
- equivalent are pairs of messages in L t and L 2 when they have the same
semantic representations,
- congruent are pairs of messages in L j and L 2 when they have the same
semantic representations and consist of the same number of equivalent for-
matives arranged in the same order (Krzeszowski 1971, Marton 1968).
It follows from the definition that a pair of congruent constructions in ad-
dition to the same deep structures must meet the requirement of the sameness
of their surface structures. This claim on congruence is a very strong one espe-
cially the sameness of the number and order of formatives, and is suited for
very delicate contrastive analyses of restricted subsystems of grammar of re-
lated languages. It severely limits the number of congruent constructions.
Consider:
(1) a. Marek zaprosit Susan do Polski.
'Mark invited Susan to Poland.'
b. Mark invited Susan to Poland.
(2) a. Chiopiec zaprosit dziewczyng do Polski.

'The boy invited the girl to Poland.'
b. The boy invited the girl to Poland.
286 Andrzej Kopczynski
(3) a. Ten wysoki chtopiec zaprosit Susan do Polski.

'This tall boy invited Susan to Poland.'
b. Mark has invited Susan to Poland.
(4) a. Czlowiek, ktorego widziates wczoraj w autobusie,
zaprosit Susan do Polski.
'The man that you saw on the bus yesterday invited
Susan to Poland.'
b. The man that you saw on the bus yesterday invited Susan
to Poland.
(5) a. Marek zaprosit Susan do Polski.
'Mark invited (preterite) Susan to Poland.'
b. Mark has invited Susan to Poland.
(6) a. Marek zaprosit pi$knq dziewczyn, ktorq widziates

w zeszfym roku, do Polski.
'Mark invited that beautiful girl that you saw last year to
Poland.'
b. Mark invited Susan to Poland.
(7) a. Zaprosit em Susan do Polski.
invited Susan to Poland.'
b. I invited Susan to Poland.
If we assume that there is referential congruity between these pairs of sen-

tences, we can say that all of them are equivalent translations of each other.
However, only sentences in (1) can be said to be congruent according to the
above definition. Sentences (2), (3), (5) and (6) do not have the same number
of formatives because in (2) English requires the definite article absent in the
Polish sentence, in (3) and (6) the Polish sentences contain a structure of
modification absent in English. In (4) the sentences are not congruent because
of the articles and the fact that adverbials are differently placed. In (7) the
Polish sentence does not contain the subject pronoun which in Polish is ex-
pressed in the ending of the verb.
For analyzing translations another concept is useful, one that separates
translations in (1), (2), (4), (5) and (7) from those in (3) and (6), i.e., literal
translations from paraphrases. To define the latter the concept of equivalence
is sufficient, to define the former we must arrive at a concept of formally
closest possible translation. Therefore, our claim on what in contrastive ana-
lysis is called congruence must be weaker and must exclude differences be-
tween languages. We shall call the formally closest possible translations 'cor-
respondent' and define them as follows:
correspondent are pairs of messages in L t and L 2 when they have the same
semantic representations and are word-for-word translations of each other
inasmuch as the rules of grammar of either language permit.
Correspondent are both those cases where word-for-word translation is pos-
sible and those where such translation is impossible because of structural or
cultural differences between languages provided they are the closest possible
translations. Thus,
(8) a. George visited Agnes yesterday.

b. Jerzy odwiedzii Agnieszk wczoraj.
'George visited Agnes yesterday.'
(9) a. There were many people present at the reception.
b. Na przyjeciu by to wiele ludzi.
'At the reception were many people.'
(10) a. This policeman is trigger-happy.
b. Ten policjant przy byle okazji strzela.
'This policeman shoots at every occasion.'
By our definition all of the above sentences are correspondent because all of
them are the closest possible translations overriding language differences.
From the point of view of congruence only sentences in (8) meet this con-
dition and the rest of them are equivalent.
By contrast, consider the following:
(11) a. George is getting married tomorrow.
b. Cztowiek, ktorego kochasz, zeni sif jutro.
'The man you love is getting married tomorrow.'
(12) a. George is getting married tomorrow,

b. Jerzy zeni si? jutro.
'George is getting married tomorrow.'
(13) a. The president of the United States paid an official visit

to Warsaw.
b. Prezydent Stanow Zjednoczonych odwiedzii Warszaw?.
'The President of the United States visited Warsaw.'
(14) a. The President of the United States paid an official visit
to Warsaw.
b) Prezydent Stanow Zjednoczonych zlozyl oficjalnq
wizyte w Warszawie.
'The President of the United States paid an official visit
to Warsaw.'
Only (12) and (14) are correspondent translations. (11) and (13) are not
because literal translations are possible but the translator, for some reason,
chose to use paraphrases. The reason for such a choice cannot be explained in
purely linguistic terms and one must take into consideration a larger context.
Hence the need for a concept of communicative appropriateness, which can
be defined as follows:
communicatively appropriate are pairs of messages in L t and L 2 when they
perform the same (= similar) communicative functions in terms of
the speaker, his status and the status of his receptors,
the speaker's intention in performing the speech act,
the speaker's stance toward the speech act and the receptors,
the receptors' stance toward the speech act and the speaker,
the illocutionary force of the speech act,
the verbal, paralinguistic and non-verbal form of the speech act,
the existing norms of interaction and interpretation,
the setting.
Here are some examples of translations in which some element of communi-
cative appropriateness has been missed.
(15) a. You're right, I believe,

b. Ma pan racj.
'You're right.'
(16) a. John might be aware of Jane's wedding,

b. Jan wie slubie Janki.
'John knows about Jane's wedding.'
(17) a. A: Shall we go to the movies?
: (nod of the head in agreement)
b. A'.Pojdziemy do kina?
'Shall we go to the movies?'
:Dobrze.
O.K.'
(18) a. Get lost!

b. Ten pan uprzejmie pana prosi, aby pan oddalii.
'This gentleman kindly asks you to go away.'
In translations (15) and (16) exponents of the speaker's stance toward the
proposition have been missed I believe, might. In (17) the non-verbal signal
of agreement was replaced by its verbalized counterpart dobrze O . K . ' All
of the non-verbal signals must be missed in SI because the interpreter is not
visible. They can be easily expressed in CI.
In (18) the form of the translation is inappropriate in terms of style.

All of the above - except possibly in (17) entail a change of the illocu-
tionary force of the utterances.
Recapitulating, we claim that the three linguistic concepts of equivalence,
congruence and correspondence are useful in analyzing translations. All of
them have one feature in common: semantic identity (= similarity) of con-
structions under consideration. Equivalence does not posit rigid constraints
on formal similarity. Correspondence is a yardstick for the greatest formal
fidelity of translation overriding language differences. Congruence tells us
whether such fidelity is possible focussing on language differences.
In addition to the linguistic concepts, there is a need for a broader notion
of communicative appropriateness. It helps us view translations in terms of
similarity of their communicative functions. It should also be an instrument
for explaining certain linguistic choices.
Positing correspondence in translation is not to claim that literal transla-
tion is superior to non-literal translation (although in informative translation
other things being equal - literal translation is probably more desirable).
The above concepts are first of all useful in observing translation as an activity,
particularly observing the strategies of the interpreter in what he does to the
input text.
All of the above concepts also chart a cline along which we should be able
to tell where the borderline between correct and erroneous translation has to
be drawn.
Let us assume that the ideal is a correspondent and communicatively ap-
propriate translation, i.e., a translation that does not contain any optional de-
viations from T!.
The requirements as to the quality of translation, however, should be de-
pendent on its goals. In literary translation the goal is aesthetic and the trans-
lator has to take into consideration the poetic form of the message in addi-
tion to the content. In anthropological translation the goal is an explication
of the cultural content of the source message and the proper form of trans-
lation is of secondary import. In informative translation, of which interpret-
ing is a part, the primary goal is the transfer of information, i.e., the presenta-
tion of the content of T j (cf. also Casagrande 1954, Ure 1964). The con-
straints on the form of informative translation are much less stringent than on
literary translation.
There is still another dimension to consider: the oral-written scale. We
want to claim that in oral informative translation the requirements as to the
form must be less demanding that in written informative translation. This fact
follows from differences between the spoken and written mode. Let us men-
tion some of the differences: in the spoken mode the form of the text is more
discourse-bound and contains a greater number of elliptical constructions

with the speaker relying more on extratextual referential coherence than on
textual cohesion (cf. Widdowson 1973); a looser inter- and intrasentential
organization with less complexity greater reliance on parataxis than hypo-
taxis; sentences frequently run together due to a neglect of oral demarcation
cues; numerous repetitions, hesitations, false starts, etc.
From a purely linguistic point of view deviations from the content or form
of T! can occur at four linguistic levels: semantic, syntactic, morphonological
and stylistic.
We can say that in conference interpreting non-correspondent rendition is
correct when deviations from Ti are stylistic in character if by style we mean
the expression of the same basic meaning of the message by means of differ-
ent linguistic forms. For instance, a T x expressed by a complex sentence
which was translated by two simple sentences can be considered as having the
same basic meaning and expressed by two stylistic variants:
(19) After the delegation held talks with the Prime Minister, it
toured the city.
Delegacja odbyla rozmowy Premierem. Nast^pnie
zwiedziia miasto.
'The delegation held talks with the Prime Minister. Then it
toured the city.'
Traditionally speaking, deviations at the syntactic and semantic level presup-

pose erroneous translation. The problem is not so simple, however. A syntac-
tic deviation violates the L 2 norm. It certainly impairs communication but
does not necessarily cause its complete blockage. If, e.g., the sentence
(20) Oficjalna delegacja jutro przyjedzie

'The official delegation will come tomorrow'
is translated
The official delegation tomorrow will come
the obvious outcome is a syntactic error. A question arises whether such an
error is also an error in translation. It seems that from the point of view of the
goal of interpreting, i.e. the expression of the thesis of T j , it is not an error in
translation if communication has not been blocked.
There is a need then of distinguishing between errors of linguistic com-
petence and errors of translational competence depending on the goal of trans-
lation. Not all errors of linguistic competence must necessarily be errors of
translation competence, and vice versa. If the sentence
(21) Delegacja oficjalna jutro przyjedzie

'The official delegation will come tomorrow'
is translated as
The official delegation will come the day after tomorrow
we will have an error of translational competence although there is no error
of linguistic competence.
Deviations can also occur at the level of communicative appropriateness.
The basic questions to be asked here are: Does the translation render the in-
tention of the speaker, his attitude towards the message and the audience? Is
the translated message taken in the same way in L 2 as it is in Lj? Is the form
of T 2 appropriate in terms of the communicative context? Does it have the
same illocutionary force? Does T 2 render paralinguistic and non-linguistic fea-
tures of T!?
All of these factors have to be taken into consideration in evaluating the
quality of translation. Before we proceed to that, however, let us briefly dis-
cuss the error itself.
Generally speaking there are two main definitions of the error that are
distinguished: one concentrating on the code itself and the other viewing an
error from the point of view of the communicative goal of the speech act for
the latter (cf. Enkvist 1973):
an error is an utterance which deviates from the adopted norm (= usually
the standard variety of L 2 );
- an error is an utterance which blocks communication (= very broadly con-
ceived, e.g., information, aesthetics, cultural content, etc.); thus not every
error in the code must necessarily be an error in communication if the goal
of communication has ben fulfilled.
There is another dichotomy generally recognized: errors of competence, i.e.,
those that violate rules unknown to the speaker and errors of performance,
those that violate rules known to the speaker (cf. Corder 1971, 1967 and
many others). This distinction, however, causes many difficulties as we shall
see below.
Recapitulating the above it seems that it is necessary to distinguish the fol-
lowing categories of errors:
a) errors of translation
b) errors of linguistic competence
c) errors of linguistic performance
d) errors of communicative appropriateness.
a) Errors of translation is the most general category overriding the other three
in the sense that an error of translation may be caused by any of the remain-
ing categories or a combination thereof. It results in a non-equivalent rendi-

tion, i.e., it changes the meaning to the point of blocking communication,
b) and c) The competence-performance dichotomy in the treatment of errors
is based on the inner knowledge of the speaker and extremely difficult to
get at in practice. One of the major procedures available to us at the pre-
sent state of our knowledge is a direct elicitation from the speaker whether
or not he knows the rule involved. This procedure, however, is not 100%
reliable because the speaker may know the rule when asked but will violate
it in speaking because he has not internalized it. Therefore two more aspects
are worth considering in the treatment of errors of competence and per-
formance. The error can be viewed in two ways as product and as pro-
cess. The product aspect of the error is the objective outcome of the error
as it is heard by the hearer. No matter whether the speaker knows the rule
or not, when he says "The two Ministers has expressed. . .", the net result
for the hearer is a violation of the grammatical rule. In other words, the
hearer receives the error as if it were an error of competence. There is an-
other aspect: the question whether the error is identifiable by the hearer
or not. The error above is easily identifiable but there are others that are
not unless the hearer compares them with T j , and in the situation of inter-
preting it is usually not possible. Most of the grammatical errors are easily
identifiable but a lot of lexical errors are not unless they overtly violate
e.g. collocational rules. Identifiable errors of performance are slips, hesita-
tions, repetitions, etc.
Considering the above, it will probably be wise to classify errors in conference
interpreting along the following categories:
errors of competence (= "as if they were errors of competence from the
point of view of the hearer")
errors of performance (= identifiable errors of performance)
ommissions and additions
errors of appropriateness
errors of translation.
Below we shall give examples of each category on the basis of an experiment
conducted with students of interpreting at Warsaw University (four CI and
four SI subjects translating from Polish into English and English into Polish,
see Kopczynski 1980).
Errors of competence
Syntactic and phrasal errors of competence invariably result in a grammatical-
ly deviant form in L 2 . They are therefore easy to identify by the hearer. They
create noise in the channel and can block communication altogether. Here are
some examples:
(22) faulty word order

. . . za zdrowie wszystkich obecnych tutaj przyjaciot i
towarzyszy...
' . . . to the health of all our friends and comrades present
here . . . '
- to the health of all the present here our friends and com-
rades
It is characteristic that all of the word order errors were made in SI. It is a
type of error that reflects a strategy of the interpreter: he keeps too close to
the speaker and follows the Polish word order.
(23) agreement
. . . tak narod chinskijak amerykanski cechuje. ..
'. . . the Chinese and American nations are characterized . . .
the Chinese and American nations is characterized
Among the phrasal errors the most common are those affecting the noun
phrase, especially those in the use of the article and prepositions, e.g.
(24) a. . . . Przewodniczacy Mao Tse-Tung. ..
' . . . Chairman Mao Tse-Tung . . . '
- the President Mao Tse-Tung
b. . . . przy tej okazji
. .on this occasion'
- in this occasion
They are committed with great regularity, which is one of the indications
that they are true errors of competence.
As opposed to grammatical errors lexical ones are not easily identifiable.
Some of the types found were: broadening of Lx meaning, narrowing of L j
meaning, use of a wrong meaning of a polysemous item, use of an L 2 item
similar to L t item, use of an L 2 item similar to another L 2 item, etc., e.g.
(25) a. . . . wszystkich obecnych tutaj przyjaciot i towarzyszy...

' . . . all the friends and comrades present here . . .'
- all people present here
b. . . . I ask you to join me in raising your glasses ...

- aby wzniesli szklanki -
'in raising your tumblers'
c. . . . through the wonder of telecommunications ...

dzigki cudowi komunikacji
'through the wonder of transportation''
d. . . . sita napedowa. ..
'motive force'
- motoring force
Errors of performance
Among errors of performance we have included all those identifiable disturb-
ances which deviate from its fluent rendition by an ideal speaker-interpreter,
e.g.
(26) hesitation:
a. exchanging . . . eh . . . of views on issues of . . . eh
interest
b. - chcialbym . .. mm . . . ziozyc ho id
'I'd like . . . eh . . . to pay homage . . . '
(27) stuttering:
- The Chinese peo . . . peo . . . people is a great peo . . .
peo . . . people
(28) repetition:
the . . . the . . . Chinese government
(29) false start:

a. mozemy starac sig ... y ... t% przepasc . . . nad tq prze-
pasciq wybudowac most
'we can try . . . eh . . . this gulf . . . to bridge this gulf
b. when the common . . . mm . . . hopes w i l l . . . mm . . .
be . . . mm . . . come true
Errors of performance were the largest group found in the analyzed texts.
There was almost no single utterance both in SI and CI without them. There-
fore it can be assumed that they are the exponents of
- characteristic features of the spoken subcode: evidence of planning and
monitoring speech while interpreting; in this sense they can be considered
not so much as errors but rather as part and parcel of spontaneous speech;
- deficiencies in linguistic competence: they precede, i.e., signal errors of
competence;
- deficiencies in translational competence: difficulties in selecting a gram-
matical construction or a lexical item to fit the preceding text; difficulties
in choosing the optimal ear-voice span in SI and in storing L t material in CI.
Omissions and additions

Omissions are the next largest group. In our analysis we have taken into con-
sideration only optional omissions and additions. Obligatory omissions and
additions are those that arise from differences in structures and cultures of L t
and L 2 , e.g.
(30) The parliament began its session on Friday

- Parlament rozpoczql obrady w piqtek -
'Parliament began session on Friday'
(31) Mr President and Mrs Nixon . . .

- Partie Prezydencie i Szanowna Pani Nixon... -
'Mr President and Distinguished Mrs Nixon . ..'
In (30) the words the and its are omitted in Polish because of differences in
structures and in (31) Szanowna 'Distinguished' is a necessary addition arising
from a difference in cultural conventions.
By contrast:
(32) . . . chcialbym powitac Prezydenta Nixona oraz innych

naszych amerykanskich gosci.. .
- I wish to extend our welcome to President Nixon and our
[other] American guests -
In the translation the word other was missed without any linguistic or cultural
motivation. It's an example of an optional omission.
In our material there is a definitive pattern among omissions and additions
which points to a strategy in tackling Tj by the interpreter. The overwhelming
majority of omissions were those of modifying elements of different kind:
from an omission of a determiner, an adjective, an adverbial to that of relative
phrases and clauses and adverbial clauses. The strategy is apparently to focus
attention in the first instance on the constitutive part of T x the main clause,
the main verb, the head noun. The modifying elements are evidently viewed
as redundant and frequently dismissed.
The causes of those errors cannot be unequivocally determined; some of
them are probably errors of performance (memory lapses etc.), others are er-
rors of receptive competence.
Although we found a small number of additions in the translated texts,
they followed the same pattern - most characteristic additions were those of
modifying elements.
Errors of appropriateness
In our corpus they involved inappropriate usage in terms of register and in-
appropriate expression of illocutionary force.
(33) a. Przewodniczacy Mao Tse-Tung

'Chairman Mao Tse-Tung'
President Mao Tse-Tung -
b. . . . chciaibym powitac Prezydenta Nixona i jego mal-
zonkf...
' . . . and his wife...'
to extend our welcome to President Nixon and
Mrs. Nixon -
c. Jestesmy przekonani...
'we are confident'
we hope -
There is an overlap between errors of appropriateness and other ones, e.g.,

a lot of lexical errors are at the same time errors of appropriateness in that
they affect the choice of terminology within a register.
Errors of translation
As mentioned above errors of translation are those that block communication
and are caused by errors of competence, performance or appropriateness, or a
combination thereof.
Theoretically speaking T 2 in relation to Tj can be:
(i) correspondent and communicatively appropriate, i.e., identical (= sim-

ilar) in meaning and form, e.g.
Narod amerykanski jest wielkim - The American nation is a great
narodem. nation.
'The American nation is a great
nation.'
(ii) equivalent and communicatively appropriate, i.e., identical in meaning

but not in form, and without errors of competence, performance and
appropriateness, e.g.
Never have I heard American Nigdy jeszcze amerykanska

music played better in a foreign muzyka nie by la grana lepiej za
land. granicq. -
'Never has American music been
played better in a foreign land.'
practically equivalent, i.e., similar in meaning, intelligible but with errors
of competence, e.g.
Ministrowie wyrazili. .. The Ministers has expressed .
'The Ministers have expressed ..
practically equivalent with errors of performance, e.g.
At this moment through the won- - We tejchwili chocia ... eh...
der of telecommunications more dzi^ki cudowi telekomunikacji
people are seeing and hearing what wigcej ludzi moze widziec i
we say than on any other occasion slyszec, co mowimy niz
in the whole history of the world. kiedykolwiek w okazji. . . w
calej historii swiata. -
'At this moment although eh
thanks to the wonder of tele-
communications more people
can see and hear what we say
than any other occa - in the
whole history of the world.'
practically equivalent with errors of appropriateness, e.g.

a. Chairman Mao . . . Prezydent Mao . ..
'President Mao . . .'
b. Panie Prezydencie i Szanowna Mr President and Mrs Presi-
Pani Nicon . . . dent -
'. . . and Airs. Nixon
practically equivalent with any combination of the above types of errors:
non-equivalent without identifiable errors of competence, performance
and appropriateness, i.e., a T 2 which is well-formed in terms of L 2 but
which changes the sense of !, e.g.
In the spirit of frankness which - Mam nadziejf, ze ten duch
I hope will characterize our talks szczerosci b^dzie dominowai w
this week.. . czasie naszych rozmow dzisiej-
szego wieczora . . .
' - this evening -'
298 Andizej Kopczynski
(viii) non-equivalent with errors of competence, e.g.

Ludzie i tylko ludzie sq silq - This people and only people
napgdowq w tworzeniu historii than can lead in the history of
swiata . . . the world
'The people, and the people
alone, are the motive force in
the making of world history.'
(ix) non-equivalent with errors of performance, e.g.
On behalf of all your American - W imieniu wszystkich
guests I wish to thank you for amerykanskich gosci chciaibym
the incomparable hospitality for podziqkowac narodowi
which the Chinese people are chinskiemu . . . eh . . . swym
justly famous throughout the siynnym na catym swiecie.
world. O n behalf of all the American
guests I wish to thank the Chi-
nese nation eh - in its famous
throughout the world.'
(x) non-equivalent with errors of appropriateness, e.g.
Jestesmy przekonani, ze We hope that the day will
nadejadzie dzien. come.
'We're confident that the day will
come.'
(xi) non-equivalent with any combination of the above types of errors.
Cases (i) and (ii) meet all the requirements of correct translations as they were
defined above.
There may be some questions concerning cases (iii)(vi). What we have called
'practically equivalent' T 2 's must necessarily change the exact sense of T t but
they are still understandable from the context in spite of the errors they con-
tain. They are borderline cases which do not block communication although
they are formally erroneous. It seems to us that it is reasonable to distinguish
this category of T 2 's in goal-oriented concept of the error adopted here to ac-
count for the fact that a lot of formally erroneous oral translation performs
its communicative function of getting the message across. The problem is how
to exactly distinguish them from cases (vii)-(ix). One method is to do it
operationally by asking a group of bilingual judges about the sense of Tj and
T 2 : if the meanings coincide the T 2 should be subsumed under 'practically
equivalent'. Another method that can be used - although sparingly is back-
translation.
Non-equivalent T 2 's are those that block communication. From the above
typology it follows that only cases (vii)(xi) are errors of translation, (vii) be-
ing the 'purest' case of such error.
There is finally one more factor that should be taken into consideration
when we talk about errors in conference interpreting (and written translation
as well): the problem of identiflability of the error. Along this scale we can
have a continuum of texts. The two extremes would be:
1. an apparently flawless text in terms of L 2 grammar and appropriateness
but with numerous errors of competence, performance and appropriateness
that are not identifiable, e.g. case (vii) above.
2. a text which contains identifiable errors of performance and/or competence
and appropriateness but which is still practically equivalent (cases iii-vi).
If along this scale we find texts of type (1) and (2) which both are non-equi-
valent the type (1) text is especially damaging to communication because in
such a case the hearers are not alerted to the imprecision of translation. In the
type (2) text the hearers are at least aware that something is wrong.
Errors of various kinds, especially errors of performance, are part and
parcel of interpreting. Therefore the criteria for evaluating oral translation
should be less stringent than for written translation and the level of expecta-
tion lower. But the minimum limit must be set and that is practically equi-
valent translation, i.e., one that may not be flawless but which carries over the
referential content of the original.
We might say that successful conference interpreting is successful not be-
cause it is free of errors but in spite of errors.
References
Barik, H. A.
1969 A study of simultaneous interpretation, unpublished PhD diss. (University of
North Carolina).
1973 "Simultaneous interpretation: temporal and quantitative data", Language
and speech 1 6 : 2 3 7 - 2 7 0 .
1975 "Simultaneous interpretation: qualitative and linguistic data", Language and
speech 1 8 : 2 7 2 - 2 9 7 .
Casagrande, J. B.
1954 "The ends of translation", IRAL 2 0 . 4 : 3 3 5 - 3 4 0 .
Chernov, G. V.
1978 Teoria i praktika sinkhronnogo perevoda [Theory and practice of simultane-
ous interpretation) (Moskva: Mezhdunarodnoye Otnoshenya).
Corder, S. P.
1967 "The significance of learners' errors", IRAL 5 . 6 : 1 6 1 - 1 7 0 .
1971 "Idiosyncratic dialects and error analysis", IRAL 9 . 2 : 1 4 7 - 1 6 0 .
Enkvist, . E.
1973 "Should we count errors or measure success", Errata, edited by Jan Svartvik
(Lund: CWK Gleerup), 1 6 - 2 3 .
Gerver, D.
1969 "The effects of source language presentation rate on the performance of
simultaneous conference interpreters", Proceedings of the 2nd Louisville
Conference on Rate and/or Frequency Controlled Speech, edited by E. Foulke
(Louisville: University of Louisville), 1 6 2 - 1 8 4 .
1971 Simultaneous interpretation and human information processing, unpublished
PhD diss. (Oxford University).
1976 "Empirical studies of simultaneous interpretation: a review and a model",
Translation: application and research, edited by R. W. Brislin (New York:
Gardner Press), 1 6 5 - 2 0 7 .
Goldman-Eisler, F.
1968 Psycholinguistics: experiments in spontaneous speech (London: Academic
Press).
Kopczynski, A.
1980 Conference interpreting: some linguistic and communicative problems
(Poznan: UAM).
Krzeszowski, T. P.
1971 "Equivalence, congruence and deep structure", Papers in contrastive lin-
guistics, edited by G. Nickel (Cambridge: University Press), 3 7 - 4 8 .
Marton, W.
1968 "Equivalence and congruence in transformational contrastive studies",
Studia Anglica Posnaniensia 1 : 5 3 - 6 2 .
Ure, J. L.
1964 "Types of translation and translatability", flaie/ 10.1: 5 - 1 1 .
Van Hoof, M.
1962 Theorie et pratique de ('interpretation (Munich: Max Hueber).
Widdowson, H.
1973 "Directions in the teaching of discourse", Theoretical linguistic models in
applied linguistics, edited by S. P. Corder and E. Roulet (Brussels: Didier),
65-76.
TOMASZ P. KRZESZOWSKI
Tertium comparationis
"To compare them would be tantamount to putting ten-ton

lorries and banana skins into the same class on the grounds
that neither ought to be left on footpaths!" Carl James
One of the reasons why contrastive analysis (CA) continues to perform the
role of Cinderella of Linguistics is the fact that its most fundamental concept
remains as hazy as ever. Although such notions as semantic equivalence, trans-
lation equivalence, congruence, etc., are the standard jargon of the trade and
have received considerable attention from CA-theorists, tertium comparationis
(TC), the concept that lies at the heart of any comparison (eo ipso at the heart
of CA), remains remarkably neglected.
The existing CAs implicitly involve various platforms of interlinguistic refer-
ence, determined by specific linguistic models which they employ and specific
levels of analysis which they embrace. Thus different TCs are used for com-
parisons in lexicology, in phonology and in syntax. With the exception of the
omnipresent initial assumption concerning the semantic identity or at least
similarity of the compared phenomena, in few of these studies explicit men-
tion of any TC is made, leave alone any attempts at justifying a particular
choice. In the existing literature two types of TC have been employed and dis-
cussed: formal correspondence and semantic equivalence. Due to the fact
that neither concept is very sharp a peculiar schizophrenic situation emerged:
while, on one hand, it was fortunate that neither TC has been consistently
used in CA, it was, on the other hand, unfortunate that no satisfactory ana-
lysis of the relevance of these two kinds of TC in CA has been attempted.
Even a cursory glance at the wealth of the existing CAs suffices to notice
that the two types of TC are not the only ones that are used in practice, even
if theoretical discussions tend to be limited to just those two (e.g. Lado 1957,
Spalatin 1969, Ivir 1969,1970). 1 Moreover, neither of the two types of TC is
based on a homogenous set of assumptions determining practical and theoret-
ical status of CA.
302 Tomasz P. Krzeszowski
The present paper will outline three related topics: 1. the inherent am-
biguity of the concepts 'formal' and 'semantic' TC, 2. other possible TCs rele-
vant to CA; 3. interdependencies between various types of TC.
Formal and semantic likeness have been both used in contrasting various
aspects of languages. The two kinds of resemblance have also received some
attention in theoretical discussions. It has been pointed out, for instance, that
formal likeness alone cannot serve as TC without support from semantic equi-
valence (Liston 1970, Lipinska 1977). At best a comparison based on formal
criteria alone is incomplete, at worst it cannot be performed at all, and in
many cases it is misleading (see also Spalatin 1969).
If, for example, one compares Polish and English personal pronouns, a
formal analysis will ascertain the equivalence between the English you and
the Polish ty/wy and will be accurate as far as it goes. Such an analysis is in-
complete as it leaves out such forms as PanjPani and other possible equivalents
of you. These equivalents can only be established if criteria other than formal
are employed (for details see Krzeszowski 1976). English articles cannot be
compared with anything in those languages in which there are no articles, if
only formal criteria are considered. Finally, in the case of such phenomena as
the present perfect tense in English and passe compose in French the formal
analysis is misleading since the formal similarity is not matched by semantic
similarity, a kind of situation which often causes considerable learning pro-
blems (see Politzer 1968). Therefore, it is generally recognized that a con-
trastive analysis based on purely formal criteria falls short of both theoretical
and practical expectations.
Somewhat less obviously a contrastive analysis based on semantic similarity
alone can also be inadequate and misleading. In the contrastive practice se-
mantic equivalence is often erroneously identified with translation equivalence:
To establish that these (systems of deictics, supplied by me] are comparable, we first
need to show their contextual equivalence; this can be done most simply by reference
to translation (Halliday et al. 1964:115).
One can easily show that translation equivalents are often semantically non-
equivalent (for discussion and examples see Krzeszowski 1974, 1979). It ap-
pears that semantic equivalence must be constrained formally, while trans-
lation equivalence does not have to be so constrained. When one translates
one departs from semantic equivalence due to three types of reasons: 1. errors
in translation (which will be disregarded here); 2. formal properties of respec-
tive languages; 3. what is generally loosely termed 'stylistic' reasons. These
three types of reasons lead to situations in which actual translation practice,
with the exception of legal texts, seldom involves translations which are
semantically equivalent. This means that only some translations can be used
as data for systematic CAs (cf. Ivir 1969), while translation as a method of
contrasting must be regarded with caution.
Translation may be viewed amorphously as the rendition of a text from one language to
another. This is translation from the standpoint of la parole: the text, the act of speech
or writing, is the thing. Or it may be viewed as a systematic comparison of two Lan-
guages: this is translation from the standpoint of la langue. (Bolinger 1966:130)
In fact the use of translation in systematic CAs is highly limited:

Translation equivalence serves merely to help us isolate items of structure with shared
meanings in the two languages (Ivir 1970:15).
In the subsequent remarks I would like to say a little more about the inade-
quacy of the approaches limited to the alternative 'formal and/or semantic'
resemblance as the only possible TC. To begin with, as has been said earlier
on in this paper, the very meaning of the two concepts is richer than is com-
monly recognized in CA. For instance, 'formal' can be extended to cover the
entire plane of expression (in Hjelmslevian terminology). In most American
studies 'formal' is restricted to word order, function words, inflections, af-
fixation and suprasegmentale. In the broader perspective 'formal' would also
embrace such aspects of expression as alliteration, rhymes, rhythm, to men-
tion just a few phenomena of clearly 'formal' nature but extending beyond
the limits of traditional interpretations of 'formal'. Many of these 'formal'
properties would find their place in the study of functionally equivalent texts.
The notion 'semantic' is often also extended to cover matters of pragmatics,
especially by those authors who identify semantic equivalence with translation
equivalence:
Our experience is that languages can be effectively contrasted only on a semantic basis,
specifically, on the basis of translation equivalence (Spalatin 1969:34).
In reality, many authors have shown that semantic equivalence is not a neces-
sary prerequisite of a good translation (cf. Riilker 1973, Krzeszowski 1974,
Kopczynski 1980). What is expected of a correct translation is pragmatic or
functional equivalence. It cannot be denied that pragmatic equivalence can
serve as TC for contrastive analyses of such matters as the structure of dis-
course, stylistic properties and quantitative aspects of texts. But syntactic
CAs, the primary concern of earlier CAs, must be conducted within the limits
of the semantic component of the language, or more specifically that part of
the semantic component which can be systematically and predictably corre-
lated with the grammatical structure of sentences.
This restricted sense of 'semantic' still embraces some aspects of meaning
which are traditionally relegated to 'pragmatics' or "interpersonal function"
of sentences (Halliday 1970). According to Halliday the systems of mood and
304 Tomasz P. Kizeszowski
modality are precisely those systems which relate sentences to their inter-
personal functions. I suggest that the notion 'sentence-semantics' should cover
those elements of 'pragmatics' which can be correlated with the structure of
sentences, even if consistency in this area is definitely out of the question: de-
clarative, interrogative and imperative sentences do not necessarily perform
the functions of statements, questions and commands, respectively. In so far
as some correlation between form and function does exist those 'functional'
aspects of sentence structure constitute the border area between pragmatics
and semantics and should be included in any semantic-syntactic CAs (for de-
tails see Krzeszowski 1974). Therefore, James (1980) suggests that for the
purposes of CA translation equivalents should be limited to those which are
both semantically and pragmatically equivalent. However, this proposal also
raises doubts. Presumably what James means by 'semantic' refers to Halliday's
'ideational' function of sentences, while Halliday's 'interpersonal' and possi-
bly 'textual' functions fall under 'pragmatic'. Under James' proposal many sen-
tences across languages would exhibit both 'ideational' (semantic) and 'inter-
personal' (pragmatic) equivalence, yet one would hardly wish to use them as
data for syntactic CAs. For example:
(1) Did he kill that dog?

(2) Zabit tego psa?
Lit. 'Killed-he that dog? '
are equivalent, both ideationally (agent, transitive, verb patient) and interper-
sonally (general question); yet they falsely suggest a relationship between syn-
tactic types represented by 1 and 2 in English and Polish. A systematic syn-
tactic correspondence would have to be ascertained between 1 and 3 rather
than between 1 and 2:
(3) Czy on zabit tego psa?

Lit. 'Whether he killed that dog? '
In another paper (Krzeszowski 1981a) I suggested that syntactic contrastive

studies can be performed on data restricted in the following way: a contrastive
grammar will take as its primary data (to be assigned the status of semanto-
syntactic equivalence) the closest approximations to grammatical word-for-
word translations and their synonymous paraphrases, if such exist.
Such constraining of primary data as basis for syntactic CAs bypasses the
inherent difficulties of the proposals suggesting the unrestricted semantic
equivalence as basis for comparison. Accepting any translation as a possible
basis for syntactic CAs leads to two mutually exclusive, but both undesirable
consequences. Either 1. no comparative generalizations are possible, as the
number of well-formed translations of a particular sentence into another lan-

guage cannot be predicted a priori, or 2. purely arbitrary decisions concerning
formal correspondences in the unconstrained translations must be made. Any
non-arbitrary decision involves circularity: the investigator has to assume
formal correspondences on the basis of syntactic and/or morphological features
which the compared texts share. This circularity is even reflected in the use of
the term "comparable" in certain contrastive grammars. For example, Stock-
well et al. (1965:65) thus write about determiners in English and Spanish:
Both English and Spanish have two sets of determiners, commonly referred to as detinue
and indefinite articles. In many respects they are comparable [my emphasis]; in others
they are different.
The circularity consists in the following: we compare in order to see what is

similar and what is different in the compared materials; we can only compare
items which are in some respect similar, but we cannot use similarity as an in-
dependent criterion for deciding how to match items for comparison, since
similarity (or difference) is to result from the comparison and not to motivate it.
To avoid this undesirable circularity, in deciding about formal correspond-
ences one needs a common TC outside the formal properties. The underlying
meaning of the closest approximations to acceptable word-for-word transla-
tions provides such a TC. Sentences and constructions sharing identical se-
mantic representations at the level of sentence semantics (but necssarily ex-
hibiting certain idiosyncratic differences due to the differences at the level of
word-semantics) are in the relation of semanto-syntactic equivalence and con-
stitute a constrained set of data for syntactic CAs. The approach through con-
strained translations does not require the initial recognition of shared syntac-
tic categories as TC for syntactic CAs. Such a recognition would illegitimately
anticipate the results of CAs (for details see Krzeszowski 1974 and 1979).
Recapitulating this section, let us say that formal properties alone do not
provide an adequate TC for syntactic CAs, while a semantic TC must be con-
strained through restricting the scope of translation equivalents as primary
data for syntactic CAs.
Such constrained, but rigorous, CAs have very limited pedagogical rele-
vance. Any extension of the scope of CAs to make them pedagogically more
useful increases the likelihood of their becoming less rigorous and hence less
respectable as a 'scientific' procedure. One has to look for ways of extending
the scope of CAs without losing any of the rigor characterizing syntactic CAs.
Formal and semantic types of TC discussed so far will not suffice as basis for
such extended CAs. Also phonetic and phonological CAs cannot rely on se-
nantic equivalence as TC. Therefore, in what follows I am going to look at
Dther possible kinds of TC. In the subsequent discussion I shall consider vari-
ous kinds of equivalence as directly determined by various kinds of TC. It will

appear that no single type of TC is sufficient as a foundation of the numerous
existing approaches to CA. Therefore, an attempt towards a typology seems
to be in order.
The first division will have to be drawn between text-bound and systematic
(projective) CAs, reflecting the familiar distinction between la parole and la
langue. Text-bound CAs embrace studies of translations, but not only, as we
shall presently see. Projective CAs are related to text-bound CAs in the same
way in which the study of language is related to the study of texts. Such
studies go beyond the primary linguistic data found in texts and aim at gener-
alizations about various aspects of the contrasted languages.
As was said a few lines earlier, the study of translation is not the only pre-
occupation within the domain of text-bound CAs. It is perfectly possible to
conduct CAs of texts which are not translations. Various kinds of quantita-
tive CAs can be performed on texts in two or more languages without the initial
prerequisite that the compared texts should be translations (cf. DuSkova 1978,
Becka 1978). Therefore, throughout the rest of this paper I will find it con-
venient to use the label 2-text to refer to any pair of texts, written or oral, in
two languages, which are used as data for CAs. Any 2-text can be transla-
tionally equivalent. A translationally equivalent 2-text is a 2-text for which
it can be asserted that its constituent texts can function as translations of one
another. A translationally equivalent 2-text is a 2-text about which no such
assertion can be made, but which can be used as a body of data for quantita-
tive CAs. Quantitative CAs make use of such 2-texts as primary data in which
the constituent texts are in the relation of statistical equivalence, i.e. a relation
based on the additional assumption that the constituent texts contain certain
similar elements which can be counted and that the results of these computa-
tions are comparable. Such an assumption can only be justified through refer-
ence to relevant systematic CAs. (For details concerning the relations between
text-bound and systematic CAs see Krzeszowski 1981 b).
Any text-bound CA is corpus-restricted if no systematic generalizations
outside the original data are made. Quantitative CAs are necessarily corpus-
restricted, even if they allow for predictions of statistical nature concerning
other similar texts. Qualitative text-bound CAs may be corpus-restricted if no
systematic generalizations are drawn concerning the compared languages.
However, they may also serve as foundation for projective generalizations, if
clearly stated constraints on the relevant 2-texts are provided. In such events
2-texts serve as primary data on which contrastive grammars as generalizations
about the compared languages are based.
For the purposes of CAs, we shall define TC as the reason why any 2-texts
and/or any elements of two languages are juxtaposed for comparison. Each
Teitium comparationis 307
type of CAs will have its own type of TC thus providing grounds for a
taxonomy. Within each type it should be possible to distinguish more specific
subtypes, unique within each type. Each type of TC will determine a different
kind of equivalence, i.e. the relation obtaining between the compared items.
It will be seen, therefore, that semantic equivalence and translation equivalence,
discussed so far, are but two of seven types of equivalence that I would like
to distinguish. Thus in addition to statistical equivalence and translation equi-
valence, providing TC for text-bound CAs, we shall distinguish several types
of systematic CAs.
In CAs constrained by semanto-syntactic equivalence it is possible to com-
pare a) constructions, b) systems and c) rules. Of these three types only CAs
of constructions are directly based on semanto-syntactic equivalence, con-
straining 2-texts for the purposes of such CAs. The comparison of systems re-
quires dependence on syntagmatic considerations in the same way in which
any paradigmatic analysis is linked to syntagmatic considerations. For ex-
ample, isolating a particular system in a particular language requires an exam-
ination of syntagmatic arrangements of elements, i.e. constructions in which
those elements appear. In CAs, likewise, no equivalence of systems can be
ascertained without assessing the equivalence of constructions in which ele-
ments of those systems appear. Investigators usually compare equivalent sys-
tems across languages, basing on intuitions their decisions to juxtapose the
relevant words, thus employing the concept of system equivalence as TC.
It takes little reflection to realize that system equivalence can be made ex-
plicit only through the examination of constructions in which elements of the
compared systems appear, i.e. via the notion of semanto-syntactic equivalence
relating the relevant 2-texts as primary data.
Likewise, any comparison of rules cannot be divorced from an implicit
comparison of constructions on which these rules operate. Therefore, the
semanto-syntactic equivalence of constructions is the central concept in syn-
tactic CAs, whether or not they are extended to comparisons of systems and
rules (for more discussion see Krzeszowski 1976).
Phonological and lexical CAs cannot be based on semanto-syntactic equi-
valence. Such studies use yet a different type of TC, which can be called sub-
stantial in so far as it is connected with the material substance outside lan-
guage but associated with language through phonological and semantic inter-
face of language (Hjelmslev 1963).
In the case of phonological CAs acoustic, articulatory and in principle also
auditory phenomena provide this substantial TC. Most phonological CAs
make reference to articulatory parameters, less frequently to acoustic ones.
I know of no CA based on auditory parameters alone.
In the case of lexical CAs, the external reality, or strictly speaking its
psychic image in the minds of language users, provides the substantial TC as
items across languages are compared with respect to differences and similar-
ities concerning their reference to various elements of the reality in the world
at large as it is reflected in the minds of language users.
Both phonological and lexical CAs are mainly paradigmatic, even if, espe-
cially in various kinds of generative frameworks, syntagmatic arrangements
are also comparable. However, in the case of phonological and lexical CAs we
deal with a situation which is a mirror image of semanto-syntactic studies:
paradigmatic studies are now central while syntagmatic studies are secondary
and impossible to conduct without the ultimate reference to substantial TC.
Much less rigorous contraints apply to 2-texts which are to be compared
with regard to their styles or registers. Any translation will do here as long as
it is functionally (pragmatically) equivalent. Naturally enough some idealiza-
tion of the data is necessary. For example, obviously erroneous translations
must be disregarded as well as translations which fail to convey adequately
some relevant pragmatic functions, especially if alternative, more adequate
translations are available. Although constraints on the suitability of 2-texts
for this sort of CA are different from those required of 2-texts as data for
semanto-syntactic CAs, they are equally important and must be stated clearly
and unequivocally, lest the CAs will fail to grasp the relevant generalizations
concerning pragmatic or stylistic aspects of the compared texts.
Pragmatic or function equivalence as TC for stylistic CAs is therefore a
relation that holds between the constituents of 2-texts selected in such a way
as to retain certain invariables guaranteeing that such 2-texts are +equivalent
functionally:
Since, however, contrasting does presuppose, apart from agreements and differences
under observation, also a certain common foundation, one proceeds from the fact that
the fundamental differentiation of styles holds in roughly the same way for all cultured
languages and that it makes itself felt in roughly the same tendencies even though not
always realized by the same means of expression. In other words: French scientific style
will be characterized by analogous stylistic tendencies as its counterpart in Czech, in the
same way as the basic features distinguishing the belles-lettres narrative style from
descriptive style will be the same in English as, say, in Italian. This is due to the impact
of social, i.e. extralinguistic communicative needs which e.g. in languages of the Euro-
pean cultural sphere in the given period bear on the whole an analogous character. That
is why we engage in comparing discourses of belles-lettres prose in various languages, or
of scientific style in various languages and so on. In this manner a certain common
foundation is gained upon which contrastive analyses of national styles can be built
(Beka 1978:131-132).
Contrastive studies based on functional (pragmatic) equivalence require a

separate extensive treatment as the number and the nature of elements which
can be compared is as yet undetermined. However, as the pragmatic CAs con-
tinue to flourish (for some examples see Fillmore and Oleksy in the present
volume) a detailed synthetic discussion will hopefully become possible.
The problem of generalizations in CAs requires some comments. Stating
any equivalence relation is itself an act of generalization. Such generalizations
are of little practical value unless they are supported by quantitative data. For
instance, in the area of system comparisons both syntagmatic and paradigmatic
criteria lead to establishing systematic equivalence between Polish and English
personal pronouns: / - ja, you - ty/wy, they - oni/one, etc. Correct, as far
as it goes, this generalization is extremely inadequate as it does not say any-
thing about the relatively low frequency of occurrence of personal pronouns
in Polish texts in contrast with a relatively high frequency of occurrence of
personal pronouns in English texts.
Even more obviously systematic generalizations about equivalence prove
inadequate in those cases when nearly synonymous equivalents are available
in the compared languages, i.e. in the majority of cases. Systematic CAs reveal
structural contrasts and similarities but give no cues as to the actual occur-
rences of the compared phenomena in 2-texts. For instance, without quanti-
tative data there is no way of determing the relative frequencies of such con-
structions as extraposed vs. non-extraposed sentences, psych-moved vs. psych-
non-moved, particle moved vs. particle-non-moved, to mention just a few.
Intuitive judgements and preliminary research clearly show that the relative
frequencies of various types of constructions characterize native vs. non-native
performance in a language. Graczyk (MS) demonstrates the inclination of
Polish learners to overuse extraposed constructions. The relative frequency of
such constructions in the compositions of Polish learners was found to be
significantly higher than in the compositions of native speakers, who used
synonymous constructions, avoiding extraposition.
Various instances of congruence or near congruence with very low prob-
ability of occurrence in actual texts suggest that the significance of results
obtained from systematic CAs is not in any way related to the significance of
results obtained in quantitative text-bound CAs. Therefore, at least from the
pedagogical point of view, quantitative CAs cannot be neglected as they pro-
vide the necessary complementary information which cannot be ignored in
preparing teaching materials, working out remedial courses, and constructing
drills and exercises.
The various kinds of CAs associated with the respective types of equivalence
are presented in the following diagram:
CAs
text-bound systematic
(2-texts equivalent given) (2-texts + equivalent available)
/
- translationally + translationally
structurally functionally
equivalent texts equivalent texts
constrained constrained
quantitative
corpus-restricted projective
systems constructions >> rules

(paradigmatic)
TCs: statistical translation system semanto- rule pragmatic

equivalence equivalence equivalence syntactic- equivalence (functional)
(1) (2) (3) equivalence (5) equivalence
v- - -ÎK^"" X (6)
\
\
phonological CAs lexical CAs
substantial
equivalence
(7)
Summing up, traditionally recognized formal, semantic and translational cor-

respondences do not provide necessary TCs for the various types of CAs, as
more subtle distinctions are required. Systematic CAs alone cannotyield prac-
tically useful results but must be amplified by quantitative CAs which investi-
gate relative frequencies of equivalent phenomena. Hopefully, statistical re-
search will further enhance the traditional workshop of contrastive linguistics.
At the same time, however, it must be stressed that quantitative CAs are im-
possible without systematic CAs as the latter establish common categories at
all levels of linguistic analysis as parameters along which quantitative measure-
ments can be made. Therefore, neither qualitative nor quantitative CAs alone
are adequate both theoretically and practically. As quantitative CAs have
been so far neglected to the detriment of contrastive linguistics in general,
I suggest that this neglect should be made up for in the years to come by
those who wish to continue their interest in the field. Particularly needed are
CAs in the area of phraseology and stylistics.
Note
1. A notable recent exception is provided by James (1980), who mentions "substantive"

TC for phonological and lexical contrastive studies, as well as formal (surface structure),
semantic (deep structure) and translation equivalence as TCs for grammatical studies.
James considers translation equivalence to be the best TC for CA, provided it embraces
both semantic and pragmatic equivalence. No attempt is made, however, to relate trans-
lation equivalence as TC to other kinds of plausible TCs within the overall landscape of
contrastive linguistics. Substance as TC is also briefly mentioned in Rusiecki 1976.
References
Bolinger, D.
1966 "Transformulation: structural translation", Acta linguistica hafniensia 9:
130-144.
Beika, J. V.
1978 "Application of quantitative methods in contrastive stylistics", Prague studies
in mathematical linguistics 6 : 1 2 9 - 1 4 7 .
DuSkova, L.
1978 "The simple sentence in Czech and English", Prague studies in mathematical
linguistics 6 : 8 3 - 9 2 .
Fillmore, C.
this volume "Remarks on contrastive pragmatics".
Graczyk, I.
no date "On writing in English as a second language: verb phrase in advanced learners'
written English", part 1, unpublished MS.
Halliday, . A. K.
1970 "Language structure and language function", New horizons in linguistics,
edited by John Lyons (Harmondsworth: Penguin Books Ltd.), 1 4 0 - 1 6 5 .
Halliday, . A. K. - A. Mcintosh - P. D. Strevens
1964 The linguistic sciences and language teaching (London: Longmans).
Hjelmslev, L.
1943 Prolegomena to a theory of language [English version 1961], translated by
F. J. Whitfield (Mardison: The University of Wisconsin Press).
Ivir, V.
1969 "Contrasting via translation", YSCECP, B. Studies 1 : 1 3 - 2 5 .
1970 "Remarks on contrastive analysis and translation", YSCECP, B. Studies 2:
14-25.
James, C.
1980 Contrastive analysis (Harlow, Essex: Longman).
Kopczynski, A.
1980 Conference interpreting: some linguistic and communicative problems
(Poznan: Adam Mickiewicz University).
Krzeszowski, T. P.
1974 Contrastive generative grammar: theoretical foundations (Lodz: Uniwersytet
Lodzki, reprinted in 1979, Tbingen: Gunther Narr Verlag).
1976 "On some linguistic limitations of classical contrastive analyses",Papers and
studies in contrastive linguistics 4 : 8 8 - 9 5 .
1981a "The problem of equivalence revisited", IRAL 19.2:113-128.
1981b "Quantitative contrastive analysis", Studia Linguistica 3 5 : 1 - 2 , 1 0 2 - 1 1 3 .
Lado, R.
1957 Linguistics across cultures (Ann Arbor: University of Michigan Press).
Lipinska-Grzegorek, M.
1977 Some problems of contrastive analysis: sentences with nouns and verbs of
sensual perception in English and Polish (Edmonton, Canada: Linguistic
Research, Inc.).
Liston, J. L.
1970 "Formal and semantic considerations in contrastive analysis", YSCECP,
B. Studies 2: 2 7 - 4 9 .
Oleksy, W.
this volume "Towards pragmatic contrastive analysis".
Politzer, R. L.
1968 "An experiment in the presentation of parallel and contrasting structures",
Language learning 1 8 : 3 5 - 4 3 .
Riilker, K.
1973 "Zur pragmatischen Invarianz bei der Translation", Neue Beitrge zu Grund-
fragen der bersetzungswissenschaft, edited by . Neubert and . Kade
(Leipzig: Athenum Verlag), 2 9 - 3 5 .
Rusiecki, J.
1976 "The development of contrastive linguistics", Interlanguage studies bulletin
1:12-44.
Spalatin, L.
1969 "Approach to contrastive analysis", YSCECP, . Studies 1:2635.
Stockwell, R. P. - S. D. Bowen - J. W. Martin
1965 The grammatical structures of English and Spanish (= Contrastive structure
series) (Chicago and London: The University of Chicago Press).
WOLFGANG KHLWEIN
Pedagogical limitations of contrastive linguistics
1 General considerations
Ever since languages have been described and compared in the course of the
history of language studies some underlying applied motives can be traced.
Philosophical interests were behind the dispute between analogists and
anomalists in ancient Greece: if language could be proved to be ordered ac-
cording to a systematic plan, this insight could be taken as evidence for equal-
ly systematic principles underlying reality in its entirety - language being part
of reality, its investigation was supposed to serve as a key to the whole struc-
ture (or non-structure) of the world as cosmos vs chaos. Central issues of this
philosophically oriented kind of language study like the relationship between
form and meaning, topic and comment, syntactic structure etc. have made
their way into 20th century descriptive linguistics. It seems remarkable, how-
ever, that the gap between this kind of philosophical application and the
rhetorical and pedagogical applications underlying language studies of the
Alexandrian school had never been properly bridged. Theoretical application
as opposed to pedagogical applications of language study?
Similarly Panini's mathematical and statistical observations in phonology
and morphology had remained rather normative and descriptive than pedagog-
ical in spite of their ultimate applied aim of enabling his contemporaries to
acquire the practical skills to cope with the older Veda texts.
More recently the Neogrammarians' aim of yielding access to older stages
of European languages resulted in comparable descriptions of these stages,
lacking, however, systematic strategies of linguistic comparison in general, and
pedagogical strategies concerning e.g. reading-comprehension skills in particular.
Likewise, the comparative potential of the 17th century attempts at esta-
blishing universal grids as a mentalist basis for the descriptions of language(s)
had never been fully exploited methodically from a contrastive point of view
nor pedagogically from an applied perspective.
Obviously the gaps between theoretical study of language comparison and
language pedagogy, so frequently complained at in our days, are not that new
314 Wolfgang Khlwein
after all (cf. Khlwein 1980:761f.). Maybe, however, contrastive linguistics

has made us more conscious of them, e.g. by distinctions like the meanwhile
well-established one drawn by Fisiak (1973:8) among four kinds of contra-
stive studies:
general theoretical vs.
specific theoretical vs.
general applied vs.
specific applied contrastive linguistic studies.
It is true, such categorizations allow to pigeonhole specific contrastive studies
nicely, but it still is a long way from e.g. general theoretical to specific applied.
For a principal early objection cf. Lee (1971:20) who held that an exhaus-
tive comparison of two languages might be altogether impossible "because
there is no acceptable model of language which takes adequate account of the
fact that various features and characteristics of a language are closely and
meaningfully associated with extra-linguistic reality".
In the same year James summed up nine arguments often brought forth
from a pedagogical point of view against contrastive linguistics:
(1) Interference from the I-! [source language] is not the sole source of er-
rors in L 2 [target language] learning. There are other sources which CA
[contrastive analysis] fails to predict. Even the unsophisticated teacher
who knows no linguistics is conscious of more errors than CA can predict.
(2) The predictions of student errors in made by CA are not reliable.
(3) CA is based on, and perpetuates, a naive view of language structure.
(4) There are no established criteria for comparability.
(5) CA endorses a teacher-centred rather than a learner-centred approach to
foreign language learning.
(6) CA only conceives of interference in one direction, from L, to L 2 .
(7) We expect the strongest habits to exert most interference, so why is it
that the weaker L j habits interfere more with L 2 [third language] than
Li habits?
(8) The degree of typological difference between Li and L 2 is not propor-
tional to the interference strength.
(9) Interference is an otiose idea: ignorance is the real cause of error.
Though James vindicated contrastive linguistics in all nine points, refuting
some of them explicitly, they have stayed alive in many ensuing discussions.
In 1972 Marton sets out his article on "Pedagogical implications of con-
trastive studies" with the statement (1972:115): "Current views on pedagog-
ical applications of contrastive studies are marked by disappointment and pes-
simism", owing to the fragmentary nature of contrastive descriptions and to
the fact that error prediction of contrastive linguistics is only rather limited.
Starting out from the same observation, found after investigating syntactic
errors made by foreign language learners and comparing them to hierarchies of
difficulty established on the basis of contrastive linguistics Whitmann and Jack-
son came to the harsh conclusion: "Contrastive analysis [ . . . ] is inadequate
theoretically and practically, to predict the interference problems of a lan-
guage learner" because "Interference, or native-to-target language transfer,
plays such a small role in language learning performance that no contrastive
analysis, no matter how well conceived, could correlate highly with perform-
ance data, at least on the level of syntax" (1972:40).
Also in 1972 Richards provided further evidence for this rather pessimistic
outlook on the state of pedagogical limitations of contrastive linguistics of
that epoch when he compared some essential factors which set apart the
learning of e.g. English as a foreign language from the study of immigrant
varieties of English, indigenous-minority varieties of English, pidginization,
creolization, and local varieties. As within the latter varieties of English limi-
tations are supposed to be rather socially imposed than individually, what
may be regarded as errors from the point of view of a foreign language learn-
ing situation should rather be seen as deviance within the latter social contexts.
Thus the frequently claimed possibility of transferring pedagogical contrastive
insight from the foreign language learning situation to these other sociolin-
guistic situations of linguistic variety should be handled with care. If at all, ac-
cording to this study contrastive linguistics can have a bearing on the 'normal'
foreign language teaching situation only, where interlingual processes do not
have a specific social role but are considered as indicators of partial learning,
in other words, where more is depending on the construction of teaching
materials: "The study of interlingual phenomena in language learning thus
leads to a focussing on the central processes of second language acquisition,
and to the study of the circumstances which give these processes significance"
(Richards 1972:185).
Despite the role conceded to contrastive linguistics in the foreign language
teaching situation, the emphasis on the part played by non-linguistic 'circum-
stances' such as individual differences in perseverance, motivation, aptitude
etc. becomes obvious.
The respective differences which result from the settings of different learn-
ing situations also bear on the pedagogical use of the interlanguage hypothesis
which is frequently claimed to be the hinge of pedagogical contrastive gram-
mars. As recently as in 1980 John strongly advocates the recognition of
f i x e d 'approximative languages' in relation to learning situations, of
certain barriers which the learner possibly never surmounts in the

learning process. Often, these barriers (caused by interference
from the other langiiage/s) are found not to be idiosyncratic with

the individual learner, but as characterising a whole community
or type of learners sharing a given first language (L x ) and learning
a second language (L 2 ) in a given language learning situation (John
1980:210).
The following examples from European respectively American contexts might

serve to illustrate John's respective classification of the role of language milieu
situations in language contact, and should consequently be kept apart from
one another:
(1) e.g. the adult guest worker who has a fully developed L t and acquires
L 2 in a socioculturally 'natural' learning situation
(2a) where e.g. English or French are learnt as a lingua franca or for special
purposes (e.g. French used in mass media in the German speaking areas
of Eastern Luxembourg or of Eastern Belgium) the learner does not
step outside his L t language milieu and thus acquires his L 2 in a situa-
tion socioculturally more or less alien to the L 2
(2b) the learning of e.g. English as a second language for educational pur-
poses in most European schools basically falls into the same language
milieu situation; here, however, deviations are rather regarded as errors,
which are less socially determined, often due to insufficient teaching,
lack of motivation, etc.
(3) e.g. in groups of second generation migrants the milieu is that of L 2 : as
a natural consequence sociocultural L 2 -features enter into Lx -acquisition
(4) e.g. immigrants in French speaking parts of Canada who come from
neither French nor English speaking countries, who have to acquire
English (e.g. for business transactions) in a French speaking milieu, i.e.
in a milieu which is neither equivalent to that of their own Li nor to
that of L 2 (English).
Experience proves John's assumption, that it is very likely in situations (1)
and (2 a) a foreign language will be developed in which many foreign language
structures tend to be fossilized at an approximate stage - which is accepted
and up to a certain degree even expected by the environment. In situation (3)
fossilized source language stages actually result and tend to impede source
language behaviour (cf. Khlwein 1981; de Matteis 1981) - and simultaneous-
ly impede contact with the source language native speaker, e.g. in our ex-
ample with their parents. In situation (4) the foreign language spoken in the
environment is itself unlike the foreign language as spoken in its own environ-
ment, in a certain sense it might itself be regarded as an approximative lan-
guage; children of people in this situation often suffer the same effect con-
cerning the native language of their parents as in situation (3).
Presumably it is situation (2 b) where pedagogical grammars on a contrastive

basis will have their strongest impact.
What then is the main aim of pedagogical grammars? Do present pedagog-
ically oriented contrastive studies come up to the respective expectations?
Up to the seventies many so-called pedagogical grammars could be distin-
guished from linguistic grammars merely on the basis of their selective prin-
ciples. To a certain extent this was justified by the general educational out-
look which reduced school teaching to the format of condensed, abridged
university teaching. As a consequence curricular planning was almost solely
subject matter oriented not only in the teaching of languages. Under these
auspices all that was expected from a pedagogical grammar was the selection
of linguistic items to be taught, at best in addition a discussion of the most
appropriate methodological form which a pedagogical grammar should apply
for the purpose of describing linguistic statements; but: the optimal linguistic
system for describing a language need not necessarily be the 'best' system for
learning a language. As late as in 1974 a shift towards the "processes and pro-
cedures whereby the facts of a language can be made available to the learner"
was advocated by Corder (1974:168), who defines: "Such an organisation and
presentation of the linguist's facts [i.e. a form which does make them acces-
sible] is called a pedagogic grammar" (1974:167). In 1979 Engels is following
up more precisely: "The final goal of every pedagogical grammar must be to
enable students to make a direct transition (without detours) from a func-
tional grammatical insight to its application in language production"(l 979:17).
As for the second question raised above even as late as in 1979 the same
author arrives at as negative an evaluation as:
Even the contrastive grammars i.e., modern, scientifically defend-

able, up-to-date comparative drafts between the mother-tongue
and a foreign language, regardless which tendency they represent,
structural or transformational, [ . . . ] are deficient from the ped-
agogical point of view (1979:14f.).
Though this statement makes clear that an applied contrastive grammar is not
yet per se a pedagogical grammar it nevertheless follows from Engels that a
pedagogical grammar must necessarily have an applied contrastive component:
Yet they [i.e. contrastive grammars] remain a kind of application

of linguistics in the literal significance of the word, although in a
contrastive grammar we are already dealing with a two-fold gram-
mar: the mother-tongue versus a foreign language and vice versa,
so that, thanks to the conscious insight into the differences, the
mother-tongue no longer hinders the learning processes(1979:15).
Where then are its deficiencies, its limitations? This would certainly be the
wrong place to give a specific review of all existing contrastive projects,
scrutinizing their individual pedagogical potentials. Nor can an encompassing
discussion of contrastive linguistics within the framework theoretical vs ap-
plied linguistics be expected here. Nor will we be heading towards a (further)
idealized model to determine the place value of contrastive analysis within the
pedagogical grammar. What could, however, be attempted, is the pinpointing
of some crucial relationships between (foreign) language learning and teaching
and contrastive analysis as viewed from some current linguistic trends, and as
they might already have become obvious from the preceding survey.
2 Specific considerations
2.0 The following four sections will concentrate on the four areas outlined
in Fig. 1.
Foreign language
learning and Linguistics
Contrastive analysis
Foreign language learning and teaching
Linguistics
Contrastive analysis
1. (traditional) theoretical and applied contrastive linguistics

2. <"> teacher orientation of contrastive linguistics
3. learner orientation of contrastive linguistics
4. ^ socio-cultural dimension of contrastive analysis
Fig. 1
2.1 As the relationship between sectors 14 in Fig. 1 is not an arithmetic

one, the requirements set by sectors 24 will have a backlash effect on the
format of 'traditional genuine contrastive linguistics'. If contrastive linguistics
is carried out autonomously compatability with 24 remains a random affair
- a procedure which would at least, to say it mildly, be uneconomical. Con-
trastive studies that do not set out from encompassing theories concerning
24, but from 'pure' linguistic theories will have a correspondingly limited
applicability. In any case, theoretical contrastive linguistic deficiencies will
have their bearing right through all other sectors, imposing corresponding
limitations upon them.
Perhaps Helbig put it too harshly when stating that confrontative studies
are "vielfach nichts anderes als die Summierung von mindestens zwei Einzel-
beschreibungen, aus der dann einige konfrontative bzw. kontrastive Schlu-
folgerungen abgeleitet werden" (1976:12), because no adequate descriptions
of specific languages are available, but it certainly is a prerequisite to have
specific descriptions which are comparable from their theoretical, methodical
and terminological designs.
Various contrastive studies still vary not only as regards the linguistic
models underlying their description of source language and target language
but also with regard to the use of different models applied to the description
of one language. There have been frequent and long arguments comparing the
efficiency of linguistic models for contrastive studies, e.g.
- favouring transformational grammar and using deep structure as a tertium
comparationis (Vater 1973:9); cf. also at a rather early stage Hllen: "Da-
mit hier sauber unterschieden und gegeneinander abgewogen werden kann,
ist ein fr beide Sprachen gltiges, verlliches Beschreibungs- und Kate-
goriensystem notwendig" (1969:597), and Marton (1972:119);cf. also to
a certain extent Zabrocki, emphasizing the datasupplying aspect of con-
trastive linguistics at the cost of creating any explanatory theory (1976).
- against transformational generative models, cf. e.g. Hartmann 1977, when
emphasizing contrastive stylistics and contrastive research into problems of
register; similarly Hartmann 1981; cf. also Chesterman, setting out from
language as capacity, as process, and as product, and attributing to trans-
formational generative grammar solely the latter aspect, and envisaging
'formidable' problems if transformational grammar were to be interpreted
from the perspectives of capacity and process, referring to its 'psycholin-
guistic fallacy' (1980).
- favouring a (generative) semantic basis, cf. Krzeszowski 1974;Markkanen,
pointing out, however, the respective difficulties for application (1979).
- favouring corpus-bound procedures or at least their connection with non-
corpus-bound ones, in any case their connection with error analysis; e.g.
Filipovic 1972/73 and 1974, and Spillner 1972:32f. when introducing

quantitative data from comic strips to contrastive linguistics.
- favouring case grammar respectively valence grammar based approaches,
cf. e.g. Di Pietro (1976) and Fink (1981), respectively.
postulating models that can cope with the description of both, verbal and
non-verbal behaviour, e.g. Barrera-Vidal - Khlwein (1975:26ff.) pointing
out to the potential of tagmemics; and Spillner when widening the scope
of contrastive linguistics to contrastive area studies (1978:160).
favouring functional comparisons on the basis of notional or referential
equivalence (Ebneter 1976), distinguishing, however, concepts or referents
which are language particular from those that are not (Esser 1980).
Not only for contrastive studies following the last mentioned approach the
notion of equivalence is crucial. Like the use of models it has often been dis-
cussed but never really been solved, partially perhaps because the most cur-
rently used criteria for equivalence are in turn determined by specific linguistic
theories.
- Setting out from Marton 1968:55 e.g. Vater (1973:9) suggested a rather
formal view:
Eine sprachliche Konstruktion (Satz, Phrase, Wort) ist dann einer

Konstruktion in einer anderen Sprache quivalent, wenn sie
a. in der gleichen Beziehung zur nchsthheren Konstruktion
steht,
b. wenn innerhalb der Konstruktion die gleichen Beziehungen
herrschen.
Both conditions, a and b, have to be met.

- Setting out from Krzeszowski 1974 e.g. Jackson (1976) and Markkanen
(1979) emphasize the importance of equivalent semantic inputs along with
structural similarity. The problems springing from the requirements of an
underlying universal semantic grid as a tertium comparationis remain.
Two frequently used operative techniques are:
(a) paraphrase equivalence: two constructions are said to be equivalent if
they can be paraphrased alike; to determine the value of the para-
phrase, however, entails the necessity of another paraphrase; we are
left with a regressum ad infinitum of paraphrasing (cf. Hllen 1973).
(b) translational equivalence (Catford 1965); "sentences that can be con-
sidered translations of each other by a competent bilingual" (Mark-
kanen 1979:23). As early as in 1975 Schwarze pointed out the funda-
mental limitation imposed on translational equivalence by truth con-
ditions:
Da die quivalenz unter Bezugnahme auf Wahrheitsbedingungen

definiert ist und da [ . . . ] nur Aussagestze wahr oder falsch sein
knnen, ist das Kriterium ['bersetzungsquivalenz'] nur auf Aus-
sagestze anzuwenden (1975:18).
Apart from the fact that both attempts do not constitute theories of equi-
valence but rather discovery techniques of equivalence the perhaps even
more important problem of the scope of equivalence remains leading to
functional and communicative criteria:
A direct approach to communicative equivalence is extremely difficult as
we have to recur to mental processes of cognition, which in turn can only
be achieved via text analysis which in turn should elucidate the underlying
features of functional equivalence. Functional equivalence includes signifi-
cation (actualized semantic meanings of a text), intralingual paradigmatic
meaning, and the communicatively relevant properties of actual textual
structure (Jger 1973:65ff.). Furthermore associative and connotative
components of meaning will have to be taken into consideration. To which
extent they can be accounted for within a (hitherto not yet accessible)
framework of 'Verstehensuniversalien' (Jger 1973:66) remains an open
question.
It is true, functional equivalence is a necessary prerequisite for communica-
tive equivalence, but it is not the sufficient prerequisite. There is a consider-
able amount of interlingual variation as to what the structures of a language
determine to be expressed (what must be expressed and what may be ex-
pressed), e.g. the distinction Russ. sinij/goluboj that may, but need not be
expressed in German or English: hellblau/dunkelblau respectively light blue/
dark blue, but, if desired, just blau respectively blue. How these differences in
scope of linguistic freedom are made use of is largely a matter of sociocultural
conventions, of the culture- and language-specific interplay of system and
norm, of frequency (e.g. the use of the passive voice in English vs German),
distribution, intralingual analogies. It is here, where contrastive linguistics
opens up to cross-cultural contact analysis and it is here, where Krzeszows-
ki's postulate of supplementing qualitative contrastive analyses by quantita-
tive ones ties in (cf. Krzeszowski in this volume). When comparing the 'Pre-
sident de la Republique' of the 5th French Republic to the Federal German
'Bundesprsident' Spillner (1978:159) ends up with the postulate of a trans-
cultural deep structure in order to avoid the tricky problem of tertium com-
parationis:
322 Wolfgang Kuhlwein
Transcultural
transcultural
deep structure
country A country
culture culture y
L2
Fig. 2
But is this transcultural deep structure not just another kind of tertium com-
parationis - which we do not happen to have ( - yet)?
2.2 It is obvious that these theoretical problems concerning the overlap lin-
guistics/contrastive analysis bear most heavily on language teachers' potential
use of contrastive studies. Pedagogical applications can never make up for
theoretical deficits. Within the respective scale of estimation of contrastive
linguistics ranging from 'pedagogical panacea' to 'pedagogical irrelevance'
Helbig (1976:15) adopts a fairly realistic position:
Er [the contrastive linguist] kann weder die Beschreibung der Ein-

zelsprachen (vor ihm) noch die fremdsprachenmethodische Auf-
bereitung (nach ihm) ersetzen; er kann weder die Auswahl und
Abfolge des Sprachstoffes allein determinieren (ohne didaktisch-
methodische Faktoren) noch die psychologischen Gegebenheiten
ausschalten.
On the other hand though theoretical contrastive linguistics often seems to

lend itself more directly to language typology, to the investigation of linguistic
interference, to the theory of translation (cf. Lee 1972; Jackson 1976:6, 21;
Khlwein Wilss 1981) than to foreign language teaching it seems to be
equally obvious that intralingual descriptions plus psychology of learning plus
foreign language teaching methodology in turn cannot replace contrastive
analysis.
For the teacher contrastive analysis is one of his necessary discovery pro-
cedures for finding out on which target language phenomena he should con-
centrate. In this function pedagogical contrastive linguistics will hardly claim
to be new. Properly supplemented by error analysis and tests it will refrain
from too ambitious predictive claims. Actually the teacher will usually follow
a reverse procedure: setting out from a fairly large corpus of errors, linguistic
error analysis will set apart psychologically, motivationally etc. determined
errors from linguistically determined ones; among the latter ones contrastive
analysis serves as a tool to separate intralingual ones from those which are or
at least could be accounted for by interlingual factors. Practical experience
indicates that in many cases it is hard, if not impossible, to draw the distinc-
tion between 'are' and 'could be' (cf. Khlwein 1972). Thus, all contrastive
linguistics can do is to indicate where interlingually determined errors might
occur. Consequently pedagogical grammars on a contrastive basis should not
be taken as the core of course design, but will rather sharpen the teacher's eye
for diagnosis. After all as early as in 1972 Marton had gathered the experi-
mental evidence against Lado's early assumption 'similar = easy' (1972:122;
cf. also Kruppa 1975:98), as it had been reinforced e.g. by Upshur:
Contrastive analyses can identify the target language habits for

which learning will be unnecessary or will be facilitated by posi-
tive transfer and those for which learning will be interfered with
because of negative transfer (1962:124).
It can hardly be explained with contrastive linguistic means that many Dutch
pupils commit more errors in those cases where Dutch hebben and zijn cor-
respond to French avoir and 2tre than in those cases where they don't.
Within the limited use of contrastive analysis for diagnostic purposes con-
cerning errors which have potential interlingual origins, the use of contrastive
linguistic procedures can, however, be extended into the didactic area lying
between diagnosis and therapy by rendering clear devices for the framing and
patterning of exercises and drills. The role which can be played by contrastive
linguistics in the final stage, in therapy, seems to be more controversial. As a
consequence, stage-models (like Vilke's proceeding from mechanist to cogni-
tive drills; cf. Filipovic 1974:11 f.) or spiral models (like Lee's gradual widen-
ing of the range of circumstances of the use of the item to be learnt; cf. also
Lee 1969:58f.; cf. also Dirven et al. 1976:138ff.) usually are preferred to
all-or-none procedures. - Whether, or rather, to which extent contrastive
linguistics should serve as a presentation technique (Darstellungsmethode;
Helbig 1976:10) in foreign language teaching methodology mainly depends on
non-linguistic factors, as mentioned above, including motivation, age,attitude,
aptitude of the learners. More than for many other applied linguistic disci-
plines Pike's statement is valid for contrastive linguistics: linguistics is not so
much concerned with what is being served by the waiter but with what is go-
ing on in the kitchen! Parallel to the extent, however, to which the pendulum
in language pedagogy is swinging towards more cognitive procedures, the im-
mediate use made of contrastive linguistics in foreign language teaching
methodology is increasing. - The respective most obvious deficits still seem

to exist in contrastive lexicology and contrastive text linguistics (cf. also Roos
1975:89ff.; Hartmann 1977 and 1981). - A further deficit is due to the illu-
sion that pedagogical contrastive grammars can rely on such simple a basis as
target language- vs source language-standard. Apart from source language
standard, (a) the respective stage of the acquisition of source language-standard
in the learner and (b) the respective dialectal/sociolectal source language vari-
ety which is most familiar to the learner (Karpf - Kettemann - Viereck 1980)
have to be considered, as in many instances source language standard merely
is a hypothetical construct for the learner. Whereas the problems relating to
'standards in contact' have secured the interest of foreign language teachers
via contrastive linguistics, the problems resulting from Variants in contact'
have still remained a fairly secluded domain of sociologists and sociolinguists,
respectively, mainly occupied with either source language or target language
solely, (a) leads to the postulate of a stronger recognition of psycholinguistic
factors, (b) to a wider recognition of sociolinguistic and socio-cultural factors,
respectively, in pedagogically oriented contrastive linguistics.
2.3 The more linguistic theory has opened towards contact and context fea-
tures determining language use, the more emphatically psycholinguists point
out that most contrastive linguistic studies are still being carried out without
a corresponding contact with the reality of language use and language acquisi-
tion, that contrastive linguistic research is carried out in vacuo, in abstracto
(cf. Slama-Cazacu 1971:190). Basically this objection seems to boil down to
the problem of psychological reality of the linguistic theories used by con-
trastive linguistics. Thus it is rather directed against linguistics in general than
against contrastive linguistics in particular. Whether, however, any study of
language should be carried out within an encompassing psycholinguistic frame-
work can neither be generally discussed let alone be solved here. It is true, the
more performance oriented linguistics is becoming, and the more performance
models underlie contrastive studies, the better it accounts for use and acquisi-
tion of language. At the same time the deplorable gap between linguistic
theqry and its application(s), though frequently quoted by language teachers,
shrinks. Instead of linguistic theory psychological (and/or behavioural) theories
yield the basis for language acquisition research, within which linguistic the-
ory has to grope for its proper problem oriented place. The distinction
between theoretical and applied linguistics dwindles, and correspondingly that
between theoretical contrastive linguistics and applied (e.g. pedagogical) con-
trastive linguistics. Such a view, close as it is to the requirements of social
reality, sounds desirable. It will certainly not dismiss applied contrastive
achievements as being wrong, e.g. the interlanguage hypothesis respectively
the concept of approximative systems (Nemser), compromise systems (Fili-

povic), Interimsgrammatiken (Khlwein) but it has to reinterpret them from
a psycholinguistic perspective. This perspective originates from a conceptual
rather than from a formal/functional grammatical basis. We certainly need
not go all the way with Sapir, Whorf, Weisgerber ('Sprachinhaltsforschung',
'sprachliche Zwischenwelt') towards linguistic relativity (cf. Dirven 1976); the
conceptual strategies ('das Formen von 'Begriffsinhalten"), which actively
and passively interrelate with the correspondingly different structures in dif-
ferent languages, would, however, have to be taken as the starting points of
research into language contact. But as far as the painstaking problem of
tertium comparationis is concerned, this view does not solve but only shifts
the problem; we do not (yet) have the necessary underlying universal concep-
tual grid either. The repeated onomasiological attempts within the history of
language study tell us a sad story. Further developments of integrational ap-
proaches as e.g. outlined by Lieb (1980:5ff.) will, however, have t o be fol-
lowed alertly.
The placing of the user respectively learner into the centre of the stage, the
ensuing bridging of the gap between theoretical linguistics and applied lingu-
istics does, however, not do away with another pedagogically important gap:
the one existing between a linguistically or psycholinguistically based descrip-
tion of language and language teaching methodology. Nevertheless, it will at
least render some essential clues. We stated above that the implications of
contrastive analysis for the methodology of foreign language teaching have
remained open to debate. If we envisage a contrastive description as based on
perceptual strategies which result in language specific concept formation, and
as using a (not yet developed) adequate metalanguage, the resulting contrastive
pedagogical 'grammar' will lend itself strongly for a rather cognitive approach
to foreign language learning. Man masters his environment symbolically by
language, by performing mathematical operations, by interpreting traffic
signs, . . . everyday, everywhere. This does not, however, imply that the
strategies used for the acquisition of the connection between concept or
extra-linguistic event and its linguistic symbol should be equally symbolic,
e.g. rendered as a generative-transformational rule on the blackboard in the
classroom. For learning purposes the formation of this connection can be
achieved on a lower semiotic level, e.g. in highly indexalic or iconic ways
which can lead to a more straightforward production of language. Though
not yet fully systematized the type of pedagogical grammar as suggested by
Engels seems to provide the best example:
The syntactic rules are compressed in graphic mediators symboliz-

ing the rule, once insight is acquired (static function), which after-
wards function as mediating stimuli in the process of internaliza-

tion (dynamic function) (1979:25).
The cognitive element of this strategy seems to rest on two levels:
- There must be a basic attitude on the learner's side to sense language spe-
cific (contrastive) diversity in perceptual strategies of coping with the
world, and to sense corresponding diversity in concept and rule formation;
cognition will also play a part in acquiring the meaning of the iconic, sym-
bolic or indexalic mediator ('insight'!), before it comes to be reduced to a
pure signal which directly triggers language production in the learner and
(next and last step) before it will become internalized, thus rendering its
materialiter presentation superfluous. If foreign language acquisition will
work along these two steps, the amount of cognitive language comparison
involved may well be taken into the bargain.
2.4 Contrasting languages on the basis of language specific perceptual and

conceptual strategies can, however, not end up in cognitive psychology but
must ultimately go as far beyond the realm of psycholinguistics as to the re-
spective determining forces which reveal different languages to be different
emanations of different cultures and habits of the respective language com-
munities. This way carries contrastive analyses from contrastive linguistics via
contrastive psycholinguistics to contrastive sociolinguistics.
It is well worth discussing whether the mediating link 'contrastive psycho-
linguistics' is necessary at all. Why not proceed from the interrelation of lan-
guage and society to the (contrastive) description of languages, from what we
are doing to what we are saying, directly (e.g. Halliday 1978)? I should like
to suggest that both positions should not be seen as a mutually exclusive alter-
native, because what can be done in a society can also be accounted for in
intraorganic terms of a psycholinguistic competence of what can be meant
and even of what can be said. 'Doing meaning saying' certainly is no one-
way road. Envisaging the abovementioned difficulties in contrasting what we
say (Halliday's lexicogrammatical potential), however, the attempt to derive
it from the pragmatic or behavioural potential of what we can do, certainly is
worthwhile being elaborated.
As for the tertium comparationis contrastive analysis under these auspices
has to set out from a satisfactory definition of culture. Necessarily it will be
rather broad, e.g. culture as seen as "Gesamtheit von Problemlsungsentwr-
fen" (Ghring 1980:75):
Kultur ist all das, was das Individuum wissen und empfinden kn-
nen mu,
1) damit es beurteilen kann, wo sich Einheimische in ihren ver-

schiedenen Rollen so verhalten, wie man es von ihnen erwartet
(Erwartungskonformitt), und wo sie von den Erwartungen ab-
weichen;
2) damit es sich in Rollen der Zielgesellschaft, die ihm offen ste-
hen, erwartungskonform verhalten kann, sofern es dies will
und nicht etwa bereit ist, die Konsequenzen aus erwartungs-
widrigem Verhalten zu tragen;
3) zur Kultur gehrt auch all das, was das Individuum wissen und
empfinden knnen mu, damit es die natrliche und die von
Menschen geprgte oder geschaffene Welt wie ein Einheimischer
wahrnehmen kann (Ghring 1980:73f.).
Perhaps the implicit pedagogical consequence that foreign language teaching

should turn into the teaching of foreign behaviour (based on contrastive cul-
tural analyses on the teacher's side) is too strong and can be mitigated, but a
contrastive approach on this basis will nevertheless extend far beyond the lin-
guistic teaching objective 'communicative competence,' and it will meet the
same sociological difficulties which the latter meets in a socioloinguistic dis-
guise, as had been pointed out by Janicki's criticism concerning contrastive
sociolinguistics as done within the communicative competence perspective:
"even if the language varieties of reference were defined intuitively, it would
not be possible to state that the two varieties compared are comparable in
sociolinguistic terms" (1979:35).
E.g. the Danish-English pedagogical contrastive project draws the necessary
consequence by widening contrastive linguistic performance analysis to con-
trastive communicative performance analysis, encompassing the learner's com-
municative needs as well as the effect on his interlocutors (Faerch 1979:17).
That this procedure might ultimately yield a description of communicative
competence in contact(cf. Clyne 1979:17ff.) need not be excluded. Actually,
the postulate of contrastive sociosemiotic analyses ties in with the old and
frequently uttered wish - often produced as an argument against older stages
of contrastive linguistics not to separate the learning of language form from
the learning of language use (cf. Lee 1969:57; cf. also Dimitrijevic's (1977:
14If.) complaint concerning the lack of contrastive cultural analyses, though,
perhaps linked too exclusively with the - equally deficient sphere of con-
trastive lexicology), from the strategies adopted for successful role-playing in
a foreign cultural and linguistic context.
As Di Pietro pointed out so clearly, it is the non-ritualized ones among
these strategies that are of extreme pedagogical significance; it is e.g. not suf-
ficient to acquire communicative competence of how to apologize in English;
to be able to grade the various forms as to their social appropriateness is of

equal importance (Di Pietro 1976:4). Admittedly there are very down-to-
earth constraints as to how far one can go within specific language courses.
The Un niveau seuil of the Council of Europe might be regarded as a minimum,
a contrastive Leech - Svartvik communicative grammar, based on contrastive
cross-cultural analyses as a maximum - ?
One obstacle might consist in the traditional allocation of applied contrastive
studies to foreign language teaching and of applied sociolinguistic studies to
first language teaching. A more serious one consists of the fact that as far as
cultural phenomena ('Landes- und Kulturkunde') had come to the attention
of linguists the latter tried to cope with them with the conceptual and metho-
logical tools provided by theories which have been designed to describe the
linguistic part of human behaviour only (cf. Spillner 1978:153 ff.) instead of
trying it the other way round.
References
Barrera-Vidal, A. - W. Khlwein
1975 Angewandte Linguistik fr den fremdsprachlichen Unterricht (Dortmund:
Lensing).
Catford, J. C.
1965 A linguistic theory of translation (London: Oxford University Press).
Chesterman, A.
1980 "Contrastive generative grammar and the psycholinguistic fallacy", Papers
and studies in contrastive linguistics 1 1 : 1 7 - 2 4 .
Clyne, M.
1979 "Communicative competences in contact", ITL 4 3 : 1 7 - 3 8 .
Corder, S. P.
1974 "Pedagogical grammars or the pedagogy of grammars?", Linguistic insights in
applied linguistics, edited by S. P. Corder and E. Roulet (Brussels, Paris:
AIMAV), 1 6 7 - 1 7 3 .
Coste, D., et al.
1976 Un niveau seuil (Strasbourg: Council of Europe).
Dimitrijevic, N. R.
1977 "Problems and implications of contrastive analysis of vocabulary and culture",
Papers and studies in contrastive linguistics 7 : 1 3 3 - 1 4 4 .
Di Pietro, R. J.
1971 Language structures in contrast (Rowley, Mass.: Newbury House).
1976 "Contrasting patterns of language use: a conversational approach", Canadian
modern language review 3 3 : 4 9 - 6 1 .
Dirven, R.
1976 "A redefinition of contrastive linguistics", IRAL 1 4 : 1 - 1 3 .
Dirven, R. - W. Hnig - W. Khlwein - G. Radden - J. Strau
1976 Die Leistung der Linguistik fr den Englischunterricht (Tbingen: Niemeyer).
Ebneter, T.
1976 Angewandte Linguistik 2 (Mnchen: Fink).
Engels, L. K.
1979 "Pedagogicalgrammars", Glottodidactica 1 2 : 1 3 - 3 4 .
Esser, J.
1980 "Contrastive analysis at the crossroads of linguistics and foreign language
teaching", IRAL 1 8 : 1 8 1 - 1 9 2 .
Faerch, C.
1979 Research in foreign language pedagogy - the PIFProject (Copenhagen: Uni-
versity of Copenhagen).
Filipovic, R.
1972-1973 "The use of a corpus in contrastive linguistics", Studio Romanica et
Anglica Zagrebiensia 3 3 - 3 6 : 4 8 9 - 5 0 0 .
1974 "The use of contrastive and error analysis to practising teachers", YSCECP-
pedagogical materials 2, edited by R. Filipovic (Zagreb: University of Zagreb),
3-17.
Fink, S.
1981 "Some thoughts on international standardization of linguistic terminology in
contrastive studies", Festschrift in honor ofR. DiPietro, edited by E. Danesi
(Newark: University of Delaware Press).
Fisiak, J.
1973 "The Polish-English contrastive project", Papers and studies in contrastive
linguistics 1 : 7 - 1 0 .
Ghring, .
1980 "Deutsch als Fremdsprache und interkulturelle Kommunikation", Fremd-
sprache Deutsch 1, edited by A. Wierlacher (Heidelberg: Groos), 7 0 - 9 0 .
Halliday, . . K.
1978 Language as social semiotic. The social interpretation of language and mean-
ing (London: Arnold).
Hartmann, R. R. K.
1977 "Contrastive textology in descriptive and applied linguistics" (Ms.).
1981 "Contrastive textology and translation", Kontrastive Linguistik und berset-
zungswissenschaft. Akten des Internationalen Kolloquiums Trier/Saarbrcken
Sept. 2 5 - 3 0 , 1978, edited by W. Khlwein, W. Wilss and G. Thome (Mn-
chen: Fink), 2 0 0 - 2 0 8 .
Heibig, G.
1976 "Zur Rolle des kontrastiven Sprachvergleichs fr den Fremdsprachenunter-
richt (Mglichkeiten, Voraussetzungen, Grenzen)", Deutsch als Fremdsprache
13:9-16.
Hllen, W.
1969 "Zwanzig englische Kernstze und ihre deutschen quivalente. Versuch einer
kontrastiven Analyse", Die Neueren Sprachen 1 8 : 5 9 0 - 6 0 0 .
1973 Linguistik und Englischunterricht. Didaktische Analysen (Heidelberg: Quelle
und Meyer).
Jackson, H.
1976 "Contrastive linguistics - What is it?", ITL 3 2 : 1 - 3 2 .
Jger, G.
1973 "Kommunikative und funktionelle quivalenz", Linguistische Arbeitsberich-
te 7 : 6 0 - 7 4 .
James, C.
1971 "The exculpation of contrastive linguistics", Papers in contrastive linguistics,
edited by G. Nickel (Cambridge: Cambridge University Press), 5 3 - 6 8 .
Janicki, K.
1979 "Contrastive sociolinguistics - Some methodological considerations", Papers
and studies in contrastive linguistics 1 0 : 3 3 - 4 0 .
John, A. P.
1980 '"Approximative languages' and language learning situations", IRAL 18:
205-216.
Karpf, A. - B. Kettemann - W. Viereck
1980 "Phonology of dialect interference in second language learning", IRAL 18:
193-208.
Kruppa, U.
1975 "Kontrastive Analyse von Interferenzerscheinungen im deutsch-englischen
Bereich", Neusprachliche Mitteilungen 2 8 : 9 2 - 9 9 .
Krzeszowski, T.
1974 Contrastive generative grammar: theoretical foundations (Lodz: University
of Lodz).
Khlwein, W.
1972 "Fehleranalyse im Bereich des englischen Vokalismus", Fehlerkunde, edited
by G. Nickel (Berlin: Cornelsen, Velhagen und Klasing), 5 1 - 6 6 .
1975 "Grundsatzfragen der kontrastiven Linguistik", Neusprachliche Mitteilungen
28:80-92.
1980 "Angewandte Linguistik", Lexikon der Germanistischen Linguistik, 2nd
completely revised and enlarged edition, edited by H.-P. Althaus et al. (T-
bingen: Niemeyer), 7 6 1 - 7 6 8 .
1981 "Soziolinguistische Aspekte der Fremdsprachvermittlung und -aneignung",
Kongrebericht der VI. Internationalen Deutschlehrertagung vom 4.-8. Au-
gust 1980 in Nrnberg,edited by . Brckner (Berlin: Langenscheidt),472-475.
Khlwein, W. - W. Wilss
1981 Kontrastive Linguistik und bersetzungswissenschaft", Kontrastive Lingui-
stik und bersetzungswissenschaft. Akten des Internationalen Kolloquiums
Trier/Saarbrcken Sept. 2 5 - 3 0 , 1978, edited by W. Khlwein, W. Wilss and
G. Thome (Mnchen: Fink), 7 - 1 7 .
Lee, W. R.
1969 "How can contrastive linguistic studies help foreign-language teaching?",
YSCECP-Studies 1, edited by R. Filipovic (Zagreb: University of Zagreb),
57-66.
1971 "Types of interference and contrasting - the kinds of research needed",
PAKS: 1 3 - 2 6 .
Lieb, H.-H.
1980 "Integrational linguistics as a basis for contrastive studies",Papers and studies
in contrastive linguistics 1 2 : 5 - 3 6 .
Markkanen, R.
1979 Tense and aspect in English and Finnish. A contrastive study (Jyvskyl:
University of Jyvskyl).
Marton, W.
1968 "Equivalence and congruence in transformational contrastive studies",
Studia Anglica Posnaniensia 1 : 5 3 - 6 2 .
1972 "Pedagogical implications of contrastive studies", Studia Anglica Posnaniensia
4:115-126.
Matteis, M. de
1981 "Zum gesprochenen Italienisch schulpflichtiger Migrantenkinder", Kongre-
bericht der VI. Internationalen Deutschlehrertagung vom 4.-8. August 1980
in Nrnberg, edited by . Brckner (Berlin: Langenscheidt), 4 8 8 - 4 9 1 .
Richards, J. C.
1972 "Social factors, interlanguage, and language learning", Language learning 22:
158-188.
Roos, .
1975 "Kontrastive Lexikologie", Contributions to applied linguistics (1), edited by
C. Gutknecht (Bern, Frankfurt/Main: Lang), 8 9 - 1 0 3 .
Schwarze, C.
1975 "Empirische Probleme des Sprachvergleichs", Linguistische Berichte 35:
10-24.
Slama-Cazacu, T.
1971 "Psycholinguistics and contrastive studies", Zagreb Conference on English
Contrastive Projects, 7-9 December, 1970, edited by R. Filipovic (Zagreb:
University of Zagreb), 1 8 8 - 2 0 6 .
Spillner, .
1972 "Kontrastive Analysen auf der Grundlage von Comic Strips", Papers from
the International Symposium on Applied Contrastive Linguistics, Stuttgart,
Oct. 11-13, 1971, edited by G. Nickel (Bielefeld: Cornelsen, Velhagen und
Klasing), 2 7 - 4 1 .
1978 "Methoden der kontrastiven Linguistik in der Frankreichkunde", Landes-
kunde und Fremdsprachenunterricht, edited by . Arndt and F. R. Weller
(Frankfurt: Diesterweg), 1 5 1 - 1 7 8 .
Upshur, J. A.
1962 "Language proficiency testing and the contrastive analysis dilemma", Lan-
guage learning 1 2 : 1 2 3 - 1 2 7 .
Vater, .
1973 "Probleme der kontrastiven Grammatik", reproduced by L.A.U.T., Series ,
No. 12 (Trier: Linguistic Agency University of Trier).
Whitman, R. L. - K. L. Jackson
1972 "The unpredictability of contrastive analysis", Language learning 2 2 : 2 9 - 4 2 .
Zabrocki, T.
1976 "On the so-called 'Theoretical Contrastive Studies'", Papers and studies in
contrastive linguistics 4 : 9 7 - 1 0 9 .
LEWIS MUKATTASH
Contrastive analysis, error analysis, and learning

difficulty*
1 Introductory
This study reports on the results of a test in English grammar that was given
to 4,835 Arab learners of English. Furthermore, it seeks to establish patterns
of difficulty in foreign language learning (English in this case), and to show
that neither contrastive analysis (CA) nor the text-bound type of error ana-
lysis (EA) can predict with any reasonable degree of adequacy the relative
degree of difficulty which foreign learners of a particular NL will encounter
in the process of learning and producing a given TL, there being a multitude
of psychological, social and educational factors that interact simultaneously
and thus affect the overall linguistic achievement of the learner.
In this study 'difficulty' is assumed to be reflected in 'error'. However, an
attempt will be made to relate difficulty to both 'linguistic difference' be-
tween NL and TL and to 'formal marking' in TL. For other interpretations of
the term 'difficulty' and for objections to equating difficulty with error, 1 see
Kellerman(1977; 1979) and the references he cites; see also Schachter(1974);
James (1977); Eckman (1977).
The present study does not purport to raise any theoretical issues and
makes no novel claims as to the psychology of foreign language learning, for
it is our contention that as Schouten (1979:4) points out, "what is wrong
with second language learning research is the fact that its mainstream is based
on theories that are supported by no facts . . . " . The only commonplace as-
sumptions that are made here are the following:
1. In learning a certain TL, some linguistic structures will be more difficult
than others for speakers of a particular NL.
2. Difficulties in foreign language learning are primarily of two types: first,
those which are typical of speakers of a given NL, and secondly, those which
are shared by all learners of a given TL regardless of their language backgrounds.
334 Lewis Mukattash
3. Although there is no one-to-one correspondence between difficulty (in the

broad sense specified above) and error, systematic errors are symptomatic of
improper learning or improper 'processing' of TL data.
4. Since little is known of 'what goes on in the learner's mind' in the process
of learning a foreign language, linguists should continue at least for some
time to come - to base their hypotheses on observations of the learner's lan-
guage behaviour.
2 Test
The test upon which this study is based was one of several test-types used to
discover and describe the prevalent grammatical errors in Jordanian English. It
consisted of 150 multiple-choice items covering eleven areas of English gram-
mar, and was based on the English language syllabus used in public secondary
schools in Jordan. All attempts were made to minimize the influence of vocab-
ulary unfamiliarity by using only words having a relatively high frequency in
the textbooks. For details about the construction of the test, and many other
aspects (e.g. pretesting, validity, reliability, etc.), the reader is referred to
Mukattash (forthcoming-b).
Before presenting the results of the test, a few points should be mentioned
in connection with the test itself:
(i) As pointed out above, the test was originally designed (together with
other test-types) to discover, describe and justify prevalent grammatical mis-
takes in Jordanian English (see Mukattash 1977, 1978, 1979, 1980, forth-
coming-a). In other words, the test was not intended to establish a hierarchy
of difficulty for Jordanian or Arab learners of English.
(ii) The limitations and restrictions imposed by the exclusive use of one test-
type should be obvious to any researcher in the field of foreign language
learning. Indeed, the conclusions and generalizations drawn from the results
of one test-type should be treated with caution.
(iii) The test does not measure the learners' production in as much as it mea-
sures their 'comprehension' and shows their ability to choose among alter-
natives, and thus it might be said to reflect their judgement about accept-
ability or grammaticalness. However, we argue below that there seems to be
a high correlation between the learner's receptive and productive competence.
To quote Chomsky(1971:37), " . . . the grammar that represents the speaker's
competence is, of course, involved in both his speaking and interpreting of
speech. . . " .
(iv) The choice of the distractors in each item was informed by the results of
previous studies in error analysis (Mukattash 1977, 1978, 1980, forthcom-
CA, error analysis and learning difficulty 335
ing-a), and thus it could be argued that the distractors biased the results, fcr
had there been different distractors, other choices would have been possible.
In spite of these shortcomings, the test has some positive aspects that
merit comment. First, the test was given to a very large number of students
and it has a very high reliability and validity, and on these grounds only it can
be expected to be informative to a very large extent. Secondly, there is a high
correlation between error-types gathered from this test and those gathered
from other test-types. Thirdly, as we shall see below, the difference in per-
formance amongst the various groups of learners who took the test is almost
consistent within each of the eleven grammatical areas that constitute it (see
Table 1, p. 337), a further indication of its validity. Finally, the problem of
avoidance (see Schachter, 1974) does not arise here, for students were forced
to choose an alternative even if they were not sure of the correct choice. As
such, we might add that information about learners' 'hunches' (see Kellerman,
1977:86) entered into the data.
3 Subjects
As pointed out above, the test was given to 4,835 students representing vari-
ous groups of learners of English in Jordan. Only five groups of learners are
relevant to this study:
I. First Group (GI): consisted of 56 fourth-year students in the Department
of English at the University of Jordan (J.U.). They had had eight years of in-
struction in English at school (56 hours a week), and four years of instruc-
tion in English (viz. practical English, language and linguistics, and literature)
at the university level.
II. Second Group (Gil): consisted of 114 final year students from 'foreign
private' secondary schools. They had begun learning English at the age of five
or six and at the time of the administration of the test they had had 1 2 - 1 3
years of instruction in English.
III. Third Group (GUI): consisted of 194 second-year English majors from
five Teachers' Training Institutes (T.T.I's). They had had eight years of in-
struction in English at school, and two years at their respective institute
(courses in practical English, linguistics and methodology). The test was ad-
ministered to them a few weeks before they graduated.
IV. Fourth Group (GIV): consisted of 1095 final year students from public
secondary schools in the city of Amman. They had had eight years of instruc-
tion in English.
V. Fifth Group (GV): consisted of 1841 final year students from public schools
in the various districts in Jordan (excluding Amman city). They had had the
same years of instruction in English and the same syllabus as GIV above.
336 Lewis Mukattash
The rationale behind choosing these five groups is as follows. First, GI and
Gil are considered to be the best in terms of language proficiency (see Mukat-
tash, forthcoming-b), and the two groups are comparable in terms of the
amount of instruction in English they had had. Secondly, GUI is meant to
show the linguistic difficulties which the English language teachers themselves
have. Thirdly, GIV and GV represent the majority of English language learners
in Jordan, and the two groups are comparable in terms of the amount of in-
struction they had had in English (8 years) and in terms of syllabus and method
of teaching.
The testees are all native speakers of Arabic, and though they speak dif-
ferent regional dialects, basically Jordanian and Palestinian Arabic, the differ-
ences between the two dialects are not sufficient on the syntax level as to
warrant differentiation between the two groups. At the time of the admin-
nistration of the test, all the testees had had a minimum of eight years of in-
struction in English.
4 Facility indices
Table 1 shows the average facility index (I.F.) of each of the eleven grammat-
ical areas for the entire population of the testees (first column) as well as for
each of the five groups listed in the preceding section. The grammatical areas
are arranged in descending order of difficulty (with respect to the 4,835 tes-
tees). 2 The table further shows the number of the items tested under each
grammatical area. (G stands for the entire population of testees, i.e. 4,835).
Before discussing the results of the test in detail, it is convenient to make a
general observation about the overall results listed in Table 1 (see the first
column). First, it will be seen that there is no appreciable difference between
the facility indices of any two adjacent grammatical areas 3 listed in Table 1:
the lowest being 0.37 and the highest 0.56. Thus if we discard a difference of
0.05 in the facility index, we can group the eleven grammatical areas into
three main groups: (i) quantifiers, adverbs, and verbals; (ii) sentence connec-
tors, nominals, prepositions, modals and relatives; (iii) adjectivals, articles and
pronouns. This seems to suggest that it is not feasible to talk about the relative
difficulty of an entire TL grammatical area (e.g. prepositions) for learners of
a particular NL. This statement does not hold true if the entire TL grammat-
ical category has no corresponding categories in L! (viz. the absence of articles
in some languages).
Returning to the 'scale' of difficulty suggested by the figures in Table 1,
one would assume that if difficulty is to be equated with linguistic difference 4
(Lado 1957), it is to be expected that this scale should be the same for the
Table 1
average I.F.
areas and
no. of items
G GI Gil GUI GIV GV
Quantifiers (12) 0.37 0.66 0.60 0.41 0.34 0.31
Adverbs (12) 0.38 0.74 0.72 0.47 0.40 0.31
Verbals (35) 0.40 0.82 0.71 0.50 0.41 0.34
Connectors (ID 0.44 0.81 0,76 0.58 0.45 0.38
Nominals (10) 0.45 0.81 0.74 0.57 0.45 0.42
Prepositions (24) 0.46 0.85 0.78 0.60 0.48 0.39
Modals (9) 0.48 0.78 0.77 . 0.60 0.50 0.42
Relatives (8) 0.48 0.88 0.83 0.62 0.52 0.41
Adjectivals (8) 0.50 0.91 0.85 0.61 0.46 0.43
Articles (15) 0.52 0.84 0.83 0.65 0.54 0.46
Pronouns (6) 0.56 0.93 0.86 0.74 0.58 0.48
various groups of testees since they share the same NL. In fact, this is not the
case. A brief look at the figures in Table 1 will convince us that different
scales of difficulty 5 emerge for the different groups of learners, for only in
the case of quantifiers and pronouns do we find agreement amongst the five
groups of learners with respect to the relative degree of difficulty. For instance,
verbals rank third on the suggested scale of difficulty in the case of three
groups, and second for one group, but sixth for another group. This suggests
that it does not seem to be possible to talk of 'scales of difficulty' for learners
of a particular TL even if they have the same linguistic background, the same
years of instruction in TL and the same syllabus (e.g. GIV and GV). This
statement is further reinforced by the fact that the scale suggested in Table 1
338 Lewis Mukattash
differs in various respects from other scales that were arrived at through the
use of other testing techniques; difficulty being equated with error in all cases
(see Mukattash 1977; 1978). This variation amongst the various groups of
learners with respect to difficulty is a function of a multitude of psycholog-
ical, social and educational factors (see Nickel 1971; 1980; Mukattash forth-
coming-!)).
Linguistic difference is just one factor that contributes to difficulty in for-
eign language learning. Amongst the factors that emerge in this study is the
educational background. For example, members of Group IV and Group V
followed exactly the same syllabus and had exactly the same years of instruc-
tion in English. Furthermore, the two groups used the same textbooks and
experienced exactly the same method of teaching. Yet, the overall perform-
ance of Group IV on the test is significantly higher than that of Group V.
Furthermore, the results of the two groups with respect to the relative degree
of difficulty are distinctly different. For instance, nominale rank fourth on
the suggested scale of difficulty (see Table 1) for Group IV and seventh for
Group V. Similarly, whereas adjectivals rank sixth for Group IV, they rank
ninth for Group V. The only explanation for this discrepancy between the
results of the two groups seems to reside in the quality of English language
instruction which the two groups had had. Schools in the Amman area (i.e.
the Capital) attract the best teachers, at least as far as command of English is
concerned. 6 To further explain the point under discussion, let us consider the
figures in the fourth column in Table 1, i.e. average I.F. of second-year Eng-
lish majors from five T.T.I's. As pointed out above, the testees were given the
test a few weeks before they graduated to become teachers of English in
primary schools, especially in the outlying districts in Jordan. If the teachers
themselves commit all types of errors, it follows that they will inculcate in
the learners such linguistic errors and shortcomings as they themselves have
fossilized.
The nature and type of instruction in foreign language also plays a notice-
able role with respect to difficulty. Members of Group I (i.e. English majors
at J.U.) and Group II (i.e. final year students from foreign private schools)
had had almost the same amount of instruction in English at the time of the
administration of the test. Their results, nonetheless, indicate that they differ
with respect to difficulty. For instance, whereas modals rank third on the
scale of difficulty for GI, they rank sixth in the case of Gil. Furthermore, we
notice that members of GI have less difficulty in all the grammatical areas
tested. The only explanation that can be offered is that at the time of admi-
nistration of the test, university students (i.e. Group I) had had two courses
in English grammar (mainly traditional) at least, and thus they might be said
to have had more awareness of the facts of English grammar. Maturity might
be another factor responsible for this discrepancy: the ages of members of GI

ranged between 21 23, whereas the ages of members of Gil ranged between
17-19.
5 Linguistic difference
An attempt was made to establish patterns of difficulty across the various

grammatical areas included in the test. No syntactic or semantic relations
were found to hold between the items that proved to be most difficult (e.g.
those whose I.F. is 0.30 or below). Furthermore, it was not possible to ac-
count for all of the difficult items in terms of linguististic difference between
NL and TL. However, it was possible to establish with a reasonable degree of
adequacy patterns of difficulty within each grammatical area.
This is not the place to present or discuss patterns of difficulty in all the
grammatical areas tested. It will suffice for the purposes of this study to dis-
cuss the patterns of difficulty within one grammatical area, namely the verbal
system. This decision is based on more than one consideration. First, the
items testing the use of the verbal forms (35 in number) are representative of
the grammatical categories of tense, aspect, phase and voice (see Palmer 1974).
Secondly, the English verbal system presents a serious problem for Arab
learners of English (see Mukattash 1977; 1978) and possibly for all learners
of English as a foreign language. To quote Palmer (1974:1) on this matter,
"The most difficult part of any language is the part that concerns the verb".
Thirdly, the contrasts between the Arabic verbal system and the English verbal
system are linguistically interesting in the sense that in Arabic certain gram-
matical categories relating to the verbal system are mutually exclusive.
In what follows we shall briefly present an informal and brief account of
the basic facts of the verbal system in Arabic, and compare them with the
facts of the English verbal system to show points of correspondence and
divergence. Only those facts which are deemed to be relevant to our discus-
sion will be presented below.
Excluding what is traditionally referred to as the 'future' tense, we recog-
nize two tenses in Arabic: the present and the past. The present form of the
verb, exemplified by yaktubu 'he writes' is always two ways ambiguous, for
- depending on the linguistic or situational context it may be interpreted
in a temporal progressive sense (viz. 'he is writing now'), or in a non-progres-
sive 'factual' or 'iterative' sense ('he writes English well'; 'he writes letters
every day'). This verbal form cannot combine with the grammatical features
of 'phase' and 'aspect'. In other words, the English present progressive7 and
present perfect have no corresponding verbal forms in Arabic, and neither
340 Lewis Mukattash
has the present perfect progressive. The basic senses expressed by the English
present progressive (viz. temporariness, duration, futurity, etc.) are expressed
by means of the verb yaktubu, which corresponds formally to the English
simple present. As the English present perfect has no corresponding verbal
form in Arabic, the senses expressed by it are partially expressed by the Arabic
past form of the verb: viz. kataba. The senses expressed by the present per-
fect progressives are, however, partially expressed by the verb yaktubu, which
corresponds formally to the simple present and semantically (i.e. use) to both
the simple present and the present progressive.
This information about the correspondences between English and Arabic
present tenses 8 is represented diagramatically in Table 2, where FC stands for
formal correspondence; SC for semantic correspondence or strictly speaking
'use'(see Widdowson 1978:121); and for partial semantic correspondence.
Table 2
Arabic/FC English Arabic/SC
yaktubu writes yaktubu

is writing -
has been writing P/yaktubu
has written P/kataba (past)
In contradistinction to the present verbal form, the past verbal form in Arabic
combines with the morpheme kan ('be' in the past) to express anteriority in
the past, and thus might be said to correspond formally and semantically to
the English past perfect: viz. kana taraka 'he had left'. Similarly, the English
past progressive has a corresponding verbal form in Arabic which consists of
the sequence kana + V (inf) (e.g. kana yaktubu 'he was writing'). The English
past perfect progressive does not, however, have any corresponding verbal
Table 3
Arabic/FC English Arabic/SC
kataba wrote kataba

kana yaktubu was writing kana yaktubu
kana kataba had written kana kataba
had been writing P/kana yaktubu
form in Arabic. The basic sense associated with it is partially expressed by the
use of the Arabic past progressive (viz. kanayaktubu 'he was writing'). These
correspondences are diagrammatically represented in Table 3.
If linguistic difference is taken to be the basic source of difficulty in foreign
language learning and production, then we would expect the following diffi-
culties for Arab learners of English with respect to facts of the verbal system
discussed in the preceding two paragraphs:
1. The past perfect progressive, the present perfect progressive, the present
perfect and the present progressive, lacking formal correspondence in NL, will
prove to be the most difficult verbal forms for Arab learners.
2. The simple present having both formal and semantic correspondence will
be easier than the present perfect, the present progressive and the present per-
fect progressive.
3. The simple past, the past perfect and the past progressive, having both
formal and semantic correspondence in NL, will not prove to be difficult.
Table 4 shows the average I.F. of each of the verbal forms included in the
test. The first column shows the average I.F. for all the 4,835 testees, whereas
Table 4
average I.F.
category
G GI Gil
simple present 0.63 0.99 0.87
present perfect 0.46 0.94 0.77
present progressive 0.56 0.92 0.79
present perfect progressive 0.37 0.90 0.75
simple past 0.42 0.86 0.74
past perfect 0.38 0.78 0.69
past progressive 0.47 0.93 0.87
past perfect progressive 0.31 0.63 0.42

342 Lewis Mukattash
the second and third columns show the average I.F. of these forms for GI (stu-
dents from J.U.) and Gil (students from foreign private schools) respectively.
A careful study of the figures in Table 4 shows us that many of the predic-
tions made by CA are in fact true. First, the present perfect progressive and
the past perfect progressive prove to be the most difficult categories (I.F. 0.37
and 0.31 respectively). Secondly, the present perfect is more difficult than
both the simple present and the present progressive (I.F. 0.46,0.63 and 0.56
respectively). Thirdly, the simple present is easier than the present perfect,
the present progressive and the present perfect progressive.
The following cases cannot, however, be accounted for in terms of CA pre-
dictions, and constitute empirical confirmation that not all predictions made
by CA are valid:
1. The past perfect, which has both formal and semantic correspondence,
turned out to be more difficult than the present perfect, which has no formal
correspondence in NL (I.F. 0.38 and 0.46 respectively).
2. The past progressive, which has both formal and semantic correspondence
in NL, turned out to be more difficult than the present progressive, which
lacks formal correspondence (I.F. 0.47 and 0.56 respectively).
3. The simple past turned out to be more difficult than the past progressive
(I.F. 0.42 and 0.47 respectively) when both have formal and semantic cor-
respondences in NL.
Like the figures in Table 1, the figures in Table 4 clearly demonstrate that
the relative degree of difficulty of TL structures varies from one group to an-
other. For instance, while the past perfect progressive proves to be the most
difficult for the entire population (i.e. the 4,835 testees), the present perfect
progressive ranks second on the suggested scale of difficulty for the entire
population, but it ranks fourth in the case of GI (English majors at J.U.) and
Gil (students from foreign private schools). Similarly, while the present per-
fect ranks seventh in the case of GI, it ranks fifth in the case of Gil. In fact if
we examine the figures in the second and third columns (i.e. GI and Gil), we
notice that the facility indices of all the verbal categories are consistently
higher in the case of GI. The difference apparent in Table 4 (between GI and
Gil) was confirmed by a ' t ' test carried out on the means of the two groups;
the value o f ' t ' came out at 3.138, significant at 0.05. The only explanation
that one can think of is that members of GI, as pointed out above, had had at
least two courses in English grammar at J.U. prior to the administration of
the test. A major component of one of these grammar courses deals exclusive-
ly with the English verbal system.
It is quite obvious from the data presented in this section that not all the
predictions made by CA are valid. This is not by any means a new finding, for
it has been repeatedly demonstrated that CA predicts difficulties that do not
occur and fails to account for all of the learning problems that occur in the
classroom (see Lee 1968; Whitman et al. 1972; see also Schachter (1974:206)
and the references she cites).
6 Formal marking
I argue elsewhere (Mukattash 1980) that the more formal marking 9 a verbal
form incorporates, the more difficult it will be for Arab learners of English
regardless of the formal and semantic correspondences between NL and TL.
I also argue that this is almost always the case except possibly in the case of
the present progressive, which is introduced at the very beginning of the Eng-
lish language syllabus in elementary schools and which has a reasonably high
frequency of occurrence in the textbooks, particularly in the initial stage.
A comparison of the figures in Table 4 will testify to this assumption. The
only exception in this study is the past progressive, which turned out to be
easier than the simple past (I.F. 0.47 and 0.41 respectively).
The principle of formal marking also accounts for the students' avoidance
of the marked verbal forms in both speech and writing, as exemplified by
their preference of the simple past and the simple present. Table 5 shows the
frequency of occurrence of the various verbal forms (appropriate and errone-
ous) which were encountered in fifty essays written by first-year students at
the University of Jordan (see Mukattash 1978:255).
Table 5
category frequency per cent
simple present 1,259 74.4

simple past 324 19.1
present perfect 10 0.7
past perfect 17 1.0
present progressive 33 1.9
past progressive 11 0.7
present perfect progressive
past perfect progressive
simple future 38 2.2
future perfect
future progressive
total 1,693 100%

344 Lewis Mukattash
The figures in Table 5 are self-explanatory. The only point that is worth men-
tioning in this respect is that error counts based on the study of free compo-
sitions are misleading if the correlation between the frequency of use and the
frequency of errors is not calculated. This is one of the serious objections to
EA, which often, being text-bound, fails to account for the avoidance phe-
nomenon (see Schachter 1974). Indeed, the figures in Table 5 show that there
is not a single occurrence of the past perfect progressive in the fifty essays
and that there is only one occurrence of the present perfect progressive,
when - in fact - the present study as well as other test-types show us that
these two verbal forms are the most difficult for Arabs.
The principle of formal marking also accounts for the students' preference
of the active form of the verb, and their tendency to avoid the passive in both
speech and writing. Apart from two counter-examples (the simple past and
the past perfect), the figures in Table 6, which are extracted from the present
study, show that the average I.F. of the active verbal form is higher than its
corresponding passive form (for the entire population of testees).
Table 6
average I.F.
category
active passive
simple present 0.63 0.36

simple past 0.42 ? 0.46
present perfect 0.46 0.35
past perfect 0.38 ? 0.43
present progressive 0.56 0.36
past progressive 0.47 0.27
Further empirical confirmation of the principle of formal marking is pro-

vided by the following small-scale study, which was carried out independently.
Fifty students from the University of Jordan were given thirty Arabic sen-
tences to translate into English. Ten of these sentences were in the passive
voice. 10 Analysis of the translations showed the following:
(i) Only 117 sentences (23.4%) were correct with respect to the use of the
passive voice. Errors in the use of tense were not considered.
(ii) In 63 cases (12.6%) the passive voice was used, but there were mistakes
in verbal forms: e.g. had been build.
(iii) In 196 cases (39.2%) students used the active form of the verb in those
sentences that incorporated an explicit agentive phrase: viz. The windows
must clean by the workers daily.
(iv) In 124 cases (24.8%), and in an attempt to avoid the use of passive sen-
tences, students shifted the agentive phrase from sentence final position to
initial position, thus converting the passive sentence into an active one. 11
What is interesting to note in this respect is the fact that when the agent was
not explicitly expressed in the Arabic text, some students reconstructed a
generic agent and rendered the passive sentences in Arabic as active sentences
in English: viz. Samira gets angry when she is asked to do her duty (literal
translation), rendered as: Samira gets angry when someone asks her to do her
duty.
The data gathered from this translation test clearly demonstrate that Arabs
have a tendency to use active sentences even in contexts that require the use
of the passive.
It ought to be mentioned, however, that as in the case of linguistic differ-
ence, the principle of formal marking does not account for the whole range of
difficulties, there being variations amongst the various groups of learners. The
figures in Table 4 show us that in the case of GI, the past progressive is easier
than the simple past: I.F. 0.93 and 0.86 respectively. Notice further that in
the case of GI, the average I.F. of the present perfect progressive is higher
than that of the simple past (I.F. 0.90 and 0.86 respectively).
7 Concluding remarks
My claim in this study is that although it is not possible to set up a hierarchy

of difficulty from "an exclusively linguistic point of view"(Nickel 1971:193),
it is nonetheless possible, with the use of different testing techniques and
elicitation procedures, to establish patterns of difficulty for learners of a given
NL learning a certain target language. These patterns do, however, vary in
their degree of intensity according to a multitude of psychological, pedagog-
ical, social and national factors. Amongst the factors that have been shown
to bear on difficulty in this study are: the nature of instruction in the foreign
language; the age of the learner; the factor of chronology; and the frequency
of occurrence of TL structures in the syllabus. For any generalizations about
the relative degree of difficulty of TL structures to be valid, all the aforemen-
tioned variables at least should be studied systematically and in depth (see al-
so Nickel 1971; 1980).
It is by now a truism that insights gained from CA with respect to linguistic
difference and anticipated difficulty in TL cannot be true of all speakers of a
346 Lewis Mukattash
particular NL. NL is just one factor contributing to difficulty, whose role

seems to diminish as learners begin to develop awareness of the facts of TL.
Invoking independent linguistic principles (e.g. formal marking) to account
for difficulty may be helpful in a good number of cases, but as in the case of
linguistic difference, such principles run into difficulties. (Eckman (1977)
tries to account for difficulty in terms of markedness relations 12 stated in
universal grammar; but see Kellerman's (1979) criticism of his analysis; see
also footnote 1.)
Other conclusions that can legitimately be drawn from this study are:
i. There is a high correlation between difficulty and avoidance. The present
perfect progressive and the past perfect progressive, which proved to be the
most difficult verbal forms in this study, had a zero frequency of occurrence
in free compositons (see Table 5). Avoidance should thus be taken as an in-
dicator of either difficulty or ignorance (see James 1977).
ii. There is a correlation between patterns of difficulties that emerge in lan-
guage 'production' and those that emerge in language 'recognition': this is
evidenced by the similarity between error-types gathered from the two test-
types.
iii. Not all patterns of difficulty may be accounted for in terms of linguistic
difference or in terms of certain independent language principles (e.g. formal
marking; see also Eckman's(1977) 'typological markedness'). Applied linguists
will continue to account for difficulties in terms of broad and sometimes
vague labels. But this is not unexpected in the absence of attested and veri-
fied hypotheses about foreign language learning.
Notes
* I would like to thank Carl James for reading and commenting o n an earlier draft of
this paper. I alone am responsible for any errors in the text.
1. The following quotation from Kellerman (1979:43) shows the extent of the contro-
versy :
. . we have no satisfactory definition of 'difficulty'. It is not clear whether linguistic
difference = difficulty, and whether the degree of difference = the degree of diffi-
culty, whether error = difficulty or whether markedness hierarchies can predict diffi-
culty. Nor is it clear whether difficulty is something experienced by the learner or
may be something hypothetically attributed to him when he makes mistakes or
avoids."
2. The average facility indices of the various grammatical areas for the entire population
(i.e. 4,835 testees) vary in various respects from those for the five groups listed in
Table 1. First, in the case of the five groups, modals turn out to be more difficult
than prepositions (I.F. 0.61 and 0.62 respectively). Secondly, nominals and connec-
tors have the same I.F. (0.60), and so do adjectivals and relatives (0.65). This is also
true of articles and pronouns (I.F. 0.66). See, however, footnote 3.
3. Admittedly, it is not satisfying to simply state the rank of each grammatical area on
a scale of difficulty. In fact this may be misleading, particularly in the case where
there is no noticeable difference between the average facility indices of two or more
grammatical areas. To illustrate this point, let us compare the average I.F. of sentence
connectors with that of nominals (viz. 0.44 and 0.45 respectively). This should not
be interpreted to mean that sentence connectors are more difficult than nominals.
This observation is based on the fact that the average I.F. of one of the items testing
the use of connectors was found to be 0.19; whereas the lowest I.F. of the items test-
ing the use of nominals was 0.34. This means that the one item whose I.F. was 0.19
lowered the average I.F. of connectors, thus making it lower than that of nominals.
This difference is, in fact, more apparent than real.
4. Orthodox CA (see Lado 1957; Stockwell et al. 1965) views learning difficulty and
linguistic difference as being directly and proportionally related. The view of equating
difficulty with linguistic difference has been repeatedly discredited and criticized on
theoretical, methodological and empirical grounds. For a summary of the basic ob-
jections to CA, see Tran-Thi-Chau (1975:124); see also Lee (1968), and Whitman et
al. (1972).
5. The most well-known and explicit hierarchy of difficulty based on CA is that pro-
posed by Stockwell, Bowen and Martin (1965) in The grammatical structures of Eng-
lish and Spanish. For a criticism of the hierarchy see: Nickel (1971), Whitman et al.
(1972), and Tran-Thi-Chau (1975).
6. Being the Capital, Amman also provides students with more opportunities for mixing
with native speakers of English than any other place.
7. In colloquial Jordanian and Palestinian Arabic, the English present progressive has a
corresponding verbal form that consists of the sequence lam+V. The morpheme
1am does not occur in Standard Arabic, and its linguistic status is still unkown. Mit-
chell (1978:238) refers to it as 'particle'. This might be another reason why the pre-
sent progressive does not constitute a difficulty for Jordanian learners of English.
8. All the verbal inflections in Arabic which relate to number, gender and person will be
ignored here.
9. It could be the case that what is involved here is what Tran-Thi-Chau (1975:133)
refers to as the 'systematic complexity' of the English verb system.
10. Eckman (1977:321) wrongly points out that Arabic does not have passive sentences
with expressed agents.
11. As in the case of English, agentive phrases in Arabic occur in the final position of
passive sentences.
12. Eckman (1977:320) defines markedness in the following way: "A phenomenon A in
some language is more marked than if the presence of A in a language implies the
presence of B; but the presence of does not imply the presence of A." His basic
assumption is that areas of TL which differ from NL will be difficult if they are more
marked than NL, but they will not be difficult if they are not more marked.
References
Chomsky, Noam
1971 "Formal discussion of Wick Miller and Susan Ervin's: The development of
grammar in child language". The acquisition of language, edited by Ursula
Bellugi and Roger Brown (Chicago: The University of Chicago Press), 3 5 - 3 9 .
Eckman, Fred R.
1977 "Markedness and the contrastive analysis hypothesis", Language learning
27.2:315-330.
James, Carl
1977 "The ignorance hypothesis in interlanguage studies", Interlanguage studies
bulletin 2 . 2 : 1 5 2 - 1 6 5 .
348 Lewis Mukattash
Kellerman, Eric
1977 "Towards a characterisation of the strategy of transfer in second language
learning", Interlanguage studies bulletin 2 . 1 : 5 8 - 1 4 5 .
1979 "The problem with difficulty", Interlanguage studies bulletin 4 . 1 : 2 7 - 4 7 .
Lado, Robert
1957 Linguistics across cultures (Ann Arbor, Michigan: The University of Chicago
Press).
Lee, W. R.
1968 "Thoughts on contrastive linguistics in the context of language teaching",
Report on the nineteenth annual round table meeting on linguistics and lan-
guage studies, edited by James E. Alatis (Washington, Georgetown University
Press), 1 8 5 - 1 9 4 .
Mitchell, T. F.
1978 "Educated spoken Arabic in Egypt and the Levant, with special reference
to participle and tense", Journal of linguistics 1 4 : 2 2 7 - 2 5 8 .
Mukattash, Lewis
1977 Problematic areas in English syntax for Jordanian students (Amman: The
University of Jordan).
1978 "A pilot project in common grammatical errors in Jordanian English", Inter-
language studies bulletin 3.2:250-291.
1979 Further studies in Jordanian English (Amman: The University of Jordan).
1980 "Yes/No questions and the contrastive analysis hypothesis", ELTJ. 34.2:
133-145.
forth-a " ^ - q u e s t i o n s in English: a problem for Arab students", to appear i n I R A L .
forth-b "English language proficiency in Jordan: educational, socio-economic and
other factors", to appear in Dirasat (The University of Jordan).
Nickel, Gerhard
1971 "Variables in a hierarchy of difficulty", Working papers in linguistics 3.4:
185-194.
1980 "Some pedagogical implications of error analysis and contrastive linguistics",
Tijdschrift van de VUB 2 1 : 6 0 - 7 0 .
Palmer, F. R.
1974 The English verb (2nd edition) (London: Longman).
Schachter, Jacquelyn
1974 "An error in error analysis", Language learning 2 4 . 2 : 2 0 5 - 2 1 4 .
Schouten, . . H.
1979 "The missing data in second language research", Interlanguage studies bulletin
4.1:3-14.
Stockwell, R. P. - J. D. Bowen - J. W. Martin
1965 The grammatical structures, of English and Spanish (Chicago: The University
of Chicago Press).
Tran-Thi-Chau
1975 "Error analysis, contrastive analysis, and students' perception: a study of
difficulty in second-language learning", IRAL 13.2:119-143.
Whitman, Randal L. - Kenneth L. Jackson
1972 "The unpredictability of contrastive analysis", Language learning 22.1:29-41.
Widdowson, H. G.
1978 Teaching language as communication (Oxford: Oxford University Press).
WIES LAW OLEKSY
Towards pragmatic contrastive analysis
0 It is almost a truism to observe that contrastive analysis (CA) has always

been based, one way or another, on the existing theories of language and the
range of linguistic phenomena dealt with has been broad and varied. Thus,
there exist contrastive analyses rooted in the traditional grammar, structural
grammar, stratificational grammar and generative-transformational grammar
including both 'traditional' transformational approaches, as well as case gram-
mar and generative semantics. Occasionally, other types of framework have
been applied, e.g. functional sentence perspective, performative analysis, etc.
The evaluation of the applicability of these theoretical frameworks to CA can
be found in numerous works, e.g. Di Pietro 1968, Moulton 1968, Dingwall
1964, Fisiak 1975,Szwedek 1975,Rusiecki 1976,Sajavaara 1977,Krzeszowski
1979, to mention just a few names.
It is impossible and unnecessary to deal here with the complexity of pro-
blems, both theoretical and practical ones, involved in CA. It suffices to say
that CA, whatever theoretical framework it has been based on, has been
oriented towards contrasting phonological, syntactic and semantic systems or
sub-systemes of L! and Lj, or Li and L 2 and L 3 , . . . etc. Thus, CA has been
predominantly concerned with the analysis of problems pertaining to language
structure rather than language use. More significantly, the basic theoretical
apparatus of CA with its key notions such as comparability, congruence, and
equivalence has been devised so as to deal with language form and not lan-
guage use. Accordingly, it seems correct to observe that in order to be able
to tackle problems of language use CA cannot be restricted to just analysing
the sounds, the grammar, and the vocabulary of Lx and L 2 . It must go beyond
the limitations imposed on linguistic analysis by the form-centered theories
of language in an attempt to analyze the use of language in a sociocultural set-
ting. By form-centered theories of language I mean such linguistic traditions
as structuralism or transformationalism. In this context, speech act theory
wl be considered to be a use-centered theory. The distinction between the
former and the latter shows up in what has been the focal point of interest in
350 Wiestaw Oleksy
the respective theories, i.e., the basic units of analysis. The form-centered
theories have been sentence-bound and syntactic considerations have been
prevailing in the research of the representatives of this linguistic orientation.
The above is also true of the transformational tradition, even though it re-
presents more interest in idealizations: rules are said to account for the native
speaker's competence.
There is a general agreement as to the validity of the study of language in
its sociocultural setting, both among the theoretical linguists and philosophers
of language as well as those concerned with language learning/teaching. How-
ever, little has been written on the nature of processes involved in the study
of social setting of language from the contrastive point of view.
In this paper I wish to propose that for CA to be able to cope with lan-
guage use in a sociocultural setting it must be based on or, at least, related to
linguistic pragmatics understood very broadly as the science of language use
(Haberland - Mey 1977). In other words, I wish to propose a model of Prag-
matic Contrastive Analysis (henceforth PCA). Because of the preliminary
nature of the enterprise, the limitations of this paper, and the lack of empir-
ical research in the field, this paper should be treated as a handful of sugges-
tions pointing to some possibilities rather than offering proofs or definite ana-
lyses. The model attempted here is programmatic rather than normative.
1 A model of PCA
A careful reader of Papers and studies in contrastive linguistics (Fisiak, ed.)

might recall that PSiCL 10 (1979) carried two papers in one of which prag-
matics was referred to (Wachtel 1979) and in another one the author (Riley
1979) proposes a "Model of contrastive pragmalinguistics", i.e., exactly what
the present author wishes to do in this paper. For reasons of time and space
the discussion of Riley's proposals will not be dealt with here for they deserve
a separate treatment. Suffice it to say that he attempts a contrastive model of
discourse whereas my aim in this paper is to present a model of contrastive
analysis that would be capable of dealing with a communicative act in con-
creto rather than with discourse structure in extenso.
Returning now to PCA I propose that it should accommodate the concept
of communicative act. Furthermore, I propose that PCA can be conveniently
carried out if it contains the following four components:
A. Locutionary Component LC
B. Illocutionary Component Ill.C
C. Pragmatico-Contrastive Component PCC
D. Interactive Component IC
Towards pragmatic CA 351
1.1 LC comprizes what Bach and Harnish (1979:3) call 'Utterance Act'and
'Locutionary Act'. I shall not devote much attention to LC for reasons of
scope but it is worth noting that anyone who wishes to account for the locu-
tionary content of an expression used by a speaker S on a given occasion will
have to solve the problem of the choice of the apparatus that adequately ac-
counts for the expression in L 1 and, in addition, allows for a contrastive treat-
ment of the equivalent expression in L 2 . That is to say, the analyst will have
to decide on a particular type of contrastive grammar.
It can be added in passing that LC should account for both the syntactic
(including phonological) and the semantic characterization of expressions in
L! and L 2 . Typically, the term "expression in Lj and L 2 " will refer to a sen-
tence. Needless to say, most existing CAs have been limited to what I call
here LC. Furthermore, the existing CAs based on generative grammars have
been constructed so as to account for ideal objects, i.e., well-formed sentences.
This "virtue" of CA inherited from generative grammars makes the former
and the latter a rather inadequate mechanism to deal with linguistic com-
munication which, as is well known, allows for incorrectness. For example,
(1) and (2) below can both express the speaker's intention:
(1) John me give this thing yesterday.

(2) John gave me this thing yesterday.
From the communicative perspective alone both sentences are equally effec-
tive.
Before I leave LC and pass on to the discussion of Ill.C. I shall dwell a bit
on the opposition sentence/utterance. The understanding of the relation be-
tween sentence and utterance throws a good deal of light, I think, on the type
of connection that exists between LC and Ill.C. To put the matter succinctly
I shall quote Katz (1977:14) who wrote the following in this connection:
Sentence meaning is the meaning of a sentence type in the lan-
guage, whereas utterance meaning is the meaning of a particular
use, or token, of a sentence type on that particular occasion.
Grammars represent sentence meaning because the meaning of a
sentence in the language is the meaning it has by virtue of its hav-
ing a particular grammatical structure. Pragmatic theories repre-
sent utterance meaning because the meaning of an utterance is
the meaning it has by virtue of its being a specific spatiotemporal
occurrence of a sentence in a particular contextual structure.
1.2 The Illocutionary Component. The basic unit of analysis in Ill.C is the
communicative act. Generally speaking, communicative acts can be realized
352 Wiesiaw Oleksy
verbally and nonverbally. (A nod can be as effective communicatively as an

utterance of a full sentence, e.g.: Yes, I agree with you at this point.) My in-
terest in this paper is the verbal communication and for this reason I shall deal
with Speech Act (SA) as the basic unit of analysis of verbal communication.
It is worth noting that at the level of LC S's utterance will be interpreted
from the point of view of its propositional and referential content. That is to
say, that by assigning a propositional interpretation to S's utterance it will be
possible to define what the speaker means in saying something. This, however,
does not account for what the speaker does in saying something. The task of
interpreting of what the speaker does boils down, in fact, to the identification
of the type of SA that the speaker has performed. This task will be allocated
to Ill.C.
The above approach to what the speaker means in saying something as op-
posed to what he does in saying something implies that an analysis of SAs
must accomodate the fact that in linguistic communication meaning some-
thing must be distinguished from doing something and that the latter does
not have to be deduced from the former in a straightforward fashion. Thus
the important task of the analyst is to propose mechanisms by means of
which it would be possible to account for what the speaker does in saying
something. To put the matter succinctly, I will claim that what the speaker
does in saying something is performing all kinds of communicative acts. In
this paper my interest is in the speech acts and for this reason the discussion
of the problems alluded to above will be limited to proposing ways by means
of which the type of SA that the speaker has performed could be defined.
The definition of the type of SA that the speaker has performed will be re-
ferred to as the assignment of the Speech Act Value (SAV) to the speaker's
utterance.
Suggested below are three possible steps leading to the assignment of SAV
to S's utterance:
i. Illocutionary force (IF) of S's utterance must be defined.
ii. Felicity conditions (FC) for SA which S has performed must be specified.
iii. Sociocultural context (SC) in which SA typically occurs must be specified.
IF expresses the communicative property of the utterance which was used

by S in the given context. IF can be assigned to S's utterance in the following
circumstances:
a. Performative clause is expressed overtly in S's utterance and, therefore,
IF is, as if automatically, assigned to S's utterance unless there is some
other evidence, e.g. linguistic or contextual, to the contrary. If the assign-
ment of IF is done on the basis of the presence of the performative clause
then it can be said that S has performed a direct (literal) SA whose name is
usually coined after the predicate verb which occurs in the performative
clause.
b. Performative clause does not occur overtly but there are other linguistic
means of signalling IF, e.g., the occurrence of please can be said to signal
requests (cf. Sadock 1974:120). Intonation can also be used as a device to
signal the type of SA being performed, (a) and (b) above can be cases of
the literal use of an expression by S, unless (a) is the case of an embedded
performative or a hedged performative. If the latter is the case then S's ut-
terance can not be interpreted literally and the assignment of IF must be
done as is indicated in (c) below.
c. Performative clause does not occur overtly in S's utterance and there are
no linguistic signals present in S's utterance indicating IF. If (c) is the case
then S's utterance can be interpreted literally or nonliterally. In the latter
case, it is claimed, the identification of SA which S has performed will be
possible on the basis of FC and SC, and/or on the basis of what Grice called
implicature.
Felicity conditions (FC) are conditions which guarantee that an SA per-
formed by S in communicative context (CC) is effective, i.e., it expresses S's
intention to the addressee A who, in turn, decodes S's intention. Therefore,
the essential constituents which are present in defining FC are the following:
a. S's assumptions about himself and about A.
b. S's assumptions about the empirical setting in which the performance of
SA takes place.
Now I will propose FCs for the speech act Question (FCQ) which are based
on Oleksy 1979a and 1979 b.
As was said earlier in this paper, FC's are what the speaker assumes about
the addressee, about the content of what is being communicated, and about
the empirical setting in which an SA occurs. Judgments concerning felicity of
an SA rely heavily on general principles of conversation as well as on the
unique conditions for the SA being performed by the speaker at a given time
and in a given context. All this amounts to saying that SAs are not performed
in a vacuum and that the communicative context is an integral part of verbal
exchange. In the case of Question the communicative context in which this
SA typically occurs is the so-called erotetic context. (I coined the term after
Belnap's (1969) erotetic situation or, as he put it, the asking questions-in-
circumstances.) A typical erotetic context (EC) will have the following con-
stituents:
(3) S = the speaker
A = the addressee
An = the answer
354 Wiesiaw Oleksy
Q = the asking of a question

R = the state of affairs (non-linguistics)
Assuming that everything that is present in a typical EC is given in (3) above
I want to propose that the following set of assumptions held by the speaker
constitute the felicity conditions for Question.
(4) 1. S is ignorant of R
2. S wants to know R
3. S believes that he will know R if Q
4. S believes that A knows R
5. S believes that A is willing to share R with him
6. S believes that A will supply An
7. S believes that R is present in An
8. S believes that Q is appropriate in EC
It is easy to see that 41_s can be further subcategorized. Accordingly, 4 _ 3
are the assumptions that the speaker is making about himself and 4 4 _ 6 are
the assumptions the speaker is making about the addressee. 4 7 and 4 3> jointly,
are the assumptions S is making about the conversational context in which
Question is performed. 4 3 also constitutes a cognitive motivation for S to per-
form the speech act he is performing. 4 7 , in turn, expresses S's belief that the
cognitive motivation for the performance of SAQ will be satisfied. Finally, 4 8
accounts for the fact that S's utterance is purposeful and appropriate in the
communicative context in which the verbal exchange takes place.
It is believed that it is possible to formulate FCs for other SAs and that by
doing so it is possible to distinguish one SA from another.
By sociocultural context (SC) I understand those conditions which pertain to

the felicitous use of a speech act relative to social norms acceptable in the
culture to which the participants speaking a language L belong. At this point
it is assumed that "the participants speaking a laguage L" refers to native
speakers and "a language L" is the same language for the participants. I shall
not attempt any definition of such terms as "social norm" or "culture" be-
cause this is not at issue. What is at issue is the provision of a general frame
within which to study speech acts. The main stress, it seems, should be placed
on the use of a speech act in certain culturally, institutionally, and situation-
ally restricted contexts rather than on the communicative abilities of the user
in different societal strata.
Without going into details one can say that SC accounts for those aspects
of the verbal exchange which characterize the participants and the extra-
linguistic context in which the verbal exchange is performed. It is suggested
that SC can be considered in terms of sociocultural restrictions imposed on
the participants by social norms accepted in the given speech community and
culture at the given time. Thus, sociocultural restrictions will have to account
for several things pertaining to the explanation of how and to what extent
such sociolinguistic parameters as, e.g., sex, age, status, etc., of the participants
on the one hand, and the cultural (institutional) and situational contexts on
the other interact with FCs.
Closely connected with SC seems to be the role relationship that exists be-
tween the speaker and the addressee. It is worth adding that from the point
of view of the speech act analyst it is crucial how the participants contribute
to the ongoing verbal exchange, i.e., what role in regard to the speaker/ad-
dressee distinction each of them assumes at the time of the speech act. Such
roles as, e.g., father-son, student-teacher, boss-secretary, friend-friend, etc.,
are relevant because they have a bearing on the felicitous use of certain speech
acts. For example, a 10-year-old boy cannot be said to have performed a (fe-
licitous) command if he uttered (5) to his father with an intention to issue a
command.
(5) Father, get me a bike at once.
One of the conditions for the felicitous use of commands is that the speaker
(who is issuing a command) is in a superior position in relation to the addres-
see. Since this is not the case in the situation dealt with above (5) is not effec-
tive: the speaker has failed to produce a felicitous command. However, on an-
other occasion the same boy may have performed a felicitous command, for
instance upon uttering (6) to his younger brother:
(6) Tom, get me a bike at once.
In the analysis of SAs it is crucial to remember that role relationship should
be defined for the given speaker and the given addressee relative to the spe-
cific communicative context in which SA is performed. It seems thus best to
consider role relationship as a set consisting of two members: one for the
speaker and one for the addressee. If any of the members in the set changes
membership, e.g., the addressee becomes the speaker, etc., or a new partici-
pant becomes involved in the verbal exchange or the communicative context
in which the participants are performing changes, a change in the role relation-
ship may be brought about. Kando (1977:246) points out to the fact that
sometimes two roles, although performed at different times, may result in
creating a role conflict in the individual. This is the case, for instance, of the
teaching assistant in a university who has the difficult task of reconciling
allegiance to the student body with identification with the faculty. For our
purposes it is enough to note that role relationship changes with the changing
conditions in which the participants are performing. Of course, it remains to
356 Wies taw Oleksy
be explained on the basis of empirical data what the elements of the changing
conditions are and how they contribute to the ongoing verbal exchange.
Perhaps what is most interesting and relevant to the study of SAs from the
point of view of SC is the choice of strategies for the performance of SAs by
the speakers on the one hand and the appropriateness of strategies in certain
contexts on the other. This brings me to the last part of the section dealing
with the Illocutionary Component:
Strategies for the performance of speech acts. Above I have tried to deal
with two things: one was connected with how it was possible to define a
felicitous SA and another one was connected with how the functioning of
SAs in socioculturally restricted contexts could be approached. Now I wish
to address the following question: How does a speaker perform the intended
act? The answer to this question involves whatFraser(1978:7) called strategies
for the performance of speech acts. He defined strategy in the following way:
" . . . the term strategy [ . . . ] refers to the particular choice of sentential form
and meaning which the speaker employs in order to perform the intended act".
The choice of a particular strategy may vary from speaker to speaker and
is related to the linguistic repertoire of the speaker on the one hand and the
speaker's assessment of the communicative context in which a speech act is
performed on the other. It seems obvious that the output of the sociocultural
analysis of speech acts can throw some light on the choice of strategies. Of
particular relevance will be the role relationship that exists between the
speaker and the addressee at the time of the speech act, the degree of famil-
iarity between the speaker and the addressee, the participants' age and sex, and
the institutional and situational context in which a speech act is performed.
All the abovementioned factors are assumed to be pertinent to the choice of
strategies for the performance of a particular speech act. It has recently been
suggested (Walters 1979:279) that closely related to the choice of strategies
for requests is the politeness of request forms.
The problems dealt with above can be summed up under one general label
which I call pragmatic competence. In my understanding of the term, prag-
matic competence is the ability of an individual to use an SA in the given com-
municative context with a particular strategy in order to obtain maximum
communicative and social goals. This treatment of pragmatic competence is
reminiscent of Edmondson's(1980:2) social competence.
In what follows I shall demonstrate a few strategies for the performance
of a direct question. One obvious strategy for any speech act is to announce
the intended force of the utterance with which a speech act is performed.
This can be done if IF (and hence the intended meaning) of the speech act
being performed occurs overtly in the sentential form which represents a given
strategy. (7), (8), and (9) are examples of such overt manifestations of the
strategy for the performance of the following speech acts: a question, a pro-
mise, and a warning; respectively (7), (8), and (9).
(7) I want you to tell me "x".
(8) I promise I'll do "x".
(9) I warn you not to do "x".
But more than often, the speakers prefer not to announce their intentions in
performing a speech act and choose some less direct means of conveying to
the addressee the force of the speech act. (1023) are the strategies for the
performance of a direct question.
(10) What's your name?
(11) I want to know your name.

(12) Please, tell me your name.
(13) Can you tell me your name?
(14) YouH tell me your name, won't you?
(15) Tell me your name, will you?
(16) YouH tell me your name, OK?
(17) Is your name "X"?
(18) Are you Ms. "X"?
(19) You haven't told me your name yet.
(20) And your name i s . . . ?
(21) I must know your name.
(22) You must tell me your name.
(23) I'd be happy if you told me your name.
etc.
The above examples show that there are many ways to ask a question about
someone's name. Common to all utterances throughout (1023) is the speak-
er's desire to know someone's name.
It can be noticed that some of the strategies in (10-23) will be restricted
to certain communicative contexts. For example, it seems unlikely that (20)
and (23) could be used in the same context. Although a certain degree of pre-
dictability of the occurrence of strategies is possible, yet it seems advisable to
358 Wiesiaw Oleksy
rely on empirical data in this matter. The elicitation of data in this type of re-
search is of primary importance. It will be suggested that at the beginning
stage of data collecting the following procedure may prove useful:
a. elicitation of strategies: either retrospectively or experimentally
b. establishing hierarchies of strategies on the basis of the frequency of use as
reflected in the experimental data
c. differentiation among strategies relative to communicative context
d. differentiation among strategies relative to other pragmatic factors, such as,
e.g., politeness, mitigation, etc.
The elicitation of strategies can be based on pre-arranged situations, e.g., role
playing, in which the interactants would assume the roles of the speaker and
the addressee. These pre-arranged situations can be constructed in such a way
that they contain the following factors:
a. setting the communicative context in which S A will be performed is pro-
vided;
b. participants the role relationship between the speaker and the addressee
is indicated;
c. goal - the type of SA to be performed by the speaker is specified.
By way of illustration an example is provided. In the pre-arranged situation
demonstrated below the speaker is expected to produce a strategy for the per-
formance of the SA of Question. Presumably, the data obtained in this ex-
periment would be related to strategies retrospectively proposed in (10-23)
above.
a. setting A friend of yours has invited you to a party in his appartment.
b. participants At the party you've noticed someone you think you'd met
before but you can't recall her name.
c. goal You want to know her name. What would you say to her?
To finish this section it is worth pointing out that despite the fact that some
pioneering work on matters alluded to above has already been done (cf.
Fraser 1978, House Kasper 1979, Walters 1979) the research methodology
pertaining to the elicitation of data for the analysis of pragmatic competence
is waiting to be developed.
1.3 Pragmatico-contrastive component. In what follows I will try to show

how the analysis of SA proposed in the preceding section can be applied to CA.
At the outset it must be mentioned that a contrastive treatment of SAs has
not, as yet, received much attention to the best knowledge of this author. For
this reason it is necessary to provide theoretical grounds for the comparison
of SAs across languages.
In the first place, it must be assumed that an SA in Lx is comparable to an
SA in L 2 . This assumption is based on the belief that functions to which lin-
guistic expressions can be put are essentially the same across languages. The
above claim has, in fact, been recently put forth by Fraser, Rintell and Walters
(1979) as reported in Walters 1979:278. I repeat this claim (and two other
claims) here for convenience.
i. Every language makes available to the user the same basic set of speech
acts such as requesting, apologizing, declaring, promising, and the like,
with the exception of certain culture-specific ritualized acts such as baptiz-
ing, doubling at bridge, excommunicating and the like.
ii. Every language makes available basically the same set of strategies - se-
mantic formulae for performing a given speech act.
iii. Languages will differ significantly with respect to both when a particular
speech act ought to be, ought not to be, or may be performed, and with
what strategy.
From the contrastive point of view it is thus necessary to state conditions
which must be met for an SA to be equivalent in L t and L 2 . In order to do
this it will be suggested that the mechanisms proposed in the section above
for the identification of SA can be employed. Accordingly, an SA can be said
to be equivalent across languages if it is characterized by the same IF and the
same set of FCs. This is expressed in (24).
(24) SA,L! = SA,L2 if a. I F S A > L ] =IFSA,L2
b- FC S A j L i = F C s a
When the equivalence of an SAin Lj and L2 has been established what remains
to be done is to find out how a given SA is realized in Lx and L 2 . That is to
say, it is necessary to identify linguistic expressions in L t and L 2 which are
used by speakers of these languages for performing a particular SA. At this
stage, the procedures for the elicitation of strategies for the performance of
SAs outlined in the section above can be conveniently applied.
If the above suggested procedure is applied the equivalence of expressions
used by speakers in Li and L2 will be related to the fact that they are realiza-
tions of the same SA. In other words, if an SA in L { is equivalent to an SA in
L2 then the linguistic expressions which have been used in L 1 for performing
that SA must be equivalent to the linguistic expressions used for performing
the same SA in L 2 , for it would be absurd to hold that an SA in question is
equivalent in L t and L 2 and its realizations are not.
The above discussion can be summed up by saying that the comparability
of SAs across languages is guaranteed by the fact that SAs, as communicative
categories, are equivalent ex definitione: the speech act of requesting in Eng-
lish is, for example, equivalent to the speech act of requesting in Polish.
Similarly, IF and FCs which are used for the identification of a particular SA
are assumed to be language universal.
360 Wiesiaw Oleksy
However, it is also possible to approach the question of the equivalence of

SAs from another direction. Namely, given a linguistic expression X t in L,
with which a speaker Sj has performed a speech act Y n in the communica-
tive context Zlt it is worth investigating what linguistic expression X2 in L2
can be used by the speaker S 2 to perform an equivalent speech act Y\_2 in a
socioculturally comparable communicative context Z 2 . This solution touches
upon what I call pragmatic equivalence (PEq.) of linguistic expressions in Lj
and L 2 . It is then proposed that pragmatic equivalence of linguistic expres-
sions across languages be stated in terms of the equivalence of the speech act
which they realize, i.e., with which a given equivalent SA can be performed.
Thus I propose the following formula by means of which PEq. of linguistic
expressions in Li and L2 can be accounted for:
(25) A linguistic expression X! L t is pragmatically equivalent to

a linguistic expression X 2 L 2 if both Xj and X2 can be used
to perform the same SA in L t and L 2 .
The above is expressed in (26).
(26) PEq.(X 1 L 1 ,X 2 L 2 ) = SA(L 1 ,L 2 )
Assuming that PEq. of linguistic expressions in L! and L 2 has been correctly

accounted for in (26) I now wish to address myself to the question of how
the analysis of SAs proposed in the first part of this paper can be applied to
CA, i.e., how pragmatically equivalent expressions can be analyzed in L t and
L 2 . The discussion will be limited to pointing out some of the applications
of the Locutionary Component (LC) and the Illocutionary Component (Ill.C)
to CA. The Interactive Component will not be dealt with here for reasons of
space.
A contrastive treatment of linguistic expressions at the level of LC should
account for the syntactic and semantic characterization of linguistic expres-
sions which have been used to perform Eq.SA in Lj and L 2 . This stage of
analysis is important because it throws a good deal of light upon the formal
properties of linguistic expressions. At this level of analysis, as has long been
done in traditional CAs, formal properties of expressions in L t and L2 must
be defined from the point of view of formal equivalence (FEq.) which they
may or may not exhibit (cf. Krzeszowski 1979:9193 for a discussion on
equivalence in CA). It is worthwhile mentioning that linguistic expressions
across languages which exhibit pragmatic equivalence (PEq.) do not have to
exhibit formal equivalence (FEq.). This claim can be supported by the find-
ings reported in House - Kasper 1979:10 to the effect that (27) in English is
PEq. to (28) in German, though it is clear that they are not FEq.
Towards pragmatic C A 361
(27) Can you close the window?

(28) Du solltest das Fenster zumachen.
(29) You should close the window.
(30) Kannst du das Fenster zumachen?
On the other hand, (27) in English and (30) in German are FEq. and PEq., but
it was discovered experimentally by House and Kasper that in the data that
they investigated the frequency of use showed more correlation between (27)
and (28) though they are not FEq. than between (27) and (30) which are FEq.
The above discussion seems to support the claim about a necessity of the
introduction of the distinction between pragmatic equivalence and formal
equivalence into CA.
In 1.2, the section dealing with the Illocutionary Component (Ill.C), I have
distinguished the following criteria which characterize SA: (a) IF, (b) FC,
(c) SC. Also, I have indicated that (d) role relationship holding between the
speaker and the addressee and (e) strategies for the performance of SA are of
primary importance. IF and FCs have been closely connected with the identi-
fication of SA as an object of linguistic analysis. Furthermore, IF and FCs
have been proposed as indicators of equivalent SAs across languages.
In the discussion that follows some aspects of (c), (d) and (e) above will be
touched upon. While (a) and (b) seem to be useful in the characterization of
SAs irrespectively of the language and culture of the speaker, (c), (d) and (e).
are language and culture specific. Following the claims put forth by Fraser,
Rintell and Walters (1979) it will be proposed that a contrastive analysis of
SAs should incorporate (c), (d) and (e) above. By doing so it will be possible
to provide answers to questions about the culturally, institutionally and
situationally restricted contexts in which a particular SA ought to be, ought
not to be, or may be used.
A pragmatic contrastive analysis (PCA) of SA should thus demonstrate
how an equivalent SA functions in the cultures to which speakers of L, and
L 2 belong. It is believed that an adequate contrastive analysis of equivalent
SAs cannot be performed without the incorporation of the sociocultural com-
ponent, if the analysis is to demonstrate how SA under analysis functions in
culturally distinct societies.
The functioning of Eq.SA in culturally distinct societies can be carried out
successfully if strategies for the performance of a given SA are related to such
factors as role relationship holding between the speaker and the addressee and
other pragmatic factors such as politeness, mitigation, level of directness (cf.
House - Kasper 1979:3-12). It must be stressed at this point that the identi-
362 Wiesiaw Oleksy
fication of pragmatic factors and their interaction with linguistic and socio-
cultural factors must also become a matter of interest for a contrastive linguist.
Although research on the use of SAs across languages is scarce it is justi-
fied to claim, on the basis of the research alluded to in the first part of this
paper, that pragmatic factors together with sociocultural restrictions are es-
sential and underly speakers' decisions as to the choice of a particular strategy
for performing an SA. The aim of PCA is to isolate and explain these factors.
It goes without saying that the impact of pragmatic and sociocultural fac-
tors on speakers' decisions concerning the choice of a strategy can be per-
ceived differently across languages and cultures. Accordingly, the following
claim is put forth:
(31) The choice of strategy for the performance of SA is related
to pragmatic factors and sociocultural restrictions on the
use of a given SA, which is language specific.
It is believed that this claim should augment the claims put forth by Frser,
Rintell and Walters (1979).
It can be noticed that the choice of strategies is also related to what has
been previously called pragmatic competence. If the notion of pragmatic
competence is applied to foreign language learning it becomes obvious that
learners of a foreign language will display different degrees of familiarity with
sociocultural restrictions and pragmatic factors governing the use of a parti-
cular strategy for the performance of SAs. In other words, learners will en-
counter difficulties with regard to the acquisition of pragmatic competence:
their verbal behaviour in a foreign language will then display pragmatic errors
(cf. House Kasper 1979:2). It seems thus sensible to suggest that the proper
place where pragmatic errors can be studied is pragmatic error analysis. This
suggestion will not be elaborated any longer in this paper.
Summing up the last part of this paper it can be said that PCA should focus
on the identification and differentiation of linguistic expressions used by
speakers of Lj and L 2 for the performance of SAs and, in particular, PCA
should explain the motivation which underlies decisions made by the speaker
as to the choice of a particular linguistic expression (i.e., a strategy) to per-
form a given SA. As has been repeatedly pointed out in this paper research
methodology for PCA should be confronted with experimental data.
The present paper should be treated as an outline of problems to be solved
and areas to be investigated. However, it is believed that what has been sug-
gested here will prove useful in handling matters pertaining to the understand-
ing of how speakers across languages manipulate linguistic expressions to per-
form different societal tasks.
References
Bach, Kent - Robert M. Harnish

1979 Linguistic communication and speech acts (Cambridge, Mass.: The MIT Press).
Belnap, Nuel
1969 "Questions: their presuppositions and how they can fail to arise", The log-
ical way of doing things, edited by Karel Lambert (New Haven: University
Press), 2 3 - 3 7 .
Dingwall, W.
1964 "Transformational generative grammar and contrastive analysis", Language
learning 1 4 : 1 4 7 - 1 6 0 .
Di Pietro, Robert
1968 "Contrastive analysis and the notions of deep and surface g r a m m a r " , M o n o -
graph series on language and linguistics, 19th round table, edited by James
Alatis (Washington, D.C.: Georgetown University Press), 6 5 - 8 0 .
Edmondson, Willis
1980 "On saying you're sorry", unpublished MS.
Fisiak, Jacek
1975 "The contrastive analysis of phonological systems", Kwartalnik neofilo-
logiczny 2 2 : 3 4 1 - 3 5 1 .
Fraser, Bruce
1978 "Acquiring social competence in a second language", RELC journal 1 1 : 1 - 2 1 .
Fraser, Bruce - E. Rintell - J. Walters
1979 "An approach to conducting research on the acquisition of pragmatic com-
petence in a second language", Discourse analysis in second language acquisi-
tion, edited by D. Larsen-Freeman (Rowley, Mass.: Newbury House), 2 2 - 3 7 .
Gazdar, Geral
1979 Pragmatics (New York: Academic Press).
Haberland, Hartmut - Jacob L. Mey
1977 "Editorial: linguistics and pragmatics", Journal of pragmatics 1:113.
House, Juliane - Gabriele Kasper
1979 "Politeness markers in English and German", unpublished MS.
Kando, Thomas
1977 Social interaction (St. Louis: The C.V. Mosby Company).
Katz, Jerrold, J.
1977 Prepositional structure and illocutionary force (New York: Thomas Y. Cro-
well).
Krzeszowski, Tomasz P.
1979 Contrastive generative grammar (Tbingen: Gunter Narr Verlag).
Moulton, William
1968 "The use of models in contrastive linguistics", Monograph series on language
and linguistics, 19th annual round table, edited by James Alatis (Washington,
D.C.: Georgetown University Press), 2 7 - 3 8 .
Oleksy, Wiestaw
1979a "Felicity conditions and questions", paper read at the Departmental Seminar,
Department of General Linguistics, University of Pittsburgh.
1979b Question in English and Polish. Semantics and pragmatics ( E d m o n t o n : Lin-
guistic Research, Inc.).
Riley, Philipp
1979 "Towards a contrastive pragmalinguistics", PSiCL 1 0 : 5 7 - 7 9 .
Rusiecki, Jan
1976 "The development of contrastive linguistics", Interlanguage studies bulletin
1:12-44.
364 Wiesiaw Oleksy
Sadock, Jerrold
1974 Towards a linguistic theory of speech acts (New York: Academic Press).
Sajavaara, Kari
1977 "Contrastive linguistics past and present and a communicative approach",
Contrastive papers, edited by Kari Sajavaara and Jaakko Lehtonen(Jyvskyl:
University Press), 9 - 3 1 .
Stalnaker, Robert
1972 "Pragmatics", Semantics for natural language, edited by Donald Davidson
and Gilbert Harman (Dordrecht: D. Reidel), 3 8 0 - 3 9 8 .
1980 "Review of Gerald Gazdar, Pragmatics", Language 5 6 : 9 0 2 - 9 0 5 .
Szwedek, Aleksander
1975 Word order, sentence stress and reference in English and Polish (Edmonton:
Linguistic Research, Inc.).
Wachtel, Tom
1979 "A question of imperatives", PSiCL 1 0 : 5 - 3 3 .
Walters, Joel
1979 "Strategies for requesting in Spanish and English structural similarities and
pragmatic difference", Language learning 2 9 : 2 7 7 - 2 9 5 .
JERZY RUBACH
Rule ordering in phonological interference
1 Introduction
Today it is a commonplace to say that rules of a native language cause inter-

ference in foreign language acquisition. In this paper we are interested in how
these interfering rules actually apply. In particular we want to look at the
problem of rule ordering.
Theoretically speaking one could assume either of the following three po-
sitions:
(i) in phonological interference rule ordering is irrelevant, i.e. native rules
are transfered into a foreign language at random orders ;
(ii) all ordering relationships are transfered from the native language to the
target language;
(iii) or an intermediate position: in the process of interference rules are ordered
partially, i.e. some orders are transfered and some are not.
Let us observe that each of these views carries different implications for the
theory of contrastive studies in general and the theory of language interference
in particular. Position (i) implies that it is necessary to merely discover the
rules of the native language in an adequate interference analysis. Their order-
ing need not be studied. Position (ii) is a much stronger claim: apart from
stating the rules one must also know their order of application to predict how
language interference will work. Position (iii) requires a still deeper study of
the native system: one must know the rules, their ordering and in addition
some sort of principle that could tell us which type of rules will and which
will not cause rule ordered interference.
In what follows we shall look at each of these three views in detail in order
to find arguments to choose between them. Our data come from language in-
terference in the speech of Poles learning foreign languages.
366 Jerzy Rubach
2 Unordered rules
Let us look at two rules in Polish: Surface Palatalization and Regressive De-
voicing. They are both low phonetic and exceptionless. Furthermore, they
both cause interference.
Surface Palatalization applies before /i j/ and it affects consonants both in-
side words and across word boundaries:
(1) [ + c o s ] ^ [ ^ c h k ] / - ( I - s e g ] ) |+3jh]
In the case of velars /k g x/ rule (1) effects a change in the place of articula-
tion since velars are fronted to the prevelar (postpalatal) position. We thus
have [k' g' x'] in:
(2) kilometr 'kilometre', gimnastyka 'gymnastics', histeria
'hystery', obtoki 'clouds', drogi 'roads', podstuchiwac
'listen in', krzak jezyny 'blackberry bush', dach i ofcna 'roof
and windows'.
In all other cases the place of articulation is not affected by Surface Palatali-
zation. Consonants become palatalized but they retain their original place and
manner of articulation. For example, /1/ changes to /t'/ but it is still a dental
stop. Rule (1) should therefore be kept distinct from the well knownmorpho-
phonemic process of Anterior Palatalization (cf. Gussmann 1978 a, Rubach
1981). The latter changes dentals into palatals, for example, /t/ becomes an
alveolo-palatal affricate [tp] as in kat 'hangman' kaci (pi.).
In (3) below we list some examples showing how Surface Palatalization af-
fects labial and coronal consonants:
(3) / p b m f v / ^ [ p ' b ' m ' f ' v ' ] :
pisk 'scream',pierwszy 'first', biuro 'office', mieszkac live',
fiotek 'violet', witio 'wine';
/t d sz/ [ t ' d ' s'z']:
tiara 'tiara', diabet 'devil', sinus 'sine', zjawisko 'phenom-
enon', brat Ireny 'Irene's brother';
/I r n/ - [ ']:
litosc 'mercy', rizotto 'risotto', dzban jagod 'jar of black-
berries';
/ / . rv' / / jvf-i
/s c dz/->[s c dz J:
Chirac (French name), imazinizm 'imaginism', Chile 'Chile',
dzinsy 'jeans', pokaz je 'show them'.
Thus Surface Palatalization is an allophonic type of process.

Regressive Devoicing is a well known morphophonemic rule of Polish.
Obstruents are unvoiced in the following environments: (a) before a pause,
(b) before voiceless obstruents inside words and across word boundaries and
(c) in the dialects of central and northern Poland also before sonorants but
only at word boundaries:
Regressive Devoicing affects stops, fricatives and affricates:
(5) (a) brzeg 'shore',giaz 'stone', brydz 'bridge';

(b) hrabia 'count' - hrabstwo 'county', tawa 'bench'
lawka (dimin.), idziemy 'we are going' - idzcie 'go'
(imper. pi.), wozy 'carts' - woz stomy 'cart-load of
straw';
(c) zawody 'professions' - zawod lekarza 'medical profes-
sion', zawod onkologa 'oncological profession', garaze
'garages' - garaz nasz 'our garage', wtadze 'authorities'
- wtadz miejskich 'city authorities' (gen.), etc.
As has already been mentioned, both Surface Palatalization and Regressive

Devoicing cause interference in the acquisition of foreign languages by Poles.
Thus English consonants are incorrectly palatalized in words such as people
*[p'], teach *[t'], leak *[1'], history *[x'], take it *[k'], etc. Rule (4) is re-
sponsible for incorrect voiceless obstruents, for example, *[p] and *[t] in
obstacle, obscene, read, read slowly, read loudly, etc. Words such as kid when
used in the context which meets rule (4) are mispronounced as *[k'it]: inter-
ference from Surface Palatalization and Regressive Devoicing.
In he can read easily the d of read is mispronounced as *[t'], i.e. both in-
terfering rules apply. The question is in what order they apply. Observe that
either order will give the same result:
(6) (a) *rid# # i- (b) *rid # # i-

rit # # i Devoicing (4) rid'# # i Surf. Pal. (1)
rit' # # i Surf. Pal. (1) rit' # # i Devoicing (4)
One is therefore led to the conclusion that in phonological interference rule

ordering is irrelevant. In other words, it seems that we have a transfer of rules
but no transfer of their order of application.
368 Jerzy Rubach
3 Ordered rules
Let us now consider some more complex examples of phonological interference.

In particular, let us look at mistakes which result from the transfer of two
other rules of voice assimilation: Regressive Voicing and Progressive Devoicing.
Apart from Regressive Devoicing, Polish has a rule of Regressive Voicing: 1
(7) [+obstr]- [+voiced] / - ([seg]) [ ^ f c e d ]
Rule (7) refers to the dialects of central and northern Poland and it expresses
the generalization that obstruents are voiced before voiced obstruents both
inside words and across word boundaries:
(8) liczyc 'count' - liczba 'number', prosic 'to request'
prosba 'request', skonczze 'DO finish', tap go 'catch him',
gtos dziecka 'child's voice', sprzgt wojskowy 'military
equipment', etc.
Thus, for example, the // of liczyc 'to count' is replaced by /d2/ in liczba
'number' due to the voicing environment of /b/.
At the same time Polish has a rule of Progressive Devoicing:
(9) [ + o b s t r I - [-voiced] / [-voiced] -

I +continJ
Rule (9) applies only inside words and it affects fricatives:
(10) bitewny [v] 'war-like' - bitwa [tf] 'battle'
modlitewnik [v] 'prayer-book' - modlitwa [tf] 'prayer'
panstewek [v] 'states'(dimin.,gen.) panstwo [tf] 'state'
rzodkiewek [v] 'radishes'(dimin.,gen.)-rzodkwi[ki] 'radish'
(gen.)
pochewek [v] 'sheaths' (dimin., gen.) pochwa[xi] 'sheath',
etc.
Progressive Devoicing interacts with the rule of morphophonemic Anterior
Palatalization. The latter derives /z/ from underlying //r// 2 in the environment
of front vowels: 3 bar 'bar' barze (loc.), rower 'bicycle' rowerze (loc.),
biodro 'hip' - biodrze (loc.), kadra 'staff - kadrze (loc.), etc. If /2/ derived
via Anterior Palatalization happens to be preceded by a voiceless consonant
then it is further unvoiced to /s/: saletra 'saltpetre' -saletrze (loc,),musztra
'drill' - musztrze (loc.), wicher 'wind' - wichrze (loc.), kufer 'trunk' kufrze
(loc.), etc. We thus have the following derivations for the loc.sg. of bar 'bar',
biodro 'hip' and saletra 'saltpetre':
(11) barze biodrze saletrze

//-r+e -dr+e -tr+e//
2+e df+e tf+e Palatalization
t5+e Progr. Dev. (9)
Both Voicing and Progressive Devoicing are late and entirely exceptionless
rules. Voicing is clearly responsible for mistakes such as this *[z] value, Miss
*[z] Brown, a bit *[d] better and help *[b] David. Interference occurs also
internally: *[z] is pronounced in misbehave, misguided, etc. where the cor-
rect form is [s]. Observe that Voicing and Progressive Devoicing have partially
overlapping environments. The examples quoted so far all refer to the portion
of Voicing which could not be in conflict with Progressive Devoicing. The latter
cannot apply across word boundaries and it cannot affect noncontinuants.
Consequently, it could not produce *[f] for [v] in this value nor *[p] for [b]
in misbehave. However, all instances of word internal clusters of fricatives or
stops followed by fricatives are potential inputs to both rules. Thus English
svelte may, in theory, be mispronounced in two ways:
(i) as [zv] via Voicing
(ii)or as [sf] via Progressive Devoicing.
If our conclusion from section 2 that in phonological interference rules are
transfered at random orders is true, then both of these mispronunciations
should be current with Polish learners of English. Alternatively, only [zv]
should be attested if for some reason Progressive Devoicing could not cause
interference. Unfortunately for our theory none of these alternatives is true.
Progressive Devoicing does cause interference and what is more there is no pos-
sibility of mispronouncing svelte as *[zvelt]. Learners end up saying *[sfelt]
with perfect consistency, i.e. we have interference from Progressive Devoicing.
In English there are very few words that have clusters of obstruents of which
the second member is a fricative and which would disagree in the value for
voicing. There are, however, many relevant examples in German and Russian:
(12) (a) zwanzig 'twenty', zwei 'two', zwischen 'between',

Zwirn 'thread'; 4
(b) tvoy *your', kvartira 'room', Moskva 'Moscow', Sved
'Swede', tsvet 'flower', etc.
These data confirm our earlier observation: in both German and Russian Poles
pronounce [f] instead of the correct [v]. We thus have interference from Pro-
gressive Devoicing and not from Regressive Voicing though in theory both
could apply. At the same time Voicing does cause interference in contexts
which could not be affected by Progressive Devoicing: was *[z] wollen Sie
'what do you wish', konets *[dz] voyny 'end of war'. The question is why
370 Jerzy Rubach
speakers give preference to Progressive Devoicing over Voicing. As usual, the

answer should be sought in the structure of the speaker's native phonology.
Observe that all the words given in (10) as examples for Progressive De-
voicing could theoretically undergo Voicing since the second member in each
obstruent cluster is voiced at the underlying level. It must therefore be true
that Progressive Devoicing is ordered extrinsically before Voicing and the
former bleeds the latter. Furthermore, this ordering is transfered into foreign
languages, which ultimately leads to the mispronunciations under discussion.5
The derivation proceeds as follows:
(13) Pol. bitwa Pol. liczba Russ. tvoy
'battle' 'number' 'your'
/bitv+a/ /li+b+a/ /tvoj/
bitf+a - *tfoj Progr. Devoic. (9)
lidZ+b+a - Voicing (7)
b'itf+a 1'idZ+b+a - Surf. Pal. (1)
The word liczba 'number' is not affected by Progressive Devoicing since the
second member of the cluster is a stop. Neither can Progressive Devoicing ap-
ply if there is a word boundary. Compare Pol. gtos walki 'sound of battle'
with English svelte and this value:
(14) gtos walki svelte this value
II-S# # v - sv- -S# #v-//
*sf - Progr. Devoic. (9)
z##v - *z##v Voicing (7)
Summing up, our conclusion from section 2 is not true. In phonological inter-
ference rules are transfered along with their ordering. Random ordering of
Surface Palatalization and Regressive Devoicing shown by the derivation in
(6) is simply a consequence of the fact that also in Polish phonology the two
rules do not interact in any crucial way.
4 Partial ordering
The conclusion that rule ordering is transfered in phonological interference

leads to the following expectation: in cases where a given structure meets the
environment of two different rules, interference will always be caused by the
rule which is ordered earlier in the speaker's native system. In the light of this
generalization we propose to consider the interaction of the so-called 1st Velar
Palatalization and Surface Palatalization.
As is well known, velar obstruents in Polish are palatalized 6 in the environ-
ment of front vowels. Thus //k g x// are changed to / d i 5/ as shown by
the following examples:
(15) krok 'step' - kroczek (dimin.), mozg 'brain' mozdzek

(dimin.), dach 'roof - daszek (dimin.), krzyk 'shout' -
krzyczec 'to shout', sluch 'hearing' - styszec 'to hear', etc.
In many cases the palatalizing environment is concealed on the surface by the

operation of Retraction, the rule which changes /i/ to back nonround A/ Re-
traction applies in the environment after nonhigh coronal consonants (for
details see Gussmann 1978b and Rubach 1981). Naturally postalveolar
stridents / di 5/, which are the outputs of 1 s t Velar, constitute an environ-
ment for Retraction. We thus have surface [i] in, for example:
(16) krok 'step' kroczyc 'to step', strach 'fear' - straszyc
'frighten', miazga 'squash' - miazdzyc 'to squash', chlopak
'boy' chtopaczysko (augmentative), brzuch 'belly'
brzuszysko (aug.), sok 'juice' - soczysty 'juicy', puch 'fluff
- puszysty 'fluffy'.
There is no doubt that the suffixes causing 1 s t Velar in the words in (16) are
derived from underlying //i//, compare kroczyc 'to step'vs.ghipota 'foolish-
ness' - oglupic 'to fool', odnowa 'renewal' odnowic 'renew', etc. where
l/i/l surfaces phonetically since the environment of Retraction is not met.
In sum, 1 s t Velar applies before front vowels:7
(17) +obstr " -high ,
- c o r o n U Ucoron / - " 0
L+high J L+strid J L"backJ
The voiced postalveolar affricate /di/ derived from //g// via 1 s t Velar is further
changed to /2/ by Spirantization:
(18) waga 'scale' - wazyc 'weigh', ubogi 'poor' - ubozec 'be-
come poor', rg Tiorn' - rozek (dimin.), rtoga 'leg' -
nozysko (aug.) - nozyna (dimin.), snieg 'snow' sniezysty
(adj.) - sniezyca 'snow-storm', wilgotny 'moist' - zwilzyc
'moisten', skarga 'complaint' - skarzyc 'complain'.
Spirantization applies in the environment after sonorants, hence not in cases

such as mozg 'brain' - mozdzek (dimin.):
(19) +strid
+voiced
[+contin] / [+sonor]
-anter
-high
1 s t Velar is not only perfectly productive but also entirely exceptionless as a

rule (but see below). It affects all kinds of borrowings, no matter how well
established or very recent:
372 Jerzy Rubach
(20) dyftong 'diphthong' dyftonzek (dimin.), brelok 'breloque'

breloczek (dimin.), befsztyk 'beefsteak' befsztyczek
(dimin.), kielich 'goblet' kieliszek 'glass', Wartburg (car
make) Wartburzek (dimin.), pech 'bad luck' - peszyc
(verb), etc.
Now let us see how 1 s t Velar is ordered with respect to Surface Palatalization,
rule (1), discussed in section 2. There is plenty of very compelling evidence
to show that 1 s t Velar must precede Surface Palatalization. The arguments
come from various interactions with a number of other rules in Polish. In this
paper we have introduced briefly only one of the relevant rules, Retraction,
so let us limit our discussion to the interaction between 1 s t Velar, Retraction
and Surface Palatalization 8 .
Recall that Retraction changes /i/ to [i] after nonhigh coronal consonants.
It enters therefore into a feeding relationship with 1 s t Velar which produces
/ df 5/ from //k g x//, i.e. noncoronals are changed into nonhigh coronals and
1 s t Velar feeds Retraction. At the same time Retraction bleeds Surface Palat-
alization since it destroys the environment for the latter by deriving / i / . Un-
like Surface Palatalization, Retraction is constrained to apply only inside
words, i.e. it is blocked by word boundaries. Let us look at the derivation for
kroczyc 'to step', the verb derived from krok 'step' by adding the verbalizing
suffix I/i/l and the infinitive ending //tp//. This derivation is contrasted with
that for palacz i 'smoker and', where I III I is the underlying segment and
forms part of the suffix -acz added to the verbal stem of palic 'to smoke':
(21) kroczyc palacz i

//krok+i+te pal+a# # i//
kroi+i+tp - ist Velar (17)
kroi+i+tp - Retraction
pal+a' # # i Surf. Pal. (1)
Clearly the rules must be ordered as in (21). Otherwise we would not be able
to derive [i] in kroczyc 'to step'. Had Surface Palatalization applied before
Retraction, the /5/ of kroczyc 'to step' would have been made [+high] and
Retraction would not have applied: *[kro'it(5], In sum, 1 s t Velar precedes
Surface Palatalization.
Let us now return to the facts of phonological interference. Recall that in
section 3 we arrived at the following generalization: in the process of inter-
ference, rules are transfered along with their ordering and the rules which are
ordered early in the native system apply as early in phonological interference.
If this generalization is true, then words such as taking, leaking, cracking, etc.
should have a tendency to be mispronounced by Poles as *[tejii]]) *[ticiij],
*[kreCiij ], i.e. the interference should come from 1 s t Velar. Unfortunately
for our theory such mistakes are not on record. Worse, they are impossible.
Poles do have interference in taking, leaking, cracking, etc. but it is the inter-
ference from Surface Palatalization: [k] is replaced by [k'] in the environment
of /i/. Thus the true generalization about rule ordering and phonological inter-
ference must be different. We have missed something of real significance in
our theory. As has been the case before, the answer should be sought in the
structure of Polish phonology and phonological theory.
It has been recently proposed (cf. Mascaro 1976, Halle 1978, Kiparsky,
forthcoming, and Rubach 1981) that all phonological rules fall into two
classes: cyclic rules (in the new sense, see below) and postcyclic rules. Cyclic
rules as a bloc precede postcyclic rules so that in a derivation we cannot have
any intermixing of these types of rules. The details and the mechanics of
cyclic phonology need not concern us here (for discussion see the references
above). However, the basic idea is necessary.
A rule which is designated as cyclic applies in accordance with the prin-
ciple of Strict Cyclicity (for a formal statement see Halle 1978). Most funda-
mentally, Strict Cyclicity incorporates the notion of 'derived environment'
(in the sense of Kiparsky 1973), i.e. cyclic rules may apply only if either of
the following two conditions is met:
(i) there is a morpheme boundary
(ii) or a segment, which is crucial to the application of a given rule, has been
derived by another rule applying earlier on the same cycle.
Thus, if condition (ii) is not met, a cyclic rule cannot apply morpheme inter-
nally.
First Velar and Spirantization are cyclic rules. Their cyclicity follows from
the principle that cyclic and postcyclic rules form separate blocks. In Rubach
(1981) I have shown that both of these rules must apply before rules which
have been found to be cyclic (e.g. Anterior Palatalization, Lower), con-
sequently, 1 s t Velar and Spirantization must be cyclic themselves. The data
bear out his claim. First Velar is entirely exceptionless and 100% productive
at morpheme boundaries (the examples quoted earlier in (16), (18) and (20)).
However, it does not apply morpheme-internally. Thus borrowings undergo
1 s t Velar if there is a morpheme boundary but never if velars are followed by
front vowels inside the same morpheme:
(22) brelok 'breloque' - breloczek (dimin.), dyftong 'diphthong'

- dyftonzek (dimin.), and others quoted earlier in (20)
versus:
sugerowac 'suggest', agent 'agent', drogeria 'drugstore',
gigant 'giant', ginekolog 'gynaecologist', keiner Vaiter',
spiker 'announcer', kibic 'fan', kilof 'pickax', herbata 'tea',
374 Jerzy Rubach
chemia 'chemistry', hipoteza 'hypothesis', historia 'history',

etc.
The productivity of 1 s t Velar and Spirantization is best documented by their

application to words which are very recent borrowings, e.g. hotdog 'hot dog'
- hotdozek (dimin.).
Spirantization as a cyclic rule applies in accordance with condition (ii), i.e.
to segments which are derived, in this case via 1 s t Velar. Thus the /dZ/ which
comes from //g// is changed to ji/ e.g. waga 'scale' - wazyc 'weigh', rg
'horn' - rozek (dimin.) while the /dz/ which is present at the underlying level
(due to borrowing) remains unaffected: brydz 'bridge'.
In contradistinction to 1 s t Velar, Surface Palatalization is a postcyclic
rule. It applies freely without regard to the morphological structure of inputs,
compare the data quoted earlier in (3).
In Rubach (in press) I have considered in detail the problem of phono-
logical interference in the light of cyclic phonology. It has been concluded
that only automatic (exceptionless) context-sensitive postcyclic rules may
cause interference. 9 It becomes clear now why 1 s t Velar does not induce any
interference in taking, leaking, cracking, plugging, etc. First Velar is cyclic
and hence it is systematically excluded from the class of interfering rules. Ob-
serve that in theory, 1 s t Velar could apply to the examples just quoted. There
is no doubt that learners of English are aware of the fact that there is a mor-
phological boundary in taking, etc. Speakers are able to use the suffix -ing
productively: they learn to add it to verbal stems without any difficulty and
they can produce novel formations with -ing. Thus the prerequisite for apply-
ing 1 s t Velar, the existence of a morphological boundary, is fulfilled. Yet, the
rule does not apply. As has already been said, the generalization is that cyclic
rules do not cause interference.
We can now solve the problem which challenged our conclusion from sec-
tion 3: why it is Surface Palatalization and not 1 s t Velar that causes interfer-
ence in taking, etc. in spite of the fact that 1 s t Velar is ordered before Surface
Palatalization and the claim is that rule ordering is transfered in phonological
interference. First Velar is cyclic while Surface Palatalization is postcyclic,
hence only the latter may cause interference. The claim about the transfer of
rule ordering is perfectly correct. However, it has to be viewed in the light of
the generalization that cyclic rules play no role in phonological interference.
First Velar cannot, as a matter of principle, apply to taking, etc., hence Sur-
face Palatalization applies, producing incorrect *[k'].
The rules considered in sections 2 and 3 are all postcyclic, context-sensitive
and exceptionless. Therefore they may, and actually do, cause interference. It
is only in this class of rules that ordering plays a role in phonological transfer.
5 Conclusion
Our findings may serve as guidelines for contrastive phonological studies

whenever these claim to be oriented towards more practical ends such as for-
eign language teaching.
The first step in an endeavour to arrive at a coherent theory of phonolog-
ical interference is to carry out a theoretical analysis of the speaker's native
language and the target language. We can thus establish the relevant systems
of rules. As has often been stressed in the past, interference may occur if we
have an overlap in rule input or rule environments. However, it is also clear
that interference appears in cases where there are no obvious correspondences
between the rules of the native and the target language. For example, English
aspiration is known to cause transfer in the speech of English people learning
foreign languages which do not have any comparable rule of aspiration, e.g.
Polish. In such cases we would predict that interference may occur only from
automatic context-sensitive postcyclic rules (this claim has been substantiated
in Rubach, in press). We must therefore know which rules are postcyclic,
etc., and hence there is a need for theoretical studies. Only these can define
the class of rules which may cause interference even if there are no corre-
spondences between the rules of the native and the target language.
We have shown in this paper that rules are transfered along with their
ordering. Consequently, a theory of phonological interference must again lean
back on theoretical investigations as only they can identify various ordering
relationships among rules.
At the beginning of this paper we outlined three different views with re-
spect to the problem of rule ordering in phonological interference. It is clear
that only the third view is correct. In section 3 we pointed out that rule
ordering was essential for an interference theory. In section 4 we narrowed
down the transfer of rule ordering to the class of automatic context-sensitive
postcyclic rules. Thus we reject the two extreme views: (i) rule ordering is ir-
relevant for the theory of phonological interference and (ii) all ordering rela-
tionships are transfered from the native language to the target language.
Finally it is worth pointing out that our findings support generative phonol-
ogy in general and specifically its recent cyclic version as the proper frame-
work in which to carry out contrastive studies. A number of earlier papers (cf.
especially Fisiak 1976, Eliasson 1978, Gussmann 1975, Rubach, in press)
have already shown the inadequacy of phonemic framework for contrastive
analysis. We want to add one more argument to all those given earlier in the
literature: it is not only the notion of a rule but also the concept of rule order-
ing that is essential for contrastive studies. The classical phonemic framework
can incorporate neither and hence it should be rejected.
376 Jerzy Rubach
Notes
1. In Rubach 1974 I argue that the two should not be collapsed. For the purposes of
this paper the point is irrelevant hence we shall by-pass it in our discussion.
2 . 1 use double slashes for underlying representation, single slashes for intermediate
stages and square brackets for phonetic representation.
3. Actually, the change from //r// to // passes through an intermediate stage of palata-
lized r. Derivational details need not concern us here, for discussion see Rubach 1981.
4. As a matter of fact, [v] in these words as well as in English svelte is partially or even
completely voiceless. This, however, is irrelevant to the point that we are making: the
only important fact is that the in question is not identified with f in English and
German. The interference mistakes lie precisely in such an identification.
5. Incidentally, observe that language interference facts may also serve as an excellent
testing ground for various linguistic theories. The data quoted here show, for example,
that rule ordering is not an artifact of generative phonology. It is an absolutely neces-
sary concept in any phonological theory.
6. The term "palatalized" is used here in the phonological sense. Phonetically speaking,
/ df 5/ are nonhigh postalveolar consonants.
7. Also before /j/, see Gussmann (1978a) and Rubach (1981).
8. Other interactions are discussed in detail in Rubach (1981). Let us also point out that
all investigators have always assumed that 1 s t Velar is an early rule (cf. for example,
Gussmann 1978a).
9. This is true in cases of no contextual overlap between the rules of the native and the
target language (cf. Rubach, in press).
References
Elisasson, Stig
1978 "Theoretical problems in Scandinavian contrastive phonology", The Nordic
languages and modern linguistics 3, edited by J. Weinstock (Austin: The Uni-
versity of Texas), 2 1 7 - 2 4 3 .
Fisiak, Jacek
1976 "Generative phonology and contrastive studies", The Canadian journal of
linguistics 2 1 : 1 7 1 - 1 7 9 .
Gussmann, Edmund
1975 "How do phonological rules apply?", Papers and studies in contrastive lin-
guistics 3, edited by J. Fisiak (Poznan: A. Mickiewicz University Press), 113
-124.
1978a Contrastive Polish-English consonantal phonology (Warszawa: PWN).
1978b Explorations in abstract phonology (Lublin: M. Curie-Sktodowska Uni-
versity).
Halle, Morris
1978 "Formal vs. functional considerations in phonology" (Indiana University
Linguistics Club).
Kiparsky, Paul
1973 "Phonological representations", Three dimensions of linguistic theory .edited
by O. Fujimura (Tokyo: TEC Company), 3 - 1 3 6 .
(forthcoming) Lexical morphology and phonology.
Mascar, Joan
1976 Catalan phonology and the phonological theory (MIT Ph. D. diss, available
from Indiana University Linguistics Club).
Rubach, Jerzy
1974 Variability of consonants in English and Polish (University of Warsaw Ph. D.
diss.).
1981 Cyclic phonology and palatalization in Polish and English (Warszawa: Uni-
versity of Warsaw).
(in press) "Rule typology and phonological interference", Theoretical issues in con-
trastive phonology, edited by S. Eliasson (Stuttgart: Julius Groos Verlag).
KARI SAJAVAARA
Psycholinguistic models, second language

acquisition, and contrastive analysis
Maybe the things you want are like cards. You don't want them
for themselves, really, though you think you do. You don't want
a card because you want the card, but because in a perfectly
arbitrary system of rules and values and in a special combination
of which you already hold a part the card has meaning. But sup-
pose you aren't sitting in a game. Then, even if you know the
rules, a card doesn't mean a thing. They all look alike.
Robert Penn Warren, All the King's Men, Harcourt, 1946.
1 General considerations
Jacek Fisiak (1980:10) points out rightly that doubt concerning the validity
and usefulness of contrastive studies "results from a number of misinterpre-
tations and misunderstandings created by such factors as the peculiar metho-
dological status of CS, the lack of a clearcut distinction in the past between
theoretical and applied CS [. . .] and the lack of a precise formulation of the
different aims of theoretical CS and applied CS as well as the confusion of the
relationship between CS, the psycholinguistic theory of interference and er-
rors, and the theory of second language learning" (see also Sajavaara 1977).
It seems therefore justified to pay more attention than heretofore to the inter-
relationship between contrastive studies and psycholinguistic processes.
The link between foreign language teaching methodology and theoretical
linguistics has always been very close. Major shifts in linguistic theory have
been regularly reflected in language teaching methodology, which implies
that linguistic models have been associated with predominant psychological
conceptualizations. Audiolingual language teaching methodology was spon-
sored by structuralism through its link with behavioural psychology, and
cognitive-code approaches became popular in language teaching simultaneous-
ly with the heyday of generative theory. In the last few years, various ap-
380 Kari Sajavaara
proaches to communicative competence have been posited alongside gram-

matical competence (e.g. Canale - Swain 1980); this has involved a certain
change in the general attitude toward the question of what language is and
how it is used.
The teaching of grammar has always been one of the central activities in
language teaching. In many cases, knowing a language has been equated with
knowing the grammar of the language concerned. It is true that grammatical
competence is part of a human being's ability to communicate, but it is total-
ly insufficient to explain the phenomena that are involved in language behav-
iour. A speaker or a hearer only behaves "as if he knew the rules" (cf. Slobin
1979:99). It is understandable that the linguist's attention has been focussed
on various elements embedded in the linguistic code, but it is not self-evident
which categories and phenomena abstracted for a linguistic analysis can be
transferred over to the analysis of human communication. Moreover, the
lexicon has been seriously neglected in recent decades as a result of the auto-
nomous status of grammar.
The theoretical linguistic models which are available at present are insuf-
ficient for the description of second language speech processing, because lin-
guists' descriptions represent final products of language acquisition. For such
descriptions, languages have been treated as verbal codes without a link with
the dynamism of the contexts in which they are used for communication and
other purposes (cf. Sajavaara - Lehtonen 1980a, 1980b) and, moreover, the
learner is never exposed to the language in its entirety at once (James 1978).
In primitive conceptualizations of human perception, sensations like vision
and hearing were presented as straightforward printouts of the surrounding
physical world: what people saw was a picture imprinted on the retina, and
what they heard could be directly represented by the acoustic and physio-
logical phenomena involved in the transmission of sounds. In linguistics and
in considerations of the production and reception of language, the processes
that take place in the speaker and the hearer have been seen as mirror-image
hierarchical processes in which the speaker's message is turned into a chain of
acoustic phenomena, carried across a channel over to the hearer, and decoded
through another hierarchical process into a replica of the speaker's message.
All such representations involve one serious defect: both the production and
the reception are creative processes, and the establishment of communication
between the two interactants is based only partially on rules which exist in
the speech community and are available to its members through socialization
and language acquisition. As important as such rules are various negotiation
processes which are created ad hoc in each individual communicative situa-
tion. The linguist's description of the linguistic system functioning in such an
interactive process does not catch the creative aspect, the rules that are made
Psycholinguistic models 381
by participants. Even the linguist's representation of the 'established' features

of language behaviour may be misleading: the elements of the linguistic and
other texture present in a communicative situation obtain their meaning in
two ways: the linguist abstracts a function for them through his attempt to
'rulify' language behaviour, which sometimes results in seeing rules where
there are none, or language users develop subconsciously a feel for the lan-
guage system without there being necessarily much overlap between these two.
In the past twenty years we have been presented with a succession of new
and 'better' linguistic models. We can agree with Slobin (1979:31) that "all
these are partial and tentative attempts to map out part of the cognitive con-
figurations underlying verbal behaviour". Since the linguistic models are re-
placed continuously by new models, it may tentatively be concluded that
none of them represents the way in which a human being processes informa-
tion. An important question remains: What is the status of grammar in the
processing of messages?
First of all, the distinction between sentences and utterances should not
be forgotten. We can agree with Smith and Wilson (1979:64), who state:
Sentences fall within the domain of competence models, utter-

ances within the domain of performance models. Sentences are
abstract objects which are not tied to a particular context, speaker
or time of utterance. Utterances, on the other hand, are datable
events, tied to particular grammars, in the sense that a sentence
is not grammatical in the absolute, but only with respect to the
rules of a certain grammar; utterances, however, may cross the
bounds of particular grammars and incorporate words and/or
constructions from many different languages, or from no lan-
guage at all.
This quotation may repeat what is a commonplace among linguists and ap-
plied linguists but its full relevance has seldom been recognized. It relates to
Lyons's distinction between language-system and language-behaviour (Lyons
1977:26ff.). In contrastive analysis which aims at applied goals, research
based on sentences (the way most of the practice has been) may be totally
misleading, since we should actually direct our attention to utterances, which
are expected to reflect what there is in a language-system. Grammars are ab-
stract descriptions of system sentences, and we do not have grammars based
on utterances of natural language use. Therefore it may be difficult to work
out details of language contrasts for applied purposes.
Speech reception and production can only be approached through refer-
ence to interaction; these processes should be related to the exchange 'game'
which is going on between two or more interlocutors. All models that neglect
382 Kari Sajavaara
the influence of interaction on the processing of speech are bound to be de-

fective. The speaker makes use of several kinds of material for his message:
what he thinks the listener knows/believes/thinks; the listener's attitudes and
feelings as revealed in the situation; mutual agreement or disagreement on
attitudes and opinions; shared knowledge; and expected information. A num-
ber of universale are always involved here which derive from the general struc-
ture of the ways of thinking (Slobin 1979:64) and the fact that the inter-
locutors know that they want or do not want to communicate something and
it is in their common interest to build upon this assumption, which, again,
results in their willingness or unwillingness to understand each other.
The success of the transmission of messages depends on a multitude of
factors. Many things can go wrong, and it is only seldom that it is possible to
see for sure that the intended message has been interpreted exactly right by
the hearer. Fortunately, approximations are sufficient in most cases without
communication being seriously handicapped. Language is typically ambiguous
and fuzzy, and the idea that linguistic elements have fixed and universal inter-
pretations within one and the same speech community is an illusion based on
normative grammars and other similar aspirations of linguists.
The conceptualization of human speech communication has also been dis-
torted by reference to theoretical models of communication. What has been
said above clearly implies that any Shannon-and-Weaver-type models of com-
munication are insufficient as models of human interaction. The acoustic on-
line signal bears a certain part of the message only, and many other para-
meters have to be considered before we can reach a more accurate picture of
human language processing. Language serves definite purposes related to the
speaker-hearer's intentions in some specific time and environment. Thus,
utterances and other behavioural phenomena are necessarily situation-specif-
ic (cf. Lyons 1977:2729). In a communicative situation, the speaker and
the hearer are faced with a problem-solving task, i.e. reaching agreement on
mutual intentions. The speaker is expected to give a sufficient number of
cues, verbal and non-verbal, for the hearer to be able to reconstruct the mes-
sage. The verbal part of the message must be linked with the speaker's 'world'
and his expectations as to what takes place in the hearer as an active part-
icipant in the situation. The hearer for his part makes use of all the cues avail-
able for him not only in the verbal part of the message to create the most
probable interpretation of how he is expected to reconstruct his knowledge.
This reconstruction requires building blocks that come from the outside the
communicative chain as seen traditionally.
The limits of this paper do not allow for an overall discussion of psycho-
linguistic models or aspects of second language acquisition, a task which
would require a full-length monograph. As far as the models are concerned,
this is not really a drawback because much of earlier 'psycholinguistics' deals,

despite the name, with rules of TG grammar or their psycholinguistic relev-
ance, which need not be entered here. The present discussion will be focussed
on three questions which are considered to be relevant for second language
acquisition and contrastive analysis from a psycholinguistic point of view:
language processing, particularly its dual, data-driven and conceptually-driven,
basis, the distinction between hierarchical and heterarchical models, and the
role of grammar in language processing. The paper will be concluded with an
overview of some central aspects of second language acquisition, and certain
conclusions will be drawn for the purposes of cross-language analysis.
Discussion of the problems of second language acquisition and cross-lan-
guage phenomena involves a multitude of questions which relate t o the lin-
guistic code, on the one hand, and communication,human interaction, speech
processing as a whole, memory, and perception, on the other. All treatments
of the topic without any recourse to these areas are bound to be defective.
2 Speech processing: static and dynamic models
The description of speech processing should aim at a representation of the

entire communicative vehicle in which the reverse and shifting roles of the
speaker and the hearer and the perspective of society are included. A distinc-
tion has to be made between a linguistically oriented approach to the pro-
cesses of production and reception and a communicatively oriented one
(Sajavaara Lehtonen 1980a): (1) A linguistically oriented approach aims at
explaining how a linguistic representation results in speech output, or is
derived from speech input, and (2) a communicatively oriented approach de-
scribes how the speaker implements his communicative intentions, how the
hearer deciphers the speaker's intentions, and how these interact in particular
situations of communication.
Thus, the choice of the model to be used in reference to language process-
ing involves a choice between a static view of language and a dynamic one.
This choice can also be characterized as one taking place between a linguistic
description and a psycholinguistic one or between a structural and an opera-
tional modelling of the chain of speech (cf. Lehtonen 1978; see also Davis
1978, Levelt - Flores d'Arcais (eds) 1977:xiv). The structural model aims
it describing the language as a set of rules abstracted and idealized from lan-
guage data. The dynamic view presupposes that grammatical rules have to be
egarded only as descriptions of certain regular structures of language, which
'unction primarily as constraints, and not as models for the mental processes
384 Kari Sajavaara
in operation when people are speaking or listening to speech (cf. Clark

Clark 1977:191ff.).
One of the most misleading features of the static models is their hierarch-
ical construction (see Lea (ed.) 1980:236ff.). There are several reasons to as-
sume that linguistic information is not processed hierarchically by proceeding
step by step through the levels of grammar, for instance in perception, from
a concrete representation to phonology and syntax, and further to more ab-
stract meaning. Instead, the functioning of the perceptual process is to be
seen as 'heterarchical': the hearer makes use of all possible knowledge avail-
able to him, phonological, syntactic, pragmatic and other, simultaneously in
terms of a time-sharing system. What is particularly important is the fact that
the speech signal does not contain all the information necessary for recognition;
a great deal is inferred from elements external to the speech input or available
in the speaker-hearer's internal data bank. In message production, similarly,
the processes cannot be seen as steps leading from higher linguistic levels to
physical speech. In all processing, surface perceptual cues are important be-
cause the perceiver uses them to construct "functional propositional units"
(Marslen-Wilson et al. 1978:244). Therefore, more attention should be paid
to the role of certain surface features (cf. also Flores d'Arcais 1978, Carroll
et al. 1978). However, such structures can be expected to be particularly im-
portant as reveal aspects of psychological operations (Carroll Bever 1976).
In reception, the primary process is knowledge-driven; it is operating

simultaneously with the secondary, input-driven analysis (see Lindsay
Norman 1977:278ff.; cf. Clark - Clark 1977:219-220). An important part
of the system is a constant retroactive reworking effect: the interpretation of
message units (words or word sequences) is open to later changes on the basis
of later information. In most cases, the speaker-hearer is unaware of the pro-
cess; what he is able to perceive is the single final option. Thus, the recogni-
tion of speech does not proceed lineally from phone to phone. There is
plenty of experimental evidence of the fact that the information for the suc-
cessive segments of the speech chain and cues used for the identification of
single elements by the listener are spread out over several segments, or cues
for several successive segments can be collapsed into a single element. In
phonetics, for instance, information which occurs later in real time can have
an effect on the interpretation of preceding structures; in Finnish, for in-
stance, the phonological status of the first-syllable vowel cannot be established
before the second syllable is heard (Lehtonen 1970). Since perception is not
accomplished segment by segment as a linear process, it must be assumed that
the processes of segmental perception operate over stretches of speech which
are longer than single elements and integrate the cues spread out in real time
Psycholinguistic m o d e l s 385
into a Gestalt, a percept skeleton, which is used as a direct input to various

identification processes (see Lehtonen - Sajavaara 1980).
Similarly to perception, also speech production takes place in units larger
than segmental speech sounds, and it does not require a hierarchical model
either (cf. Clark - Clark 1977:224-225). In ordinary natural speech, where
we have exchanges between two or more interlocutors, the basic units of pro-
duction are mostly much smaller than sentences. For instance, the opening of
an utterance is usually cued by the previous utterance; the opening element
then functions as a grammatical, semantic, and pragmatic constraint for the
subsequent part of the utterance. When a 'syntactic' unit is opened, the
speaker may not have any idea of how it is supposed to be concluded. It is
obvious that we must presuppose an overall intention or goal for the speaker's
behaviour which is to be realized through more specific plans, but it seems
doubtful whether the planning goes beyond rather restricted stretches of
utterance which can be maintained in short-term memory. Most of this is af-
fected by two of the sets of constraints mentioned by Slobin (1979:64), the
processing constraints, primarily the temporal ordering and rapid fading of
the code, and the discourse constraints, the nature and goals of human inter-
action. An indirect proof of the fact that, even at the level of phones, speech
production is not hierarchical are certain types of transfer from the native lan-
guage in the speech of second language speakers; it is often disguised by vari-
ous dynamic patterns of speech such as stress, rhythm, intonation, and pausing.
The processes of perception are those that are more easily accessible to
testing. Perception is more important than production also because languages
are acquired via perception and the original reason for many errors can be
found in perception. In the discussion of second language acquisition, per-
ception should be given the primary position also from the communicative
point of view. The native listener can understand the learner's message, how-
ever distorted, because he can apply all of his linguistic and other knowledge
to the task, while the language student has to learn to understand the 'normal'
accent of his foreign interlocutor.
The first stage of the perceptual process is concerned with precategorical
or phonetic signal processing (see Lehtonen Sajavaara 1980), which has no
specific place in the information-processing mechanisms of the central nervous
system. Features detected in this process may or may not correlate to cate-
gories abstracted in phonological analysis. The information available from
this process varies in richness in relation to the clarity of pronunciation but
an unambiguous phonological identification can never be reached through this
process alone. The second step may consist of the construction of the 'word's'
phonological structure on the basis of phonetically analysed cue information,
but this step is optional. In normal fluent speech perception, meaningful
386 Kari Sajavaara
items, words in a very broad sense, are identified directly on the basis of de-
tected cue information which constitutes the Gestalt of the items. The pro-
cesses are illustrated by Figure 1 (from Lehtonen Sajavaara 1980). If, how-
ever, the word is difficult, i.e. it cannot be immediately recognized from the
acoustic cue information, phonological categorization is needed for identifi-
cation as an auxiliary strategy; it is always available whenever the cue pattern
of the input signal finds no matching items in memory. Phonological analysis
is also needed when it is necessary to make use of the information available
in the affixes of complex derivative forms. Both channels are open all the
time, but the budding phonological process simply fades away, if it is not
needed when a new input chunk arrives. Thus, the identification of the phono-
logical segment string is possible, but not obligatory. Some experimental evi-
dence for the existence of parallel phonological processes can be found, for
instance, in recent work by Foss and his associates (Foss Blank 1980, Foss
et al., 1980). Lamminmki's study is also very revealing as regards the dual
nature of the interpretation process. It was normal for his subjects to identify
['LINGUISTIC COMPETENCE'
I OR MONITORING OF GRAMMAR,
Figure 1.
A partial model of speech perception (from Lehtonen - Sajavaara 1980a). Meaningful
items, 'words' in a broad sense, are identified directly on the basis of cue features con-
stituting the Gestalt of the items. An optional process of phonological analysis is also fed
by the same cue pattern. The two processes are assumed to be simultaneous.
the test items on the basis of features other than those which belonged to the
phonological code; whatever the speech signal, the subjects 'heard' lexical
items that were familiar to them (Lamminmki 1979).
3 Grammar in speech processing
The communicative approach implies a relocation of grammar in linguistic

conceptualizations. Some recent work by Marslen-Wilson and Tyler (1980)
is an example of research in which various on-line interactive processes are
seen in a new light. In a dynamic model, grammar refers to such character-
istics of the communicative mechanism as make it possible for a native speaker
to produce a 'grammatically acceptable' text and make it possible for a native
speaker to interpret the information embedded in a text in a predictable way.
Such grammatical rules must be represented in neurobehavioural structure
(cf. Buckingham - Hollien 1978:294), which does not however mean that
such rules are neurobehavioural structures. In the processing of speech, the
function of lexical, syntactic, and phonological structures is that of con-
straints: they reduce the number of potential guesses available for the identi-
fication of the acoustic speech input.
Slobin writes (1979:59):
The psycholinguist is concerned with rules of language use. He
wants to account for such things as how long it takes people to
understand certain forms, what happens to them in memory, and
so forth. He is not concerned with an elegant description of the
language per se, but the language user [ . . .] Processing strategies
employ knowledge organized for use the particular use of
speaking and understanding. A grammar written by a linguist is
organized for a different purpose, namely, to effectively and effi-
ciently describe a language for purposes of analysis and study.
For instance, in structural linguistics the description was based on an atom-

istic method: the language code was divided into subsystems, and for pur-
poses of efficient analysis the variation and fuzziness found in real language
communication were annulled by means of a suitable level of abstraction.
Obviously there is nothing wrong with this as long as it remains an attempt at
linguistic analysis. When, however, such constructs and items are regarded as
basic elements for instance in language teaching, i.e. linguistic analysis is con-
sidered a solution for a psycholinguistic problem, an error which may turn
out to be rather grave is committed. The assumption that constructs of lin-
388 Kari Sajavaara
guistic description are relevant units in the production and reception of mes-
sages must be considered unfounded (Lehtonen 1980b).
Lehtonen (1980b) points out that since grammatical rules are formulations
of structural regularities, or of the speaker's knowledge of such regularities, in
the sentences of the language, "an adequate description of a grammar is im-
portant for psycholinguistics, too, because in order to be able to approach the
productive mechanism we first have to learn all about the structure of its pro-
ducts". What we have to forget in this context is the hierarchical constitution
of grammar. Lehtonen concludes:
My point on grammar and the choice of an appropriate theory of

grammar for psycholinguistics could be summarized as follows:
structural grammar (including TG and other 'generative' ap-
proaches) aims at the description of the regularities in the text or
linguistic product (or at the description of the speaker's com-
petence or intuitions of the speech delivery of the ideal speaker).
The object of a dynamic, psycholinguistic, or communicative,
grammar, on the other hand, should be to describe the way in
which linguistic knowledge participates in the processes of mes-
sage production and perception.
The interrelationship between grammar and lexicon is one of the most inter-
esting and problematic points in an interactive model of message processing.
The above discussion of phonological processing implied that lexical elements
bear an important function in such processing. It could be hypothesized ac-
cordingly that information is embedded in memory in the form of lexical
items, a kind of 'words'. In accordance with Marslen-Wilson and Welsh (1978:
58), such lexical memory items can thus be defined as intersections of various
procedures operating over a range of cognitive dimensions (cf. Morton 1979,
Ellis 1979). Such a 'word' implies a pretheoretical label for a processing unit;
it contains information concerning its use in utterances and its relationships
to other words within and outside utterances. A 'word' activates various pro-
cessing procedures through this information; such procedures reach from pre-
fabricated phrases and frequent word combinations over to grammatical con-
straints and semantic fields. What comes immediately to mind when this kind
of processing model is suggested is a left-to-right probabilistic model, which
was criticized as early as 1951 by Lashley (1951), and the Markov process,
a language 'automaton' which was given up by Chomsky as a finite state
grammar (Chomsky 1957). A grammatical view of sentence processing ob-
viously makes such models unacceptable; in a dynamic psycholinguistic mod-
el, in which 'words' are given a primary status and not understood in a narrow
sense, the Markov process becomes an alternative which is worth reconsider-

ing (Lehtonen 1980b).
Cross-language comparison implies that lexical items may be detected
through different processes in different languages (see Karlsson 1979, Lehto-
nen 1980a). In Finnish, a word contains a wealth of syntactic and semantic
information for the analysis of the entire message through the suffixes that
are attached to it, while in English the same information is found in inde-
pendent morphemes attached to the word or in the order of such morphemes.
When a speaker-hearer is processing the speech input in these two languages,
he is bound to look for totally different kind of cues, in Finnish the cue fea-
tures embedded in single words, in English the data involved in the relevant
order of the elements. Moreover, there is some evidence of a non-equal status
of different types of lexical elements; in Finnish much of the processing load
seems to centre on verbal items, while in English it is the nominal parts of the
utterance that bear the major burden. In Finnish the finite verb is known to
be relatively easily recognizable in the speech chain (cf. Karlsson 1979).
4 A model of message processing
In order to integrate human information processing with data from research

on cross-language problems and second language acquisition, the Finnish-Eng-
lish Cross-Language Project has developed a tentative model of message pro-
cessing (see Figure 2). 1 The leading principle in the model is the description
of the phenomena and processes which result in communicative success in ac-
cordance with the communicatively-oriented approach which was delineated
above.
A central position in the model is occupied by a network of different types
of knowledge essential for efficient communication. In addition to grammar
and lexis, these include discourse history, facts and beliefs, and information
about language-world relationships. The network is 'governed' by the Over-
lord, a decision-maker/problem-solver, which has access to all the sources of
information all the time. The functioning of the system is affected by three
sets of intervening variables described by the three boxes below the major
box, i.e. will and intention, affective and emotional variables, and perform-
ance potential. All these are, to a certain extent at least, open to external in-
fluence (see Sajavaara 1978a), What this means is that the communicative
performance of an individual is never constant but varies under the influence
of these factors.
390 Kari Sajavaara
Figure 2.
A model of message processing (from Sajavaara Lehtonen 1980a).
The Overlord makes use of the information heterarchically, therefore it is

not normally possible to predict what the pieces of information are that result
in the 'right' interpretation of the message. There are two simultaneous pro-
cesses, the data-driven process and the conceptually-driven one; both are
necessary. The Overlord's major task is to compare both the incoming data
and the data stored in memory and make decisions as concerns the most pro-
bable interpretation of the message - also as concerns corrections to earlier
interpretations which are made necessary by new information. The function-
ing of the Overlord is dependent on data available in the data bank, because
recognition of new information takes place in reference to previous know-
ledge.
The model is three-dimensional, the third dimension representing the con-
sciousness axis. The nature of the data available in the data bank varies from
fully conscious to totally subconscious: some data which are constantly used
for purposes of interpretation never reach the level of consciousness, some lie
at the borderline, and some data are based on explicit formulations of rules
and definitions (the borderline is never strict, which means that the whole is
to be seen as a continuum).
The 'buffer', which includes short-term memory, and the interface take
care of the dosage of the incoming flow of material into chunks which can be
processed by the system (see Ellis 1979). The interface is responsible for the
reduction of the incoming data (from sound waves and letter symbols) to the
extent that the buffer can handle it. The interface must carry this function
out, at least partially, on the basis of certain physical features of the incoming
stimulus as was pointed out above. A certain number of the cues embedded
in this stimulus are language-specific, others are universal (e.g. the informa-
tion that a stimulus is human speech). The interface and the buffer are neces-
sary both in the production and in the reception of speech. The processes
that are executed here are the reception and actual transmission of linguistic
data and the control of the articulatory organs and their motor programmes
and the hearing organs and processes.
All the information available in the data bank is of equal value and can be
used simultaneously by the cognitive mechanism materializing in the Over-
lord. The syntactic, morphotactic, and pragmatic rules as well as various
semantic networks which have been activated through the identification in
the incoming signal of a certain 'key element' together with the earlier results
of the progressing analysis are available at each stage of the process as con-
straints which open up or restrict potential choices in the construction of
message content or in the production of speech. The data-driven analysis is
necessarily an on-line process which proceeds word by word but, through the
retroactive reworking effect, any interpretation may be changed later in the
light of subsequent context. The Gestalts which are an important element in
the detection mechanism are used to make predictions as concerns the nature
of forthcoming data (see Lehtonen 1980a, Sajavaara 1980).
The dual nature of the processing system together with the general func-
tioning principle of the Overlord implies that most of the assumptions con-
cerning interference in various contrastive theorizings have been based on too
straightforward an idea of what takes place in the human brain. Yet the same
model also indicates that in some respects the contrastivist conjectures have
been too complicated in assuming that all the hierarchical 'linguistic' steps are
activated all the time in language processing. The partial model in Figure 1
above predicts interference on three levels of processing in cross-language
communication: detection of cue patterns, lexical identification, and recon-
struction of phonological target structure. Contrastive implications of the
model will be taken up later.
The production process is somewhat more complicated to reconstruct than
the reception process (here it is not necessary to take any stand as concerns
392 Kari Sajavaara
the existence or non-existence of two different systems for production and

reception; it is assumed here as a working hypothesis that in a fully developed
processing system it is one and the same mechanism which works in both
directions (see Tarone 1974, Ruder - Finch 1980). An attempt by Lehtonen
and Sajavaara to reconstruct the progression of communicative events from a
speaker's viewpoint can be found in Lehtonen 1980 b as reproduced in
Figure 3. The communicative event is based on the speaker's overt communi-
cative behaviour, both verbal and non-verbal, on the one hand, and on various
non-intentional features to be detected in the speaker and/or the situation, on
the other. It is here assumed that the motive for communicative behaviour
arouses various kinds of communicative plans simultaneously, and it is at the
execution stage that certain ones of these plans are deleted, e.g. in some cases
only verbal behaviour materializes, in some it is non-verbal, in some it is both.
It is also noteworthy that non-verbal behaviour or paralinguistic elements in-
volved in verbal behaviour determine the interpretation of utterances when
the verbal and the non-verbal features suggest different interpretations (cf.
Lyons 1977:6163). Despite the fact that the verbal part is linguistically
more central (see Lyons 1977:61), the pragmatic interpretation is more high-
MESSAGE PARLOUR
VERBAL
GOAL PLANNING ex-
'EVENT' : l or, ecution covert
.
IN THE ; i-U speaker s :ommuni-
WORLD will motor rative be
and plans shaviou&/
intention p l a n s for i n t e n -
tional non-ver- ]
bal gestures
and paralanguage
t h e speakers
affective & : ' non-
physiologi- inten-
cal state tional
.feature
L e h t o n e n & S a j a v a a r a 1980
Figure 3.
In this model of the progression of communicative events from the speaker's viewpoint
(developed by the Finnish-English Cross-Language Project, see Lehtonen 1980 b), the
communicative event is seen to be based on the speaker's overt communicative behaviour,
both verbal and non-verbal, on the one hand, and on various non-intentional features to
be detected in the speaker, on the other. In addition, the 'event' in the world also may
add its own elements to the communicative event.
ly dependent on the non-verbal part, which is more closely associated with

the social and expressive function of language.
It is important to remember that a model of this kind is not meant to be a
scientific description of how the human brain functions (similarly the size of
each individual box bears no resemblance to the relevant factor's importance
in human speech processing). The modelling of the processes and phenomena
is valuable for the illustration of what is regarded as important in human
speech communication, on the one hand, and also of the structuring of re-
search problems and starting-points for experimental research. It would be
misleading to suggest that the 'black box' could be described by means of a
model (and in this sense the present model cannot be considered to be any
better than all other models that are available), but this fact does not under-
mine the reasoning behind the building of the model presented here.
5 Second language acquisition
Language competence was earlier mostly understood to mean internalized

knowledge of the verbal code: this is the inference we can make on the basis
of language teaching materials. Today it is perhaps much more difficult to
make statements as to what it means that somebody knows or does not know
a language. Despite a pronounced emphasis on communication in language
teaching, nothing much has changed in actual practice: in many cases 'com-
municative' language teaching aims at the production of grammatically well-
formed sentences on the basis of the recognition and segmentation of ele-
ments and phenomena abstracted by linguists. If communicative competence
means, however, the ability to use the mechanism described above in an ac-
ceptable way, the function of linguistic elements changes radically. Formal
teaching obviously leads to knowledge about grammatical constraints in the
form of generalized rules, which results in what could be termed as 'pseudo-
competence', while true communicative competence can be reached through
natural communication only (cf. Krashen 1978:15ff.; Widdowson 1979:246).
In language acquisition, the acquirer builds up his language system on the
basis of the input of concrete representations that is available to him. The
process works from the concrete to the abstract. In language teaching prac-
tices, the case has normally been the reverse. The learner has been provided
with a set of abstractions, a number of isolated rules at relatively autonomous
levels of description, which he is then supposed to be able to use as a basis for
the creation of concrete representations. Resulting pseudo-competence may
allow for the simulation of acceptable language behaviour in favourable cir-
cumstances (see Sajavaara 1978a:5960).
394 Kari Sajavaara
From a cross-language point of view, the major research tasks centre on

two primary areas (Sajavaara - Lehtonen 1980 a): (1) research on cross-lan-
guage interaction to find out how learners develop the ability to produce the
right kind of cues, linguistic and non-linguistic, and the ability to interpret
the cues correctly; and (2) research on message processing in the language
learner to deal with the phenomena which are outside the scope of conscious-
ness. For the purposes of foreign language teaching, the results of this research
will have to be correlated to the learner in the formal teaching situation.
Several approaches to the theory of second language acquisition have been
developed recently. We have Krashen's Monitor Model (Krashen 1978), Schu-
mann and Stauble's Acculturation Model (Schumann 1978), and Selinker and
Lamendella's (1918) neurofunctional perspective (see also Schumann 1979).
Hatch's work on input (Hatch 1979) and Seliger's on high and low input
generators (Seliger 1977) have been important contributions as well as Fill-
more's findings concerning the relevance of social skills for language acquisi-
tion. There is also the model developed by Bialystok (1978) and the work on
universal tendencies by Wode and his collaborators (Wode 1979). Many re-
search programmes concerned with various aspects of bilingualism (e.g. the
Canadian immersion programme, the Heidelberg project, and the ZISA pro-
ject, see Cummins 1979, Dittmar 1978, Meisel 1980, respectively) or with
cross-language problems (see Faerch 1979, Sajavaara Lehtonen 1980b)
have produced data relevant for second language acquisition theory. At the
moment we have a great number of parameters which are known to be present
in the second language acquisition process; there is still, however, a long way
to go before we can have an integrated picture where the formal aspects of
language fit in with various developmental features. 2 If it is assumed that
everybody, at any age, can internalize languages (cf. Krashen 1978), what is
needed is knowledge about why the process does not always result in lan-
guage proficiency. It seems that many of the intervening variables are social
or in formal teaching situations pedagogical; there are some indications
of the fact that, without the influence of such intervening variables, the de-
velopmental processes might be fairly universal (see Seliger 1980, Meisel
1980).
Language acquirers who are successful in natural settings success mean-

ing rapid progress - are more concerned with communication than with form
(see Fillmore 1976); caretakers tend to pay attention to children's messages
and not to their grammar (Snow Ferguson (eds) 1977, Waterson Snow
(eds) 1978). Learning a second language means an expansion of existing com-
municative potential to the area of another code, and therefore the learner's
role in communicative interaction requires more attention.
Recent literature on second language acquisition (see Hatch 1977,Hakuta

- Cancino 1977) is mainly concerned with learners who are acquiring second
languages in natural settings. More data are needed about language acquisition
in formal classroom settings (see also Wode 1979). In some recent work on
the acquisition of second languages, a distinction has been made between con-
scious language learning (memorization of rules) and unconscious acquisition,
which results from exposure to the target language (see Krashen 1978,
Bialystok 1978, d'Anglejan 1978). Instead of a strict dichotomy we should
consider a continuum with two extremes: at one end there is 'total' acquisi-
tion, which results in native-like performance, and at the other end a perhaps
hypothetical language system which is based entirely on explicit memoriza-
tion of rules, sufficient for the production of acceptable L j strings under
favourable circumstances (cf. Sajavaara 1978 a). In addition to the type of
exposure to the target language, the age of the learner, the type of the rule
system, the level of acculturation, and the complex of factors involved in the
Lj - L j interrelationship also affect the outcome, to mention a few of the
most prominent parameters (see Gingras (ed.) 1978). Children before the
'critical' age mainly acquire, while older people can both acquire and learn,
although explicit memorization of rules and their application to practice
obviously becomes more difficult with age. In rule systems which relate to
semantics, notional categories, pragmatics, and sociolinguistics, learners must
rely more on acquisition than on learning, and there are rules which can only
be acquired because they cannot be specified (see Krashen 1978, Gingras
1978). Optimal acquisition requires a high level of acculturation and inte-
grative motivation, while a total lack of them may block acquisition entirely
(Schumann 1978).
Conscious learning is not necessary for acquisition; acquisition is not
necessarily the result of conscious learning. Formal teaching may develop an
ability in learners to use overtly formulated rules for various classroom prac-
tices, but they are not easily transferred outside the classroom (Lamendella
1979). Krashen explains the inability of learners to use explicit knowledge
about, the language for communication as a function of his Monitor (Krashen
1978); in a more general frame of reference, e.g. the one given in Figure 2
above, it could however be described in terms of the interplay between the
conscious and subconscious strata of the information-processing mechanism.
The two channels of the identification process described in Figure 1 above,
i.e. phonological mediation and direct lexical access on the basis of Gestalt
cues, correlate with Lamendella's (1979) notions of 'Foreign Language Learn-
ing' based on the cognition hierarchy of neurofunctional systems and
'Secondary Language Acquisition' based on the communication hierarchy, and
with Krashen's (1978) dichotomy of learned vs acquired rules. In fluent pro-
396 Kari Sajavaara
cessing, which results from acquisition, the synthetic Gestalt mode is the
regular one. Language learning emphasizes an analytical approach and gives
preference to processes which are dependent on phonological mediation or
similar procedures. This analytical mode of processing is often reinforced in
the available teaching methodologies. If the foreign language is approached
through detailed acoustic, physiological, and other similar phenomena, the
student may be misled to paying attention to factors which are of relatively
small importance rather than to more comprehensive units of speech and
communication. The processing of individual phonological items hardly causes
any serious problems if we find ways of teaching the learners to 'hear' the
foreign language input correctly. In most cases it is not necessary to pay spe-
cial attention to phonology.
In language contact, intake is regulated by a 'socioaffective filter' (see
Dulay Burt 1977), which is controlled by various social and personality
factors and knowledge of the world (Sajavaara Lehtonen 1980a). The filter
can be opened by integrating the learner psychologically or socially into the
target group or by adding to his knowledge of the world. Increased operation-
ality of the new language also lowers the filter. Conscious or subconscious
perception of what makes part of the world and the language is governed by
previous experience. In this way a great deal of what is important in commu-
nication can be transferred from first-language skills.
The teacher's main task in the classroom is the opening of the filter and
the provision of a sufficient amount of language input. It is important that
the language code is not an end in itself; it should be regarded as a product of
processes whose goals are non-linguistic. When the code was the main target
of teaching, it was not considered necessary to tell pupils anything about why
it was necessary for them to carry out certain procedures in the classroom. It
was sufficient that the teacher knew. For the pupil to know about the pur-
poses of the tasks was considered hazardous or even destructive for learning
(seen as habit-formation - the basic idea was obviously derived from behav-
iouristic psychology). Now it is understood that a learner's performance may
decisively be enhanced by his own view of his own chances of success, and
therefore it may be of some importance to tell the learner what is expected
of him. He may even be taught to learn. An integral part of this work in-
volves bringing language courses closer to reality as it is experienced by the
learner, closer to the world he lives in, and the environment he feels to be his
own. This means that it is not necessary for the language teacher to operate in
a vacuum without any link with what the learners are and what they do or are
planning to do outside the classroom.
Several attempts have been made during the past decade to break away
from the grammatical syllabus and to approach language teaching problems
through a closer specification of the aims within a different framework. There

are, for instance, Wilkins's (1976) notional syllabuses, further developed in
the Threshold Level by van Ek (1980), Munby's (1978) communicative syl-
labus design, and Freihoff and Takala's taxonomy of language-use situations
(Freihoff - Takala 1974). Although all of these reveal important parameters
of what language is and how it is used, the operationalization of such para-
meters does not seem to create a break out of the vicious circle of the gram-
matical code. As has been pointed out above, and also by Widdowson in
several contexts (e.g. in a paper read at the Second Nordic Conference of Ap-
plied Linguistics at Hanasaari, Finland, in November 1979), it is possible to
divide such parameters further and further into smaller units without really
assisting the learner in his task of learning how to communicate in the other
code or, preferably, how to extend his communicative potential over to the
area of the other code. The basic skills of communication are already there,
and it is exactly these skills that are often specified in the taxonomies. The
only way seems to be through using language for various tasks whose imple-
mentation requires language skills. It would seem that the objectives of lan-
guage courses should be specified as tasks to be carried out by learners, not
as notional and situational parameters.
The four 'basic skills' have attained an important place in many recent
theoretical presentations of language teaching methodology. In many cases
the four skills are strictly separated, which is seldom the case in true language
behaviour (plentifully illustrated by Widdowson (1978)). In speech commu-
nication the roles of the speaker and the hearer alternate. Reading is also a
kind of dialogue between the reader and the writer, which implies that writing
should be the same. In the classroom, every time the teacher writes something
on the blackboard it constitutes a reading exercise, whether it is the intention
or not. At a deeper level, in the light of the model presented in Figure 2 above,
the four skills are not much different from each other either; for instance,
writing always involves certain receptive processes, listening and reading are
interrelated, and perceptual processing entails phenomena which are present
in production, etc. Therefore, even if in some cases reading comprehension
can be given as a reasonable outcome of a course, it should not be specified as
the aim; instead it should be considered a means for some other goal, e.g.
gathering information about a certain concept or complex of concepts in a
certain language.
In language teaching the following criteria should be observed: (1) Lan-
guage and communication should be seen as a means, not as an end. (2) Lan-
guage input should be offered in contexts where it is relevant for the needs of
the learners. (3) Language classes should provide a framework for meaningful
communication in the target language geared to the experience of the part-
398 Kari Sajavaara
icipants. (4) The teaching should involve tasks of information-processing and

problem-solving which are intellectually demanding. (5) The courses should
be correlated to other task demands, e.g. other subjects. (6) Formal aspects
of the language code should be taught to the extent they can be supported by
the learning objectives (e.g. formal requirements of written texts). (7) Rele-
vant rules of discourse should be given proper attention in the classroom as
regards the extent to which they govern the choice of various language items.
6 Implications for contrastive analysis
Discussion of cross-language influences based on natural language data is seri-

ously handicapped if too simplistic an approach is adopted by the researcher.
It should always be remembered that there may be different kinds of influence
at different levels of learning and there is variation between learners. Lj in-
fluence on L 2 is never a matter of a simple yes/no question. At the same time,
a number of global characteristics can be seen to operate in language process-
ing and language acquisition (cf. Wode 1980, Meisel 1980). Various kinds of
L, influence are known to exist at all levels of analysis the foreign accent
is there but the tests available at the moment are insufficient to reveal the
'critical' points of the interlanguage speech channel. Even today the best way
to approach them is through analysis of the student's speech production, be-
cause it also reveals problems in perceptual processing.
As was pointed out earlier, the history of contrastive linguistics implies
that cross-language analysis based on a static view of language is insufficient.
What is needed in addition is mapping differences and similarities in various
processes taking place in and between the speaker and the hearer in acts of
communication (Sajavaara Lehtonen 1980 b). Since many interference pro-
blems can be expected to derive from the clash of the two dynamic informa-
tion-processing systems, rather than two static rule patterns, it is impossible
to predict interference phenomena without reference to speech processing
systems. This does not mean that classical contrastive analyses are useless; on
the contrary, what is implied by the above discussion of message processing is
that such analyses are valuable for the explanation of many phenomena in-
volved, but they can only seldom be sufficient alone for the purposes of for-
eign language learning. For a 'contrastive' analysis of the linguistic codes, the
codes should be located in their proper places in the speech communication
processes across the languages, which requires the mapping of the various
parameters delineated in the message-processing model above. In this way,
contrastive analysis expands to cover the entire chains of communication in
two or more languages (see Sajavaara 1977, Sajavaara - Lehtonen 1980a).
If there is no evidence for the psychological relevance of existing grammars,

i.e. they do not represent structures and processes embedded in the brain, it
cannot be expected that structures that form part of descriptions of grammar
are necessarily structures that are transferred from L, to L 2 . Transfer is a
psychological process and only what is psychologically real can be transferred,
as is rightly pointed out by Meisel (1980). Thus, for transfer the main pro-
blem is that of the psychological reality of grammatical structures. If abstract
grammars are "just convenient fictions for representing certain processing
strategies" (Lakoff Thompson 1975:285), how do we then explain the fact
that numerous analyses of foreign language learners' errors have proved that
grammatical structures are transferred from language to language. There
should also be a way to explain the discrepancy (not so surprising as such for
foreign language teachers) that exists between second language acquirers in
natural settings (e.g. Dulay Burt 1977, Meisel 1980) and foreign-language
learners who have access to L 2 data in formal classroom situations. Slobin
(1979:61) asks whether "grammars are - universally - constrained to take
certain forms because of the ways in which language must be used" and
whether constraints of perception shape grammar. Some studies concerned
with second language acquisition in natural situations (e.g. Keller-Cohen
1979) have indicated that learners' perceptions of L 2 are based on previous
knowledge. The same is also evident in other types of language learning such
as L j , L 3 , foreign-language teaching, and relearning (Wode 1980). Now the
main issue is the nature of such previous knowledge. What are the structures
and processes like which are embedded in long-term memory and which guide
perceptions? Here most FL learners and most natural language acquirers are
in different positions (for some major differences see Ringbom 1980) because
of the differences in the nature of relevant linguistic representations. Natural
language acquirers who are uninitiated in the metalanguage of abstract gram-
mars do not transfer categories that make part of linguistic representations,
while it can be predicted that problems inherent in L 2 can be found. Most FL
learners are exposed to L 2 through syllabi which are explicitly or implicitly
grammatically structured, or explanations of how the languages ( L j or L 2 )
'function' are given in the form of abstract theoretical representations (it
makes no difference which one of the existing models is used). After the
learners have been exposed perhaps only in the 'teaching' of the mother
tongue to such grammatical descriptions, they are able to 'perceive' such
categories. In this way, the interference from L j must be, to great extent,
teaching-induced and at least partially related to what Krashen (1978) calls
the Monitor. For example, one of the most persistent errors by advanced
Finnish learners of English is the wrong use of the passive (of the type *In
this paper it is tried to show that. ..), which seems to be mainly due to the
400 Kaii Sajavaara
fact that what is called the passive in Finnish grammars is not really a passive
in the same sense as the term is applied in the grammars of Indo-European
languages. A wrong equation between the two codes results in an overgener-
alization process in the target language.
Learning a new language means not only the acquisition of new categories
but also the reorganization of the criteria which are applied as cues for the
identification process of those new categories. The consequences which the
model of message processing sketched above has for a description of data pro-
cessing problems which the language learner encounters in FL communication
can be outlined only tentatively (cf. Lehtonen Sajavaara 1980).
The model allows for the following cross-language observations:
(1) Processing is based on the learner's previous linguistic knowledge, i.e.
the idiomacy of his Lj pronunciation and the corresponding receptive idio-
macy. Successful detection of L 2 phenomena is possible only if the speaker
has had a sufficient amount of contact with spoken L 2 . Because of the
universal nature of many features in speech, the cueing system seldom
fails totally in the case of a foreign-language learner, but if the system
resorts to cues which are irrelevant in the target-language system, incorrect
identification alternatives may hamper the functioning of the entire
mechanism.
(2) Interference in various paradigms and corresponding errors in identifi-
cation and production have been the object of most contrastive studies. Inter-
pretation of these data is still problematic and it is possible that the significance
of this interference has been exaggerated.
(3) Even if appropriate cue patterns are available for the word-identifica-
tion mechanism, the Overlord may fail to recognize the item because there is
a gap in the lexicon. Thus no lexical memory item can be activated on the
basis of the incoming cue pattern. It is also possible that the lexical item is
stored in a different form (as a graphemic representation without any cor-
responding phonetic shape or as a disguised morphological stem which does
not open all the necessary derivatives).
According to Krashen (1978), speech performance is always initiated by
means of the acquired system, while the learned system is available as a
Monitor to edit the output. In acquisition-poor environments, a speaker ob-
viously relies on his Lj competence as performance initiator. The string
is then 'translated' into an L 2 string, whose grammaticality and acceptability
depend on the availability of 'rules' and on the nature of the constraints pre-
sent. Under optimal circumstances, the L 2 string is initiated and processed
entirely on the basis of the acquired L 2 system without the Monitor being
activated. Most L 2 speakers are obviously located somewhere between the
two extremes: at least occasionally they have to rely on Li systems for speech
processing. This is the case when an L 2 unit has not been acquired when
there is a gap and the Monitor fails to give the right answer. L t influence
on surface strings may be due to the fact that (1) the string has been initiated
by the acquired Li system and the Monitor has not been able to correct the
string, (2) the Monitor lacks the correct 'rule' and an L j rule is used as a repair,
or (3) strings originally initiated by the correct L 2 system are mutilated by
the learned system (Sajavaara 1978b). The L j and L 2 acquired and learned
systems are closely interlinked in the light of the model presented above,
and both systems are referred to several times during speech production,
which may be one of the reasons for the variable performance by the same
speakers in different situations.
Second language acquirers may experience problems for several reasons
(see Sajavaara 1980): (1) A rule cannot be retrieved. (2) A rule is inaccurate
or wrong. (3) The speaker refuses to apply the rule entirely or in some situa-
tions (Krashen's (1978:1011) Monitor under-users are characterized by not
applying the rules even if they know them, and Faerch and Kasper's (1980)
strategy of formal reduction relates to instances where the insufficiency of the
linguistic system makes it necessary for the speaker to make do with "less of
their interlanguage repertoire than is in fact at their disposal"), or he wants to
avoid errors or facilitate speech production to reach higher levels of fluency
without distorting the message. (4) The rule has not been acquired at all (the
learner has not reached the stage where the rule is acquired, the rule has not
been taught, the rule cannot be taught because its exact formulation is not
possible or is so complicated that it is functionally unwieldy, or the learner is
unable to internalize explicitly formulated rules). (5) The language user exag-
gerates the importance of monitoring, i.e. the control of the processing
mechanism (Krashen's (1978:10-11) Monitor over-users), and makes an at-
tempt to apply explicit rules all the time. In L 2 , t h e problems are the greater
the further away we move from Lj-type language acquisition in naturalistic
situations, because the system can be expected to be more and more defective
and based on more explicit formal properties. It also becomes more delicate
and more liable to malfunction due to external factors, since more perform-
ance potential is normally required for second language processing.
In early contrastivist hypotheses, the distance between languages was seen
to be indicative of the degree of learning difficulty. The distance is important
in an indirect way: a learner's reactions to problems in L 2 processing may re-
flect his mental image of the potential correlation between Li and L 2 struc-
tures. This results in the speaker's judgments on what is transferable or not
(see Ringbom 1979, Sharwood Smith 1979:350).
In contrastive analysis aiming at applied goals, too much attention may
have been paid to the transfer of individual structures. According to Felix
402 Kari Sajavaara
(1978:224), L x interference seems to occur in the form of sporadic erroneous

performance;he concludes that the interrelationship between L j and L 2 lies at
a much deeper level than surface or even near-surface categories. He also calls
for a more global look at the role of previous Li knowledge in the acquisition
of L 2 , which is well founded in the light of what is known of human informa-
tion processing, perceptual mechanisms, and memory. Keller-Cohen (1979)
points out that learners seek global properties of language, i.e. large analytic
chunks which are linked with clear contexts. Various kinds of surface features
may be in different positions depending on the degree of optionality of the
items; the persistent problems experienced in pronunciation may be due to
the fact that phonetic choices are seldom optional.
7 Conclusion
One of the most important observations based on what has been said above
concerns the high degree of task dependency of speech production and recep-
tion. There are few factors that remain constant under all circumstances,from
speaker to speaker, and with the same speaker. The whole system aims at ef-
ficiency, which does not however mean the economy of surface features (cf.
Sajavaara 1978a:57-59).
At the onset of the second language learning process, the learner's cue
detection mechanism is tuned to the phenomena and processes of his first lan-
guage. He tends to hear the target-language utterances in terms of categories
and structures of his native language, and it is not surprising at all that he also
substitutes elements of his first language for the target structures. Unless
various plans and programmes are 'acquired', to use Krashen's dichotomy, the
speech processing mechanism has to resort to 'learned' items. This obviously
requires more processing capacity. Insufficient acquisition can be compen-
sated by increased capacity or low task demands. Other simultaneous tasks,
emotional states, and high loads on problem-solving capacity may also affect
the processing system.
One of the most important parts of the foreign-language learning process
involves a continuous reorganization and completion of the cueing mechanism,
a 'new deal' of the cards of the communication game. When acquiring the cor-
rect Gestalts of target-language items, the learner also tunes the perceptual
mechanism to pick up such Gestalts. The whole process is subconscious for
the most part, and it is impossible to see how it could be directly interfered
with. It might seem that the only way to this reorganization is through the
right kind of input, by exposing the learner to the kind of L 2 material which
brings about the shift as a gradual process. At the present moment we lack
adequate tests to cope with the problems in second language communication.

To a certain extent, this inadequacy results from the fact that there has not
been an integrated picture of the mechanisms functioning in first-language
communication or language acquisition. The whole problem boils down to the
major issue taken up at the beginning of this paper, the distinction between
static and dynamic modelling of communicative skills. An integrated picture
- whether in Li or in L 2 - is possible only if we remember the distinction
between the student's knowledge of linguistic structures and his capacity to
communicate in the mother tongue or in a foreign language. Most of the pre-
vious research has dealt with the final product, the surface string. There is an
urgent need to develop methodologies to study the stages before the actual
utterance. More information is also needed of discourse planning such as it is
revealed, for instance, by speech errors in L x and L 2 . Phenomena which will
have to be studied include language behaviour in various interactional settings,
ability to use different channels of communication to convey messages and
intentions, and ability to understand and interpret messages transmitted by
other interactants.
Notes
1 . 1 wish to acknowledge my debt to Jaakko Lehtonen, who has participated in the pro-
duction of the model. The model should be considered as a representation of the
problem points in language processing and an attempt to correlate linguistic elements
to other elements. It is not meant to be a representation of the flow of data in the
processing mechanisms: thus the arrows indicate relationships and channels for feeding
in data (see Lehtonen 1980a).
2. No attempt will be made here to deal with the rich literature on second language learn-
ing strategies (see, e.g., Faerch - Kasper 1980).
References
d'Anglejan, A.
1978 "Language learning in and out of classrooms", Understanding second and
foreign language learning, edited by J. C. Richards (Rowley, Mass.: Newbury
House), 2 1 8 - 2 3 6 .
Bialystok, E.
1978 "A theoretical model of second language learning", Language learning 28:
69-83.
Buckingham, . - H. Hollien
1978 "A neural model for language and speech", Journal of phonetics 6 : 2 8 3 - 2 9 7 .
Canale, . - M. Swain
1980 "Theoretical bases of communicative approaches to second language teaching
and testing", Applied linguistics 1 : 1 - 4 7 .
404 Kari Sajavaara
Carroll, J. . - . G. Bever
1976 "Sentence comprehension", Language and speech (= Handbook of percep-
tion 7), edited by E. C. Carterette and M. P. Friedman (New York: Academic
Press), 3 0 0 - 3 4 4 .
Carroll, J. . - . K. Tanenhaus - T. G. Bever
1978 "The perception of relations: the interaction of structural, functional, and
contextual factors in the segmentation of sentences", Studies in the percep-
tion of language, edited by W. J. M. Levelt and G. B. Flores d'Arcais (New
York: Wiley), 1 8 7 - 2 1 8 .
Chomsky, N.
1957 Syntactic structures (The Hague: Mouton).
Clark, . . - . V. Clark
1977 Psychology and language (New York: Harcourt).
Cummins, J.
1979 "Linguistic interdependence and the educational development of bilingual
children", Review of educational research 49:222 - 2 5 1 .
Davis, S. M.
1978 "Audition and speech perception", Bases of language intervention, edited by
R. L. Schiefelbusch (Baltimore, Md.: University Park Press), 4 3 - 6 6 .
Dittmar, N.
1978 "Ordering adult learners according to language abilities", Papers from the
First Scandinavian-German Symposium on the Language of Immigrant Work-
ers and their Children, edited by N. Dittmar, H. Haberland, T. Skuttnabb-
Kangas, and U. Teleman (Roskilde: Universitetscenter), 1 1 9 - 1 4 7 .
Dulay, . - M. Burt
1977 "Remarks on creativity in language acquisition", Viewpoints on English as
a second language, edited by M. Burt, H. Dulay, and M, Finocchiaro (New
York: Regents), 9 5 - 1 2 6 .
van Ek, J.
1980 Threshold level English (Oxford: Pergamon).
Ellis, A. W.
1979 "Speech production and short-term memory", Structures and processes (=
Psycholinguistic series 2), edited by J. Morton and J. C. Marshall (London:
Paul Elek), 1 5 7 - 1 8 7 .
Faerch, C.
1979 Research in foreign language pedagogy: the PIF project (= Anglica and
Americana 7) (Copenhagen: University of Copenhagen).
Faerch, C. - G. Kasper
1980 "Processes and strategies in foreign language learning and communication",
ISB 5 : 4 7 - 1 1 8 .
Felix, S. W.
1978 Linguistische Untersuchungen zum natrlichen Zweitsprachenerwerb (Mn-
chen: Fink).
Fillmore, L. W.
1976 The second time around: cognitive and social strategies in second language
acquisition (Ph. D. dissertation, Stanford University).
Fisiak, J.
1980 "Some notes concerning contrastive linguistics", AILA bulletin 2 7 . 1 : 1 - 1 7 .
Flores d'Arcais, G. B.
1978 "The perception of complex sentences", Studies in the perception of lan-
guage, edited by W. J. M. Levelt and G. B.Flores d'Arcais (New York: Wiley),
155-186.
Foss, D. J. - M. A. Blank
1980 "Identifying the speech codes", Cognitive psychology 1 2 : 1 - 3 1 .
Foss, D. J. - D. A. Harwood - M. A. Blank

1980 "Deciphering decoding decisions: data and devices "perception and produc-
tion of fluent speech, edited by R.A. Cole (Hillsdale, N . J . : Lawrence Erl-
baum), 1 6 5 - 1 9 9 .
Freihoff, R. - S. Takala
1974 A systematic description of language teaching objectives based on the speci-
fication of language use situations (= Reports from the Language Centre 3)
(Jyvskyl: Language Centre for Finnish Universities).
Gingras, R. C. (ed.)
1978 Second language acquisition and foreign-language teaching (Arlington, Va.:
Center for Applied Linguistics).
Hakuta, . - H. Cancino
1977 "Trends in second-language-acquisition research", Harvard educational re-
view 4 7 : 2 9 4 - 3 1 6 .
Hatch, E.
1977 "Second language learning", Bilingual education: current perspectives/lin-
guistics (Arlington, Va.: Center for Applied Linguistics), 6 0 - 8 6 .
1979 "Input studies" (Paper read at the First Nordic Interlanguage Symposium at
Hanasaari, Finland, August 1979).
James, C.
1978 "Quo vadis, psycholinguistics?", International journal of psycholinguistics
5:73-81.
Karlsson, F.
1979 "Automatic morphological segmentation of Finnish word forms", Papers
from the Conference on General Linguistics, Seili, 29.-30.8.1979 (= Publica-
tions of the Linguistic Association of Finland 3), edited by K. Hkkinen and
F. Karlsson (Turku: The Linguistic Association of Finland), 5 1 - 6 4 .
Keller-Cohen, D.
1979 "Systematicity and variation in the non-native child's acquisition of conver-
sational skills", Language learning 2 9 : 2 7 - 4 4 .
Krashen, S. D.
1978 "The Monitor Model in second language acquisition", Second language acqui-
sition and foreign-language teaching, edited by R. C. Gingras (Arlington, Va.:
Center for Applied Linguistics), 1 - 2 6 .
Lakoff, G. - H. Thompson
1975 "Introducing cognitive grammar", Proceedings of the First Annual Meeting
of the Berkeley Linguistic Society, edited by C. Cogen et al. (Berkeley, Ca.:
Berkeley Linguistic Society), 2 9 5 - 3 0 9 .
Lamendella, J.
1979 "The neurofunctional basis of pattern practice", TESOL quarterly 1 3 : 5 - 1 9 .
Lamminmki, R.
1979 "The discrimination and identification of English vowels, consonants, junc-
tures, and sentence stress by Finnish comprehensive school pupils", Papers in
contrastive phonetics (= Jyvskyl Cross-Language Studies 7), edited by
J. Lehtonen and K. Sajavaara (Jyvskyl: Department of English, University
of Jyvskyl), 1 6 5 - 2 3 1 .
Lashley, K. S.
1951 "The problem of serial order in behavior", Cerebral mechanisms in behavior,
edited by L. A. Jeffres (New York: Wiley), 1 1 2 - 1 3 6 .
Lea, W. A. (ed.)
1980 Trends in speech recognition (Englewood Cliffs, N. J.: Prentice-Hall).
Lehtonen, J.
1970 Aspects of quantity in Standard Finnish (= Studio Philologica Jyvskylensia
6) (Jyvskyl: University of Jyvskyl).
406 Kari Sajavaara
1978 "How can the theory and methods of speech sciences contribute to contras-
tive analysis?" (Paper read at the 16th International Conference on Polish-
English Contrastive Linguistics, Boszkowo, Poland, December 1978). Forth-
coming.
1980 a "Psykolingvistiska aspekter p en finnes frmaga att frst skandinaviska
sprk" (Paper read at the symposium Internordisk sprkfrstelse, Rungsted-
gaard, Denmark, March 1980). Forthcoming.
1980b "The communicative approaches to speech and language and the study of
grammar" (Comment paper read at the International Workshop on Psycho-
linguistic Models of Speech Production, University of Kassel, July 1980).
Forthcoming.
Lehtonen, J. - K. Sajavaara
1980 "Phonology and speech processing in cross-language communication", Theo-
retical issues in contrastive phonology, edited by S. Eliasson (Heidelberg:
Julius Groos). Forthcoming.
Levelt, W. J. M. - G. B. Flores d'Arcais (eds)
1978 Studies in the perception of language (New York: Wiley).
Lindsay, P. H. - D. A. Norman
1977 Human information processing (New York: Academic Press).
Lyons, J.
1977 Semantics 1 - 2 (Cambridge: University Press).
Marslen-Wilson, W. D. - A. Welsh
1978 "Processing interaction and lexical access during word recognition in con-
tinuous speech", Cognitive psychology 1 0 : 2 9 - 6 3 .
Marslen-Wilson, W. D. - L. K. Tyler
1980 "The temporal structure of spoken language understanding", Cognition 8:
1-71.
Marslen-Wilson, W. D. - L. K. Tyler - M. Seidenberg
1978 "Sentence processing and the clause boundary", Studies in the perception of
language, edited by W. J. M. Levelt and G. B. Flores d'Arcais (New York:
Wiley), 2 1 9 - 2 4 6 .
Meisel, J.
1980 "Strategies of second language acquisition", Wuppertaler Arbeitspapiere zur
Sprachwissenschaft 3 : 1 - 5 3 .
Morton, J.
1979 "Word recognition", Structures and processes (= Psycholinguistic series 2),
edited by J. Morton and J. C. Marshall (London: Paul Elek), 1 0 7 - 1 5 6 .
Munby, J.
1978 Communicative syllabus design (Cambridge: University Press).
Ringbom, H.
1979 "The English of Finns, Swedes, and Swedish Finns: some concluding re-
marks", The perception and production of English: papers on interlanguage
(= AFTIL 6), edited by R. Palmberg (Abo: Abo Akademi), 7 7 - 8 5 .
1980 "On the distinction between second language acquisition and foreign language
learning", AFinLA Yearbook 1980: Papers in language learning and language
acquisition (= Publications of the Finnish Association of Applied Linguistics
AFinLA 28), edited by K. Sajavaara, A. Rsnen, and T. Hirvonen (Jyvskyl:
AFinLA), 3 7 - 4 4 .
Ruder, K. F. - A. Finch
1980 "Toward a cognitive-based model of language production" (Paper read at the
International Workshop on Psycholinguistic Models of Speech Production,
University of Kassel, July 1980). Forthcoming.
Sajavaara, K.
1977 "Contrastive linguistics past and present and a communicative approach",
Contrastive papers (= Jyvskyl contrastive studies 4), edited by K. Sajavaara
and J. Lehtonen (Jyvskyl: Department of English, University of Jyvskyl),
9-30.
1978a "The Monitor Model and monitoring in foreign-language speech communica-
tion", Second language acquisition and foreign-language teaching, edited by
R. C. Gingras (Arlington, Va.: Center for Applied Linguistics), 5 1 - 6 7 .
1978b "The Monitor Model and contrastive analysis" (Paper read at the 16th Inter-
national Conference on Polish-English Contrastive Linguistics, Boszkowo,
Poland, December 1978). Forthcoming.
1980 "Second language speech production: factors affecting fluency" (Paper read
at the International Workshop on Psycholinguistic Models of Speech Produc-
tion, University of Kassel, July 1980). Forthcoming.
Sajavaara, K. - J. Lehtonen
1980a "Language teaching and acquisition of communication", AFinLA Yearbook
1980: Papers in language learning and language acquisition (= Publications of
the Finnish Association of Applied Linguistics 28), edited by K. Sajavaara,
A. Rsnen, and T. Hirvonen (Jyvskyl: AFinLA), 2 5 - 3 5 .
1980b "Prisoners of code-centred privacy: reflections on contrastive linguistics and
related disciplines", Papers in discourse and contrastive discourse analysis (=
Jyvskyl contrastive studies 5), edited by K. Sajavaara and J . Lehtonen
(Jyvskyl: Department of English, University of Jyvskyl), 7 - 2 6 .
Schumann, J.
1978 "The acculturation model for second language acquisition", Second language
acquisition and foreign-language teaching, edited by R. C. Gingras (Arling-
ton, Va.: Center for Applied Linguistics), 2 5 - 5 0 .
1979 "Three theoretical perspectives on second language acquisition" (Paper read
at the First Nordic Symposium on Interlanguage, Hanasaari, Finland, August
1980).
Seliger, . W.
1977 "Does practice make perfect?: a study of interaction patterns and L2 com-
petence", Language learning 2 7 : 2 6 3 - 2 7 8 .
1980 "Strategy and tactic in second language acquisition" (manuscript).
Selinker, L. - J. Lamendella
1978 "Two perspectives on fossilization in interlanguage learning",ISB 3 : 1 4 3 - 1 9 1 .
Shannon, C. - W. Weaver
1949 The mathematical theory of communication (Urbana: Illinois UP).
Sharwood Smith, M.
1979 "Strategies, language transfer and the simulation of the second language
learner's mental operations", Language learning 2 9 : 3 4 5 - 3 6 1 .
Slobin, D. I.
1979 Psycholinguistics. 2nd edition(Glenview, 111.: Scott,Foresman,and Company).
Smith, N. - D. Wilson
1979 Modern linguistics: the results of Chomsky's revolution (Harmondsworth:
Penguin).
Snow, C. - C. A. Ferguson (eds)
1977 Talking to children (Cambridge: University Press).
Tarone, E.
1974 "Speech perception in second language acquisition: a suggested model",
Language learning 2 4 : 2 2 3 - 2 3 3 .
Waterson, N. - C. Snow (eds)

1978 The development of communication (New York: Wiley).
408 Kari Sajavaara
Widdowson, . G.
1978 Teaching language as communication (London: Oxford University Press).
1979 Explorations in applied linguistics (London: Oxford University Press).
Wilkins, D. A.
1976 Notional syllabuses (London: Oxford University Press).
Wode, H.
1979 Studies in second language acquisition (= Occasional papers 11) (Singapore:
SEAMEO Regional Language Centre).
1980 "Language acquisitional universals - L j , L2, pidgins, and foreign language
teaching", AFinLA Yearbook 1980: Papers in language learning and language
acquisition (= Publications of the Finnish Association of Applied Linguistics
AFinLA 28), edited by K. Sajavaara, A. Rsnen, and T. Hirvonen (Jyvskyl:
AFinLA), 5 - 1 5 .
MICHAEL SHARWOOD SMITH
Learnability and second language acquisition
As a tool in applied linguistics, contrastive analysis (CA) has had a chequered

career. The most reasonable approach to date was arrived at round about the
beginning of the seventies: contrastive insights may help in the analysis of a
certain proportion of learners' errors. No more and no less than that. A strictly
limited role for CA has been reserved by some in the area of classroom pre-
sentation, that is, mainly with mature learners. There are however two new
trends in theoretical thinking that pose a challenge to established ideas. The
first is the increasing emphasis by certain linguists and psycholinguists work-
ing within an Extended Standard Theory framework on how linguistic theory
can contribute to an understanding of language acquisition. This trend will be
referred to as the learnability issue. The second trend, which will be related to
the first, relates to the recreationist (or 'creative construction') hypothesis.
This is the trend which carries some weight in second language acquisition
studies and which claims that the central concern of such studies should be
the universal processes that all learners share irrespective of language and in-
structional background and which, it is claimed, are essentially the same as
those underlying first language acquisition. The challenge is a twofold one. It
requires that acquisition should become a central concern in linguistic theo-
rizing thus implicating theoretical and descriptive linguists as well as experi-
mental and theoretical psycholinguists; it also sets up a direct link between (a
particular brand of) linguistic theory and what is widely assumed to belong to
the field of'applied linguistics', i.e. second language acquisition or 'interlan-
guage' studies. This at least is a logical consequence of linking up the two trends.
The learnability issue directly stems from an exploration of linguistic pro-
blems which has the aim of uncovering principles of universal grammar. This
is of course in the spirit of Chomsky 1965 (p. 24). A few years ago, people
who were interested in pursuing this line, especially psycholinguists, held
the view that universal grammar might be some set of substantial universale,
even a kind of universal base, and that specific languages might be derived
from this in terms of transformations. The rich transformational component
410 Michael Sharwood Smith
gave rise to such notions as the idea that children begin with deep structures
and then gradually acquire the appropriate transformations (see McNeill 1970
and White 1980 for discussion); a similar line of attack has been taken in
sociolinguistics where, for example, simple pidgins are compared with com-
plex Creoles. Again hypotheses were developed about language processing
whereby the transformational processes in the grammar were claimed to mir-
ror the routines carried out to produce and comprehend speech. Apart from
the fact that experimental evidence was not able to satisfactorily confirm this
idea of 'derivational complexity', the theoretical backing from linguistics has
been radically reduced since the revised versions of EST assume a much
simpler transformational component which can in principle generate a large
number of unacceptable surface structures. More general constraints are for-
mulated in various parts of the grammar in order to contain its power, such as
restrictions on phrase structure, like X-theory, proper binding in logical form,
and various mechanisms, like filters, for ruling out ungrammatical surface
structures (see White 1980:1 Iff.). Universal grammar is in fact a set of general
constraints on possible grammars rather than a kind of very deep base from
which all specific grammars may be derived. The term 'grammar' is therefore
somewhat misleading when applied to UG. These general constraints are what
the child must have a priori in order to arrive at an adequate grammar for the
data it is exposed to in such a short time, so runs the argument, and consider-
ing how degenerate that data are (see Hornstein - Lightfoot 1980). A child
must be able to rule out a large number of possible accounts, and linguists
seeking to account for linguistic data must make learnability a central criterion
in choosing between possible analyses. This criterion has been characterized
by Steven Pinker as extremely stringent; he defines a number of relevant con-
ditions that a theory would have to meet, namely the learnability condition,
the equipotentiality condition and the cognitive condition. These require that
the theory explain (see Pinker 1979:215):
(1) how languages are in principle learnable (LEARNABILITY)

(2) the mechanisms that enable the child to learn any particular
human language (not just one particular one)
(EQUIPOTENTIALITY)
(3) how learning can take place effectively within a very short
time span (three years for the basic components of language)
(TIME)
(4) how learning can take place given the kind of input data the
child is actually exposed to (INPUT)
Learnability and L j acquisition 411
(5) how learning can take place in accordance with empirical find-
ings in the study of child language (DEVELOPMENTAL)
(6) how learning takes place given the child's cognitive and per-
ceptual immaturity (COGNITIVE)
The analysis of bound anaphors provides an illustrative example of the linguis-

tic and psycholinguistic implications of the learnability approach (see Horn-
stein - Lightfoot 1980, White 1980:26ff.). In her thesis on grammatical
theory and language acquisition, White discusses an attempt by Matthei to test
constraints on anaphoric reference as illustrated in the following two sentences:
a) The chickens want the pigs to tickle each other.
b) The pigs said that the chickens tickled each other.
Children were asked to act out the examples and it was hypothesized that, as-
suming children have a priori 'knowledge' of the specified subject condition
(see a) and the propositional island constraint or 'tensed S' condition (see b),
they would not relate each other with the wrong NPs (Matthei 1981). The
normal input data, it is claimed, give the child no indication that each other
is a bound anaphor and that such constraints as the SSC and PIC must be
assumed if children are ever going to provide for themselves the correct
analysis of the structure in question. Unfortunately the children did relate the
chickens with the pigs and it seems as though they had some notion of reci-
procity as well as simple backwards-reference since some children had the pigs
tickling the chickens and also vice versa. This could be counter-evidence for
the innate status of the universal constraints in question; however, White
argues that this is not even counter evidence for the status of constraints since,
in order for the constraints to operate, the child has to recognize that 'each
other' is in fact a bound anaphor. At this point one might well ask if there is
any independent way of establishing that the child has recognized the condi-
tions for PIC and SSC to operate other than that in fact it applies these prin-
ciples correctly: the argument becomes circular. One can only then fall back
on the argument that enormous limits are imposed upon what the child can
show in its production of language because of perceptual and cognitive im-
matury, and that we must in fact imagine a moment when recognition takes
place after which the PIC and SSC then start to operate. These two constraints
are linguistically motivated in that they elegantly handle a number of lingustic
problems, including those mentioned above, and they are claimed to assist in
the psycholinguistic explanation of how children can cope with the problem
of coming up with appropriate analyses of the data on the assumption that
they may never have enough evidence in time to work the analyses out by
straight induction - Pinker's 'input' condition.
Chomsky has recently stressed that even the notion of acquisition is mis-
leading as a characterization of what the child achieves and he prefers the
term "attainment" which places more emphasis on the amount of a priori
knowledge that a child has (Chomsky 1980:9) and establishes more clearly
the role of primary data (exposure/experience) as a 'triggering' one. It is im-
portant to note that the principles of UG are not generally treated as 'impera-
tives': the child is not conceived as one who (unconsciously or intuitively, of
course) goes around looking for ways of applying them. Rather, once the con-
ditions for their application are part of the child's current grammar, the prin-
ciples simply emerge to take care of the data. This almost (ironically) has a
Piagetian ring to it, that is that there are apparently certain developmental
stages where some important restructuring of the current system has to take
place: the child has developed to such an extent that it ceases simply to assi-
milate everything but accomodates itself (adapts its internal representations)
to cope with what it experiences. 2 The actual means for doing this are given:
they are part of man's biological endowment. They lie dormant until the
right moment. Chomsky has of course disassociated himself from Piagetian
thought; there does seem to be some kind of parallel nevertheless. On the
other hand, some claims seem to imply that the child also has 'working as-
sumptions' about the nature of the specific language it is acquiring (or 'attain-
ing'). Take for example the notion of filters (see Chomsky Lasnik 1977).
These are designed to rule out certain overgeneralizations that might be
thrown up by the operation of core rules. In one account (discussed in White
1980:103 ff.) filters are represented as working assumptions which hold unless
positive evidence turns up to show that the filters are violated, in which case
the child may relax the constraint by dropping the filter. The filter in itself is
a complication or addition in UG but as a mechanism that is claimed to be a
typical feature of human language, it will simplify the task for the majority of
learners. This bears directly on notions of markedness. In this account a given
filter is assumed to be the unmarked case: this is the orientation that UG gives
the child in tackling the task of attainment. White (ibid.) also discusses Carol
Chomsky's work (Chomsky 1969) on the acquisition of verbs with regard to
to-complements like promise and ask.3 The former is marked in that it always
assigns control to subject: the subject of the complement verb is always the
subject of promise. The latter is marked in that it sometimes behaves like
promise (when it has the meaning of 'question') and sometimes like the more
usual (unmarked) type of verb (when it has the meaning of 'request') (White
1980:104), for example the verb tell which follows core grammar and assigns
control to the nearest NP (the NP in the complement of the matrix verb).
C. Chomsky's findings seem to support the idea that not only were unmarked
verbs mastered first but that the unambiguously marked verb, promise, was
Learnability and L2 acquisition 413
mastered before the partially marked verb,ask. White has this to say: "There
does seem to be a correlation between grammatical complexity and difficulty
of acquisition. [ . . . ] It would be a pity if degrees of complexity could not be
captured in some way by the concept of markedness" (1980:106).
This general observation concerning the difficulty of learning the more
particular rule holds, of course, even for theories of learning where little or no
a priori knowledge is assumed. Here we are to understand it in relation to the
notion of core grammar and indeed universal grammar as well.
The recreationist trend in second language acquisiton 4 studies centres
around the assumption that the acquisition mechanisms operating in primary
(first) language acquisition and touched on above do not atrophy but may in
principle still operate for non-primary languages as well. This thesis began in
one sense as a denial of the importance of mother tongue interference (see
Dulay - Burt 1975) and in another sense as an attempt to carry out the type
of morpheme studies pionieered by Roger Brown, and de Villiers and de Vil-
liers (see Brown 1973, de Villiers - de Villiers 1974) in the field of second
language acquisition. This has been done much more extensively now in the
latter field than in the former and a much quoted study by Bailey, Madden
and Krashen (1974) suggests a universal developmental order for young and
adult language learners (and Broca's aphasics) irrespective of the mother
tongues involved or the instructional backgrounds, as far as these grammatical
morphemes were concerned. Failure by second language learners to achieve
native-like grammars of the target language is no longer regularly attributed to
the supposed atrophying of the acquisition mechanisms that take care of first
language attainment. Various cognitive and affective factors are posited as
typical obstacles to achieving native speaker level.
White's concluding observations in her thesis where she considers the
interaction between acquisitional theory and the theory of grammar touch
upon limitations that must exist as far as what predictions can be derived from
grammatical theory. These limitations are indeed factors such as cognitive and
perceptual immaturity, limitations on memory and the like (White 1980:113).
Interestingly, Gass and Ard have recently taken a similar line to argue that
second language acquisition provides better opportunities for the investigation
into principles of UG than does first language acquisition 5 (see Gass Ard,
forthcoming). They claim that it is precisely the lack of the typical obstacles
mentioned by White and others (see for example Pinker 1979 and Wexler
Culicover 1980) that makes non-primary language acquisition, by mature
learners (presumably), a more attractive proposition. To back up this claim
Gass carried out an investigation into the acquisition of rules of relativization,
basing her study on first language work by Amy Sheldon (see Sheldon 1972,
1978) and the typological claims of Keenan and Comrie(see Keenan Comrie
414. Michael Sharwood Smith
1977, Comrie - Keenan 1979). Keenan and Comrie propose an 'accessibility

hierarchy' which in effect sets up an implicational relationship as follows:
SUBJ > DIR OBJ > INDIR OBJ > OBL > GEN > OBJ of COMPL
(the symbols have been expanded to make the meaning more transparent)
This should be read to mean that if a language allows relativization out of
a position as listed in the above hierarchy, it must also allow relativization out
of all positions to the left on the list. Keenan suggests that the leftmost posi-
tion indicates the easiest or most neutral position (Keenan 1976, Gass Ard
forthcoming). Again linguistic markedness is being suggested as an indicator
of psycholinguistic markedness. Gass and Ard discuss Sheldon's 1974 study
where correct responses were ranked for each of the following structural types:
SUBJECT:SUBJECT SUBJECT:OBJECT OBJECT.SUBJECT OBJECT:OBJECT
(the first item in the pair characterizes the Head NP in the matrix sentence;
the second refers to the NP rel in the subordinate clause). If the accessibility
hierarchy had been adhered to, so to speak, the SS and OS types should have
been easiest; in fact, the rank order was: SS 0 0 OS SO (roughly: 50%, 49%,
27% and 17%, respectively). Gass' study with second language learners who
were adults and came from typologically diverse language backgrounds, and
which dealt with twelve types of relativization, was in agreement with the ac-
cessibility hierarchy (see Gass 1979). Gass and Ard (forthcoming) formulate
their conclusions as follows: "Second language acquisition is a fertile ground
for testing theories about language acquisition and particular proposed uni-
versals".
It is clear that contrasting two or more languages is provided with a very
specific orientation if the learnability issue is accepted as a central concern.
It is also clear that investigation into non-primary language learning is going
to benefit directly from this research if learnability claims are linked up with
recreationist claims: this is particularly striking where notions of markedness
are the focus of interest. Gass and Ard's claim is a very important one in this
connection if indeed the enormous cognitive and perceptual obstacles that are
supposed to obscure first language data (spontaneous and experimental) prove
to be a serious problem in child language research carried out with learnability
in mind. On the other hand, one is allowed to wonder whether factors specific
to second language acquisition might also prove exceedingly problematic as
well if learnability was made the major concern in this second branch of in-
quiry. Wexler and Culicover (1980:7ff.) point to the limitations the first lan-
guage acquirer has and include the idea that the child has to handle limited
data at any one time (cf. Nemser's "blinding flash fallacy" in Nemser 1971).
The fact that the second language acquirer has already got a fully developed
Learnability and L2 acquisition 415
grammar (the L j ) has surely to be taken into consideration as well: this 'new'
a priori knowledge may play an important role as mediated by the learner's
perception of likely relationships between L j and L 2 (see Sharwood Smith
1979, Kellerman 1980). Where the target language is perceived as distant, the
universal processes may well operate. Where a close native-target relationship
is perceived, language transfer may predominate in such a way as to make
learners adopt marked structures in agreement with their own native language
(see, however, Eckman 1977 and Kellerman 1979) before correct unmarked
structures. This is an empirical question and one cannot rule out in advance
the possibility that learnability and recreationist claims wl contribute in
some measure to the advancement of knowledge in this area. Some will no
doubt prefer to hang on to the idea that contrastive analysis is first and fore-
most a useful, heuristic tool in descriptive linguistics and as a presentation
device in teaching as well as an aid in error analysis. At this stage in the game
it is difficult to know where to put one's money unless one option can be
ruled out a priori as logically absurd or simply unrealizable. There are certain-
ly indications that more thorough investigations are needed into the links be-
tween different kinds of acquisition on the one hand and linguistic theory on
the other. The evidence put forward to date is interesting and suggestive but
hardly adequate at the moment to draw any hard and fast conclusions. For
people in second language acquisition studies it should be interesting to adapt
and use Pinker's six conditions as a framework for comparing first and second
language acquisition in the context of (relevant) current work on transforma-
tional generative grammar. For theoretical linguists, it should be interesting to
see if cross-language observations and claims may have empirical support from
second-language acquisition studies. The interdisciplinary character of con-
trastive studies, particularly as carried out in Poland, should provide an ex-
cellent forum for these purposes. In other words, it should now be possible to
have what has not happened before, that is, a principled dialogue between
pure, theoretical linguists and people investigating second language learning:
problems of learning difficulty, of cross-language analysis and general typo-
logy are provided with a common, unifying framework.
Notes
1. Thanks are due to Paul van Buren, Eric Kellerman, Rob van Oirsouw and James Pank-
hurst for supplying comments to an earlier draft of this paper at rather short notice.
2. Chomsky's position still remains distinguishable from Piaget's of course in that the
former argues for linguistic universale and the latter preferred to see things in terms of
general cognitive universals: this at least was Piaget's established position.
3. Compare the following:

Bill told John to go; *Bill told to go (but note: Bill said to go)
Bill asked John to go; Bill asked to go (marked and unmarked uses)
Bill promised John to go;Bill promised to go (uniformly marked)
4. There are of course many lines of investigation in second language acquisition studies
other than the creative construction hypothesis.
5. Interestingly, Chomsky in Chomsky 1980a:33 refers also to the second language learner
in a discussion of the each other problem.
References
Bailey, R. W. - C. Madden - S. D. Krashen

1974 "Is there a natural sequence in adult second language learning?", Language
learning 2 4 : 2 3 5 - 2 4 3 .
Brown, R.
1973 A first language (Cambridge, Mass.: Harvard Press).
Chomsky, C.
1969 The acquisition of syntax in children from 5 to 10 (Cambridge, Mass.: M.I.T.
Press).
Chomsky, N.
1965 Aspects of the theory of syntax (Cambridge, Mass.: M.I.T. Press).
1980a Rules and representations (New York: Columbia University Press).
1980b "On the representation of form and function". Text of talk presented at
CNRS Conference, Royaumont, June 1980.
1977 "Filter and control", Linguistic inquiry 11:1 46.
Comrie, . - E. Keenan
1979 "Noun phrase accessibility revisited", Language 5 5 : 6 4 9 - 6 6 4 .
Dulay, . - M. Burt
1975 "A new approach to discovering universal strategies of child language acquisi-
tion", GURT 1 9 7 5 : 2 0 9 - 2 3 3 .
Eckman, F.
1977 "Markedness and the contrastive analysis hypothesis", Language learning 27:
315-330.
Gass, S.
1979 "Language transfer and universal grammatical relations", Language learning
29:327-344.
Gass, S. - J. Ard
forthcoming "L2 data: their relevance for language universals", TESOL quarterly.
Hornstein, . - D. Lightfoot
1980 Explanation in linguistics (London: Longman). Introduction preprinted in
GLOW newsletter 3 : 2 - 1 1 .
Keenan, . - B. Comrie
1977 "Noun phrase accessibility and universal grammar", Linguistic inquiry 8:
63-99.
Kellerman, E.
1979 "The problem with difficulty", Interlanguage studies bulletin 4 : 2 7 - 4 9 .
1980 "Predicting transferability from semantic space". Paper for international cir-
culation, University of Nijmegen.
Leamability and L 2 acquisition 417
Matthei, E.
1981 "Children's interpretation of sentences containing reciprocals", Language
acquisition and linguistic theory, edited by S. L. Tavakolian (Cambridge,
Mass.: M.I.T. Press), 9 7 - 1 1 5 .
McNeill, D.
1970 The acquisition of language: the study of developmental psycholinguistics
(New York: Harper and Row).
Nemser, W.
1971 "Approximative systems of foreign language learners", IRAL 9 : 1 1 5 - 1 2 3 .
Pinker, S.
1979 "Formal models of language learning", Cognition 7 : 2 1 4 - 2 8 3 .
Sharwood Smith, M.
1980 "Strategies, language transfer and the simulation of the second language
learner's operations", Languge learning 2 9 : 3 4 5 - 3 6 1 . Reprint of same in
ISB 4 (1979):66-83.
Sheldon, A.
1972 The acquisition of relative clauses in English, unpublished Ph. D. dissertation
(Austin: University of Texas).
1978 Assumptions, methods and goals in language acquisition research (Blooming-
ton: Indiana University Linguistics Club).
de Villiers, J. - P. de Villiers
1973 "A cross-selectional analysis of grammatical morphemes in child speech",
Journal of psycholinguistic research 2 : 2 6 7 - 2 7 8 .
Wexler, K. - P. W. Culicover
1980 Formal principles of language acquisition (Cambridge, Mass.: M.I.T. Press).
White, L.
1980 Grammatical theory and language acquisition (Bloomington: Indiana Uni-
versity Linguistics Club).
ALEKSANDER SZWEDEK
Some problems of contrastive analysis

and text linguistics
I want to begin this paper with a superficial statement that it has been admit-
ted for some time now that sentence grammars are not able to cope with cer-
tain language phenomena. There are linguistic facts within and between sen-
tences (such as pronominalization, ellipsis, thematic structure) that can be
accounted for only in the framework of a larger context. As in other kinds of
linguistic comparison there are two problems in connection with text analysis:
the problem of equivalence and semantic representation on the one hand, and
comparison of certain surface phenomena on the other hand. This paper will
discuss some difficulties connected with those two problems.
1 No satisfactory semantic representation has been proposed so far, but it is

certain that such a semantic representation will have to consist of a set of se-
mantic categories and relations (cf. discussion in Krzeszowski 1974:23ff., and
Krzeszowski 1974, Ch. III). Some of the semantic features would be gram-
maticalized in a particular language, i.e. expressed in a structured way (as a
subsystem), some would be present in the meanings of lexical units. If a se-
mantic feature is grammaticalized in two languages to a similar extent we say
that a structure in Lj is equivalent to a structure y in Lj (no matter whether
the surface structures are similar or not).
Thus systems of number in English and Polish would be more extensively
equivalent (general and uniform in both languages to a similar degree) than
passives. The passive structures of the two languages overlap only partially.
For the rest of the English passives we would have to specify additional con-
ditions (e.g. Indirect Object cannot be subjectivized in Polish, etc.). When a
semantic feature is grammaticalized in one language but not in the other (i.e.
when a semantic feature is expressed grammatically in language L, and lexically,
or partly lexically, partly grammatically, in language Lj, we conclude that the
language Lj does not have an equivalent to the structure in Li).
From the definition of semantic equivalence it follows that given a uni-
versal semantic representation what is worth comparing is everything that
420 Aleksander Szwedek
follows the semantic representation in the process of derivation, i.e. language

specific categories and rules which Chafe (1976) calls "packaging". Packaging
itself seems to be governed by two tendencies, described by Grice (1975) as
maxims of quantity and manner respectively:
a) a tendency to find the most adequate and understandable way of repre-
senting meaning,
b) a tendency to make this representation as brief as possible.
The first tendency does not require any comments. The second is realized,
among others, in substitution and ellipsis. In short, the process of "packaging"
must ensure brevity and at the same time coherence. Besides ellipsis and sub-
stitution, brevity also manifests itself in the choice of lexical items. Thus, in-
stead of describing an action as
(1) move one's fist in such a way that it comes into a violent
contact with some object
we use the verb
(2) hit.
In the process of "packaging" the speaker not only omits features of the
event that are irrelevant, but also omits those features of the event that he
assumes to be known to the listener. Thus instead of saying (3) or (4)
(3) Horace got some picnic supplies out of the car. Picnic sup-
plies include food and drinks; drinks include beer, etc. The
beer was warm.
(4) Horace got some picnic supplies out of the car, such as beer,
crackers,. . . The beer was warm.
we would rather say (5)
(5) Horace got some picnic supplies out of the car. The beer
was warm.
(1) and (3), (4) can be treated in a similar way, as paraphrases of (2) and (5)
respectively. At this point I would like to agree with Chafe (1970:8788)
that "in at least a great many cases [ . . . ] the surface structures do not, in fact,
have identical meanings but only very similar meanings", and "linguistic theory
should allow for the possibility that two surface structures, even when they
may convey identical meanings (meaning being taken in the sense of con-
ceptual manifestation) nevertheless may have different perhaps quite dif-
ferent semantic (or deep) structures".
What makes paraphrases seem identical is the same referent. However,what
is said about the referent and how it is said is what makes paraphrases differ-
CA and text linguistics: some problems 421
ent. It seems that equivalent structures can be compared in a way to para-

phrases. They have the same referent, the "packaging" may be very similar or
different. One of the difficult problems here is that even if paraphrases and
equivalent structures have identical meanings they may differ in presupposi-
tions. For example, (5) is described as having as one of its presuppositions the
sentence Beer belongs to picnic supplies, which in (3) and (4) appears in the
form of a sentence on the surface. If presuppositions can be expressed in the
form of sentences (cf. also van Dijk 1972:103: "Since presuppositions are al-
ways represented as sentences, we may consider the set of presuppositions,
followed by the sentence (s) presupposing them, to be part of a text. [ . . . ]
We assume that presuppositions are preceding sentences."), there is n o way, it
seems, in which we can distinguish presuppositions from sentences in the
semantic representations (at least not in the kind of semantic representation
proposed by Lakoff (1971)), besides the obvious fact that presuppositions do
not surface.
This difficulty can be resolved in two ways:
a) no distinction would be made between presuppositions and sentences,

which means that we would have to specify conditions under which some
sentences surface and some do not.
b) the distinction between sentences and presuppositions would be kept, in
which case we would have to characterize the differences between them at
the level of semantic representations.
It is obvious, that a) and b) are variants of the same material distinction be-
tween structures that do not surface and function as presuppositions, and
structures that surface and function as sentences. Since presuppositions seem
to play a crucial role in determining equivalence it is necessary to define their
status in the semantic representation more precisely.
2 Whichever way we resolve the problem of relation between sentences and

presuppositions, we have to assume that semantic representations of structures
that are considered equivalent must be identical at least to a degree that makes
them comparable as equivalent. From that point of view what is worth com-
paring is everything that follows the identical semantic representations in the
process of derivation. It would seem that the surface structure would be rela-
tively easy to analyze and describe. One of the problems, however, is that
semantic structure categories and relations often converge in the surface
structure, and one constituent may have more functions than one. In such a
case the interaction between two or more functions of the same constituent
makes it difficult to separate them and describe them clearly. This, I think, is
why descriptions of such pairs of concepts as theme/rheme, given/new, etc.,
crucial for our understanding of text formation, have been inconsistent and
vague.
The ambiguity can be traced back to Mathesius' (1939) description of the
structure of an utterance in terms of actual sentence partition or contextual
organization. Applying this description to Czech, Mathesius stated that word
order in non-emotive utterances directly reflects the actual partition. Every
sentence consists of a starting point which precedes a nucleus or core. The
starting point is defined as "that which is known or at least obvious in the
given situation and from which the speaker proceeds", while the nucleus is
"what the speaker states about, or in regard to, the starting point of the utter-
ance". At the same time, however, Mathesius described the theme ('zaklad') as
something "that is being spoken about in the sentence", and the core ('jadro') as
"what the speaker says about this theme". Thus in the contextual organiza-
tion of the sentence Mathesius identifies the starting point with known infor-
mation. As, however, has been pointed out time and again, the two need not
coincide, particularly in languages with grammatically determined word order.
As a matter of fact Mathesius seems to have described the starting point in
three ways, as a) known information, making no distinction between known
from the text and known from the situation, b) the first element of the utter-
ance from which the speaker proceeds, c) what is spoken about.
It must be remembered, however, that while identifying "the beginning of
the sentence" is a purely mechanical procedure, telling us nothing about its
role in the text, the phrase "what is spoken about" is vague and in some cases
may not be easy to determine. It is not clear that the notion "the first element
of the sentence" is necessary at all. As is well known the beginning of the sen-
tence in Slavic languages is associated with given information, while in lan-
guages like English with the subject. If we ignore the given information con-
cept in Slavic languages and the concept of the subject in English, what is left
is that the beginning of the sentence is the beginning. On the other hand, in
non-emotive utterances "what is spoken about" in Slavic languages is the given
information and "what is spoken about" in English is the subject. A question
arises whether the topic of a non-emotive utterance is of the same nature as in
the so-called topicalized structures. Or, if we hold that what is spoken about
in Slavic languages is not the first element (or given information) then what is
spoken about would be the subject. In that case "what is spoken about"
would be the subject in both English and Slavic languages. On the other hand,
there is no doubt that we need the concept 'topic' in the description of topi-
calized structures. It seems that in topicalized structures the aim of proposing
is to emphasize or contrast an element (in some cases though, as Enkvist
(1976:5) points out, "a constituent has to be moved simply to get it out of
the way because that final position is reserved for other [ . . . ] more important
constituents"). Is then the first element in non-emotive utterances of the same

nature as in topicalized structures, or is this position simply determined by the
grammatical structure in English and the thematic structure in Slavic languages?
The complexity of this problem has increased with the introduction of the
concept of Communicative Dynamism (CD) (Firbas 1964). It seems that
Communicative Dynamism was introduced to provide matching elements for
the semantic and grammatical structure of the utterance. Those structures in
their simple form consist of three elements each. The original matching of the
elements can be shown as follows (Danes 1964):
Grammatical structure S V
Semantic structure Ag Ac G
Thematic structure T(opic) C(omment)
The question of matching at that stage involved the problem of whether and
when the verb belongs to the theme or to the rheme. To partly answer that
question Firbas introduced the notion of Transition (Tr) or transitional ele-
ment. Now the problem was not simply defining theme and rheme (equivalent
to given and new information on the basis of Mathesius' description a) above)
but also defining Transition. Thus there could be no longer a bipartite but
gradual distinction between the elements under discussion. According to
Firbas (1964:270) the degree of CD carried by a sentence element is defined
as "the extent to which the sentence element contributes to the development
of the communication, to which it 'pushes the communication forward' as it
were. It is obvious that elements conveying new, unknown information show
higher degree of CD than elements conveying known information". Theme is
defined as having the lowest and rheme as the highest degree of CD, Transition
being defined as having a higher degree of CD than theme, and lower than rheme.
So now, in addition to the question which elements belong to theme, tran-
sition and rheme, we also have to be able to determine the degree of CD of
each element. The distinctions have become even more subtle, the final, full
form of CD being ThPr (Theme Proper) - rest of Th TrPr (Transition
Proper) - rest of Tr-Rh (Rheme) to the exclusion of RhPr(Rheme Proper) -
RhPr (Firbas 1975:331). The distinction and description of these constituents
of the thematic structure is a complex procedure that has to consider all levels,
interactions between elements of different levels and within the same level,
context and situation.
To illustrate this procedure let me take an example from Firbas (1975).
According to him context, linearity and semantic structure are the most im-
portant means involved in determining degrees of CD. The context makes ele-
ments of an utterance either 'context dependent' or 'context independent';
concepts that are to be understood 'in the narrowest sense possible'. An ele-
ment is context dependent if the piece of information it conveys is derivable

(or recoverable) from the preceding verbal context and/or refers to some
element of the immediate situational context. That description does not seem
to be very precise. On the one hand the preceding context may stretch from
the immediately preceding utterance to a more distant one, and it is obvious
that the distance factor will influence the interpretation of the elements of
the utterance (cf. for example, results of Osgood's (1971) research, and also
Szwedek 1980). On the other hand a distinction between verbal and situa-
tional context must be made, since it is reflected in the language (cf. the
distinction between 'linguistic' and 'non-linguistic' anaphora in Stockwell et
al. (1973), and the discussion of textual and situational anaphora, and of
unique nouns in Szwedek 1976:75 ff.). As to the participation of the grammat-
ical and semantic structures in the thematic interpretation, Firbas says that
If the semantic agent-action-goal pattern expressed by means of

the grammatical subject-verb-object pattern is contextually inde-
pendent in its entirety or contextually dependent merely through
its agent-subject element (A) (The girl broke a vase), the following
interpretation applies. The verb will carry a lower degree of CD
than the object, but a higher degree of CD than the subject. This
is because a known or unknown agent appears to be communi-
catively less important than an unknown action and its unknown
effector result (1974:20).
and also because an object "expresses an essential amplification of the latter,

and consequently becomes communicatively more important" (1972:79).
Discussing examples like A lion killed a hunter Firbas (1974:35) says that
it is remarkable how "a reader or hearer will most naturally interpret it as
actor-action-goal, subject-verb-object, theme-transition-rheme sequences". He
goes on arguing that "linearity being a very primitive (though efficient) means
cannot but reflect the normal and natural order of phenomena as occurring in
the extralinguistic reality. Initiating an action, the actor necessarily exists
before it. Only after it has started, can the action reach or effect its goal or
produce some altogether new object." This argument cannot be taken too
seriously because, as Firbas must well know, in Slavic languages in the most
natural order the actor does not necessarily precede the action in the linear
representation. What is more, however, many goals also exist before the action
and even before the actor.
As to the relation between the verb and the object, the description seems
to be circular: an object has a higher degree of CD (contributes more to the
development of communication, i.e. is more important for communication)
because it is communicatively more important. It would be totally arbitrary
to claim that a hunter contributes more to the development of the communi-

cation than a lion. In fact a hunter (the object with the highest degree of CD
which means that it 'pushes the communication forward') may not be men-
tioned, i.e. may not 'push the communication forward', in the next few sen-
tences at all. I suspect that all three constituents of this utterance contribute
equally to the development of communication. The stress seems to be equally
distributed at least over the two nouns. To my knowledge no reason has been
given why a sentence cannot have two rhemes, like in the above example,
carrying equal degree of CD, i.e. both being able to push the communication
forward.
It is obvious that the thematic interpretation as presented by Firbas has
been influenced by and confused with the semantic interpretation: since
almost every action must have an agent we take the latter for granted, and
this to Firbas means it is communicatively less important. But just as well, in
X gave Y to Z, all , Y and must exist before the action can take place.
And yet this does not mean that , Y and have the same (lowest) degree of
CD. Firbas's explanation may also have been influenced by selectional restric-
tions. Every verb can take a restricted type of agent, and to a lesser degree a
restricted type of goal. The so-called syntactic features of the noun and fea-
tures of the verb have to match only if the noun is an agent. For example the
verb read can take a noun marked [+human] as an agent, while there is no
parallel restriction on the goal of the verb read (anything that can be 'read':
book, paper, word, thoughts, etc.). Thus, a certain crucial, essential part of
meaning is repeated in both the agent-noun and the verb which may create
an impression that the agent-noun has a low degree of CD.
In any case Firbas bases his hypothesis on three concepts:
a) contextual boundness (context bound vs non-bound)
b) order of elements (sentence initial vs final)
c) degree of CD (from lowest in theme to highest in rheme).
It seems that the features in (a) and (b) are relatively easy to identify, while
as we have seen above, the elements of (c) are not. However, the relation be-
tween the elements of (a) and (b) is not as simple as Firbas claims. If we take
Akmajian and Jackendoffs (1970) example:
(6) John hit Bill and then George hit HIM
(with the interpretation in which HIM = John) we have to recognize the fol-
lowing relations:
a) HIM is definitely context bound,
b) HIM is stressed, i.e. has the highest degree of CD (if that
means anything), i.e. is the rheme.
Thus, what we have here is a context bound rheme which contradicts Firbas's
(1972:82) description of the relation between the two features: "non-thematic
elements are always contextually independent".
Example (6) shows that one of the main problems that badly needs clarifi-
cation is what is new and what is given information in the sentence. Chafe
(1976:32) vaguely says that "givenness is a status decided on by the speaker"
and that "it is fundamentally a matter of the speaker's belief that the item is
in the addressee's consciousness, not that it is recoverable (Halliday)". Similar-
ly vague is Firbas's (1975) description of the "narrow scene" and context-
dependent elements. Elements are context-dependent, according to him, when
he piece of information they convey is derivable (or recoverable) from the
preceding verbal context and/or refers to some elements of the immediate
situational context. Also Dahl (1976:41) distinguishes between onstage con-
cepts and off-stage concepts which, he says, are similar to contextually bound
and contextually non-bound elements. On the other hand he makes a distinc-
tion between on-stage elements and definite elements when he says: ,.known
concepts whether they are on-stage or off-stage are referred to with the
help of definite noun phrases". If, however, we accept the distinction between
on-stage and known (= definite), then Dahl's description of on-stage concepts
as already present in the addressee's consciousness, and off-stage concepts as
those that must be activated, i.e. retrieved from some deeper place in his
mind (such as his long-term memory), has also a high degree of vagueness of
which Dahl accuses Chafe.
Th? problem of the context and its relation to theme/rheme seems to be
even more complex as the discussion of Akmajian and Jackendoff's (1970) ex-
ample will show. On the other hand it is hoped that the discussion will indicate
ways of solving the problem of criteria for given/new distinction. Akmajian
and Jackendoff s examples are pretty straightforward. In the sentence
(7) John hit Bill and then George hit him
if him is unstressed it refers to Bill, if it is stressed it may refer to John or
some outsider. In any case it does not, when stressed, refer to Bill. With the
standard definition of 'given', however, we would not be able to get the inter-
pretation where him refers to John, because him referring to John could not
be treated as new information and thus stressed. Therefore, in such a case we
have to look for 'new' information in other features of the relation between
John and him. It should be noticed that John is subject and agent, while him
is object and patient, which means that the relations between him and John
are different than between him and Bill, the latter pair exhibiting agreement
in grammatical function and semantic role. Similarly in a version of the above
example:
(8) John hit Bill and then he hit George

when he is unstressed, it definitely refers to John (the same function and role),
and if stressed, it may refer to Bill (different function and different role).
Again the stress signals that the pronoun does not refer to the same function
and the same role item in the preceding clause.
In Polish the two interpretations of (7) differ in the word order and the
form of the pronoun:
(9) Janek uderzyt Billa, a potem uderzyl go Jurek.
'John hit Bill, and then hit him George.'
where go = Bill.
(10) Janek uderzyl Billa, a potem Jurek uderzyl jego.
'John hit Bill, and then George hit him.'
where jego = John.
I will not discuss these examples in more detail because all conclusions
arrived at for the English example also hold for the Polish equivalents.
In some structures, whether he refers to John or Bill may have to be deter-
mined by the context. Compare, for example, (11) and (12):
(11) John went to see Bill, but he was not at home.
(12) John went to see Bill, but he had to return.
(see Dahl 1976:44ff.) for an interesting discussion of these examples, and
also of deletion of pronouns in Polish).
The analysis of examples (6) (8) shows that identity of reference is not
the only criterion for given/new distinction. It means that the concept of
'new' information has to be revised and its definition will require taking into
account the context in all aspects. This is quite clear in Chafe's (1976) dis-
cussion of givenness and contrastiveness. I am not as sure as Chafe (1976:118)
that the features 'new' and 'contrastive' are totally unrelated phenomena.
Chafe says: "Presumably the chief reason for believing that contrastive items
carry new information has been the fact that they are given high pitch". The
brief analysis presented above shows that stressed items introduce some new
elements of meaning and that is precisely why they have to be marked by
stress. Chafe is probably right that contrast usually means selection from a
limited number of possibilities (cf. Szwedek 1975 for a similar suggestions).
On the other hand if we accept his statement that "in effect, the addressee is
assumed already to have in his consciousness all the information conveyed by 7
(I killed Cock Robin) except for selecting the correct agent of death. This
selection is all the sentence conveys in the way of new information" (Chafe
1976:117), we would also have to assume that there is a difference between

'simple (?) new' and 'contrastive new' information. Chafe's sentence 7 (re-
peated here as (13))
(13) /killed Cock Robin
may be a sequence to either
(14) Who kUled Cock Robin?
or
(15) John killed Cock Robin. (No! / killed Cock Robin).
As an answer to (14), (13) is a simple reply conveying new information; as an
answer to (15), (13) expresses a contrast and at the same time conveys new
information. Selecting the correct agent of death is selecting new information.
Chafe is right in saying that contrast usually means selection from a limited
number of possibilities, which does not mean, however, that the selected item
cannot be treated as textually 'new'.
From the discussion it follows that context is a crucial factor in determin-
ing what is 'new' and what is 'given'. At the same time the discussion also
showed that those elements can be identified independently of other levels of
analysis. As part 1 of this paper demonstrated, it seems that context also in-
cludes presuppositions. It is possible that investigation of the different or
identical behaviour of 'new' information with respect to presuppositional or
sentential context may throw some light on the nature of presuppositions and
their role in the text, as well as on the nature of the distinction between new
and given information.
References
Akmajian, Adrian - Ray Jackendoff

1970 "Coreferentiality and stress", Linguistic inquiry 1 : 1 2 4 - 1 2 6 .
Chafe, Wallace
1970 Meaning and the structure of language (Chicago: The University of Chicago
Press).
1976 "Givenness, contrastiveness, definiteness, subject, topics and point of view",
Subject and topic, edited by C. Li (New York: Academic Press), 2 2 - 5 5 .
Dahl, sten
1976 "What is new information?", Reports on text linguistics, edited by
. E. Enkvist and V. Kohonen (Abo: Abo Akademi Foundation), 3 7 - 5 0 .
Dane?, FrantiSek
1964 "A three level approach to syntax", TLP 1:225 - 2 4 0 .
DaneS, FrantiSek (ed.)
1974 Papers on functional sentence perspective (Prague: Academia).
Davidson, Donald - Gilbert Harman, eds

1975 The logic of grammar (Encino, California: Dickenson),
van Dijk, Teun
1972 Some aspects of text grammar (The Hague: Mouton).
Enkvist, Nils Erik
1976 "Notes on valency, semantic scope and thematic perspective as parameters of
adverbial placement in English", Reports on text linguistics, edited by
. E. Enkvist and V. Kohonen (Abo: Abo Akademi Foundation), 5 1 - 7 4 .
Enkvist, Nils Erik - Viljo Kohonen, eds
1976 Reports on text linguistics: approaches to word order (Abo: Abo Akademi
Foundation).
Fillmore, Charles J. - Terrence Langendoen, eds
1971 Studies in linguistic semantics (New York: Holt, Rinehart).
Firbas, Jan
1964 "On defining the theme in functional sentence analysis", TLP 1 : 2 6 7 - 2 8 0 .
1972 "On the interplay of prosodic and nonprosodic means of functional sentence
perspective", The Prague School of linguistics and language teaching, edited
by V. Fried (London: OUP), 7 7 - 9 4 .
1974 "Some aspects of the Czechoslovak approach to problems of functional sen-
tence perspective", Papers on functional sentence perspective, edited by
F. Dane? (Prague: Akademia), 1 1 - 3 7 .
1975 "On the thematic and the non-thematic section of the sentence", Style and
text, edited by H. Ringbom (Stockholm: Sprkfrlaget Skriptor AB, 3 1 7 -
334.
Fried, Vilem, ed.
1972 The Prague School of linguistics and language teaching (London: OUP).
Grice, H. P.
1975 "Logic and conversation", The logic of grammar, edited by D. Davidson and
G. Harman (Encino, Ca.: Dickenson), 6475.
Krzeszowski, Tomasz P.
1974 Contrastive generative grammar: theoretical foundations (L6dz: Uniwersytet
Ldzki, reprinted in 1979, Tbingen: Gunther Narr Verlag).
Lakoff, George
1971 "The role of deduction in grammar", Studies in linguistic semantics, edited
by C. J. Fillmore and T. Langendoen (New York: Holt), 6 3 - 7 7 .
Li, Charles (ed.)
1976 Subject and topic (New York: Academic Press).
Mathesius, Vilem
1939 "O tak zvanem aktudlnim cleneni vetnem", SaS 5:171 - 1 7 4 .
Osgood, Charles
1971 "Where do sentences come from?", Semantics, edited by D. D. Steinberg and
L. A. Jakobovits (Cambridge: CUP), 4 9 7 - 5 2 9 .
Ringbom, Hikan, ed.
1975 Style and text (Stockholm: Sprkfrlaget Skriptor AB).
Steinberg, Danny D. - Leon A. Jakobovits, eds
1971 Semantics (Cambridge: CUP).
Stock well, Robert P. - Paul Schachter - Barbara Partee
1973 The major syntactic structures of English (New York: Holt, Rinehart and
Winston).
Szwedek, Aleksander
1976 Word order, sentence stress and reference in English and Polish (Edmonton,
Canada: Linguistic Research, Inc.).
1980 "Lexical cohesion in text analysis",PSiCL 1 1 : 9 5 - 1 0 0 .
TADEUSZ ZABROCKI
On the nature of movement rules

in English and Polish
I would like to argue that there is a rather fundamental difference between

Polish and English. The grammar of the latter language employs trace-leaving
transformations in the derivation of such structures as passives, specific ques-
tions and relative clauses. On the other hand, Polish seems to be a language
without transformations.
The discussion will be limited to the structure-preserving rules which move
noun phrases: NP Movement and w/2-Movement. Rules affecting other major
categories will be ignored but I assume that the conclusions of our analysis
will apply to these rules as well. It will be claimed that the derivational
history of English sentences in (1)(2) below is different from that of their
Polish counterparts. This difference is reflected in their surface structure,
shown in (3)(4), and, consequently, in the interpretive procedure which
applies to these structures and assigns thematic relations to their noun
phrases.
(1) John was beaten by Bill.

Janek zostal pobity przez Billa.
(2) Whom did John meet?
Kogo Janek spotkat?
(3) Johnj was [beaten t j by Bill]
Janek zostal [pobity przez Billa]
(4) [Whom! [did John meet ti ]]
[Kogo [Janek spotkal]]
The interpretation of the English sentences proceeds in two steps:

a) a semantic rule associated with the verb (beat in (3) and meet in (4)) as-
signs a thematic relation to the trace in the object position (I will call such
rules lexical rules of interpretation, LRI);
432 Tadeusz Zabrocki
b) the relation is then 'transferred' to the dislocated NP co-indexed with the

trace, as in the case of the passive sentence, or to the variable bound by the Q
operator, as in the case of (4). The assignment of a thematic relation in the
above described situation is thus somewhat indirect. Of crucial importance
is the recognition of an anaphoric-like relation which exists between the
trace and moved NP.
In Polish the situation seems to be different. No anaphoric relationship, no
phonetically empty anaphors are involved. Lexical rules of interpretation ap-
ply directly to lexical NPs and it is the surface syntactic function of those NPs
(read from their case ending) that provides necessary and sufficient informa-
tion for interpretation.
In presenting the justification for the approach outlined above I will deal
separately y/ith 'NP Movement' constructions and w/z-Movement. Though
both types of structures are assumed to be non-transformational in Polish
their syntactic derivation seems to be different in significant respects.
1 NP Movement
What I want to claim is that there is no NP Movement in Polish. Passives and
other constructions whose transformational derivation would involve a
change of the syntactic function of an NP, reflected in its case ending, are
base derived. The natural consequence of this is that all potential arguments
for NP Movement applicable in English should fail in Polish.
Admittedly, not many such arguments exist. The old ones, based on vari-
ous co-occurrence facts, are no longer valid in a grammar which accounts for
selectional restrictions with the help of semantic well-formedness conditions
and which employs such devices as lexical redundancy rules to account for
the relationship between sentence types.
I am not going to present any new arguments for NP Movement. Instead,
I will mention two interesting facts about English passives noticed in recent
years by other people. It seems that these facts pose problems for an analysis
which derives all passives by base rules.
The first of them concerns peculiar constraints on adjectival passives (non-
transformational, according to the Lexicalist Hypothesis). These constraints
are not shared by verbal passives.
Wasow (1977) observes that (5)-(6) are ill-formed:
(4) *John is unknown to be rich.
(5) *John looks given a book.
(6) *John was helped.

Movement rules in English and Polish 433
The italicized portions provide an adjectival context. In (4) the accusative

subject of the embedded clause was passivized, in (5) and (6), the indirect ob-
ject. The verbal passive counterparts of (4)(6) are grammatical.
(7) John is known to be rich.

(8) John was given a book.
(9) John was helped.
According to Wasow, the ill-formedness of (4)(6) is caused by an independ-

ently motivated localness constraint on lexical redundancy rules which relate
passive sentences to their active counterparts. They could not affect items
which are not related to the verb as 'closely' as the direct object, i.e. indirect
objects and accusative subjects of embedded sentences.
The second fact is the peculiar asymmetry of idioms in English.2 There are
idioms like kick the bucket which contain a transitive verb but cannot be pas-
sivized without the loss of idiomatic meaning and there are others which allow
passivization:
(10) take advantage of

advantage was taken of
What is interesting is that there are no idioms which take the form of a passive
sentence and whose active counterparts could be interpreted only literally.
A possible way to explain this would be to assume:
a) that there is a passive transformation and
b) that idiom interpretation rules apply at the deep structure level.
Under these assumptions, idioms will be listed in the lexicon as active
(base derived) structures, while the proper interpretation of the hypothetical
"exclusively passive idioms" would require a reference to their surface pas-
sive form.
No such explanation for the absence of the "exclusively passive idioms"
appears to be available within an approach which derives all passives by base
rules. The structural asymmetry between passive and active sentences, which
follows from transformational analysis, is destroyed here.
Let us examine the situation in Polish.
First of all, Polish has no structural counterparts of those English passives
which Wasow claims cannot be base derived. (11)(13), which correspond to
English ( 7 ) - ( 9 ) respectively, are all ill-formed:
(11) *Janek jest j ) tyc chorym.

(12) *Janek zostai dany ksiqzkg.

(13) *Janek zostai pomozony.
The ungrammatically of the above follows automatically when we assume

Wasow's theory of constraints on lexical rules and the absence of NP Move-
ment in Polish.3
As far as the exclusively passive idioms are concerned, a fairly large number
of them can be discovered in Polish. The following examples can be given:
(14) On nie by I w ciemi bity. (= Tie was smart')

'He was not beaten in the temple'
(15) Ona jest kuta na cztery nogi. (= 'she is clever')
'She is shod on four feet'
(16) On jest w gorqcej wodzie kqpany. (= 'he is quick tempered')
'He is bathed in hot water'
The active counterparts of these sentences can be interpreted only literally.
(17) Nie bito go w ciemi.

(18) ?(Pod)kuto jq na cztery nogi.
(19) Wykqpano go w gorqcej wodzie.
The existence of idioms like (14)(16) is, of course, to be expected if passives

are base derived.
Still another difference between the two languages which can find a prin-
cipled explanation within the approach suggested here is the absence of
pseudopassives in Polish. This property need not be stipulated but would
follow from the fact that the independently motivated PS rules for Polish do
not allow prepositional adjuncts for major lexical categories (such as those in
the English verb = particle combination).
(20) This bed was slept in.

*To tozko byto spane w.
(21) He was looked at.
*On byt patrzony na.
Interestingly, an additional motivation for a non-transformational analysis is

supplied by what appear to be exceptions to the prohibition against pseudo-
passives. These are several well-formed passive constructions in which the sub-
ject NP corresponds to the prepositional object in the related active sentence.
(22) a. Jego zastugi zostaly zapomniane przez wszystkich.

'His merits were forgotten by everyone.'
b. Wszyscy zapomnieli jego zaslugach,4
'Everyone forgot about his merits.'
(23) a. Ten kosciot jest rzadko uczszczany przez ludzi.
'This church is rarely frequented by people.'
b. Ludzie rzadko ucz$szczajq do tego koscioia.
'People rarely frequent to this church.'
A transformational analysis would involve an extremely minor rule of prepo-
sition deletion necessary to block (24):
(24) *Zasluga zostala zapomniana o.
merit was forgotten about.
The rule will delete prepositions stranded by NP movement but not by wh-
movement as (25) shows.
(25) *Czyjq zastuge wszysy zapomnielP. as a counterpart of 22b
*Ktory koscibi ucz^szczasz
which church you-frequent
The non-transformational solution is straightforward. Passive participles and
related active verbs are separate lexical items with distinct strict-subcategori-
zation conditions. Ucz^szczac, zapomniec demand a PP complement while
uczgszczany, zapomniany do not.
I have concentrated on passives here but I believe that the analysis of other
potential instances of NP Movement in Polish will lead to similar conclusions.
The rules proposed in the rather scanty literature on the subject which
might fall under the heading of NP Movement are all idiosyncractic, lexically
governed operations, many of which violate well-motivated constraints on
transformations (cf.Zabrocki (1981) for a detailed survey).
To give one example, several of the proposed rules would involve a 'de-
motion' of a subject into an object position. The subject position remains
lexically empty, presumably occupied by a trace. The trace, however, is not
'properly bound' as trace theory demands, i.e. is not c-commanded by the dis-
located NP. (26)(28) illustrate this problem.
(26) t j \jest oczywiste N P l ze J a n e kprzyjdzie ] ] extraposition
is obvious that John will-come
'It is obvious that John will come.'
(27) t j [chce mii si? spac] inversion
want to-me to-sleep
want to sleep.'
(28) t , [nie ma tu Janka ] - negative existential

not have here John
'John is not here.'
To sum up this section, the fragmentary evidence considered supports the
hypothesis that there is no NP movement transformation in Polish.
2 RTJ-Movement
All EST analyses of w/z-Movement in English assume that surface structures

of, for example, specific questions, contain a trace. Some linguists argue that
it has been left there by a transformation (Chomsky 1977), others that it is
base derived and co-indexed with the w/z-word by a rule of construal (Kster
1978, Horn 1978). I would like to argue that neither of these two approaches
can adequately handle Polish data. The discussion will be divided into three
parts. First I will try to show that three independently motivated principles
of universal grammar which make reference to traces fail in a rather interest-
ing way when applied to Polish. Sencondly, I will comment briefly on the
nature of W/J-Movement in Polish as a non-transformational movement rule.
Finally I will come back to the problem of the interpretation of w/i-Move-
ment structures.
Resnic
To rule out constructions like (29) below, Chomsky (1979) proposes the fol-
lowing principle called Resnic.
(29) *Who1 does John believe that t, will come?
Resnic:

1 > r fei cannot be free in S
nom l 1
where "free" means not co-indexed with another element
and NP nom [e] is a trace of the subject in a tensed clause
Unlike English, Polish does allow extraction of subjects of tensed that/ze

clauses. (30) is comparable in grammatically to (31).
(30) to wierzysz, ze przyjdzie?

who you-believe that will-come
(31) Kogo wierzysz, ze Janek zaprosi?

whom you-believe that John will-invite
'Whom do you believe that John will invite?'
There are ways in which this apparent violation of Resnic can be accomo-
dated within a trace-theoretic approach (PRO drop parameter). They involve
considerable theoretical subtlety and I will not discuss them here. What is
important is that the problem can be solved in a straightforwards way if we
assume that the moved w/z-word has not left a trace.
Nesting Hypothesis
Kster (1978) argues that Universal Grammar contains a condition which he
calls Nesting Hypothesis.
Nesting Hypothesis:
A marked empty node m e [not dominated immediately by VP,S
or S - T.Z] is unacceptable when it is nested within a domain of
type V* S (main projection), where V' S* are minimal
In other words (simplifying), an NP or PP internal trace can occur only at the

extreme periphery of a VP, if it is a part of an object, or at the extreme peri-
phery of S or S, if it is a part of a phrase dominated immediately by either of
these two categories.
Within a transformational theory of w/z-movement in Polish, Koster's prin-
ciple fails inexplicably to account for a perfectly grammatical character of
(32).
(32) Ktrqi Janek Vp[spotkai Np[ti dziewczynf ]]

which John met girl
'Which girl did John meet?'
Equally grammatical is (33) in whose derivation w/z-movement was preceded

by a rule which extracted dziewczyn from within the PP. If the rule leaves a
trace, then (33) violates the Nesting Hypothesis.
(33) ktorej t , g [Janek mysli dziewczynie ]]

about which John thinks girl
'About which girl does John think?'
NP/Gap Order Condition (NP/GOC)

Horn 1978 formulates and justifies the following condition on surface struc-
tures:
NP/GOC:
The following sequence will always be interpreted by relating the
leftmost NP to the rightmost trace and the rightmost NP to the
leftmost trace:
NP . . . NP . . . t . . . t
If wh-movement leaves a trace, the following sentence is a clear counterexam-
ple to the condition (Polish allows more than one w/z-word in Comp):
(34) KtOi kogo2 [ti zaprosit t 2 ]
who whom invited
'Who invited whom?'
Another apparent violation of the condition is shown in (35) where w/?-move-
ment interacts with clitic movement.
(35) Komu jej2 kazales ti pomoc t 2
whom her you-ordered to-help
'Whom did you order to help her? '
Consider, however, sentences (36) and (37) where both /-movement and
clitic movement extract elements from within NP's.
(36) Czyjeji memu2 kazales [t 2 bratu]panbc [t 2 zonie]
whose my you-ordered brother to help wife
'Whose wife did you order my brother to help? '
(37) ? *Czyjemui mej2 kanzales [tjbratu] pomoc [t 2 zom'e]?
whose my you-ordered brother to help wife
'Whose brother did you order to help my wife? '
(36) is clearly more acceptable than (37). 7 It seems that NP/GOC is obeyed in
these cases.
The situation changes radically when one of the two interacting movements
affects a whole NP instead of a part of it.
(38) [Czyjemu bratu]l m0c2radzilem t t pomoc t 2 zonie
whose brother my I-advised to-help wife
Similarly, the grammaticality of (34) could be contrasted with the ill-formed-
ness of (39).
(39) *Ktryi ktrq2 [t t chtopiec] zaprosii [t 2 dziewczyne]
which which boy invited girl
This seemingly puzzling configuration of data (in one language all traces obey
NP/GOC, in the other only the NP internal ones) can be explained rather
naturally on three assumptions:
a) Wh-Movement does not leave a trace in Polish.

b) When a noun phrase is broken apart, there still exists a grammatical de-
pendency between the separated elements. The dependency would be formal-
ized as a rule of semantic interpretation (or a processing strategy) which esta-
blishes a link between the head of the phrase and the relocated modifier.
c) NP/GOC is generalized to apply not only to trace-antecedent dependencies
but also to the one just mentioned.
The general conclusion that would be drawn from the above noted facts is
that w/z-movement and clitic movement do not leave traces in Polish.8 Con-
sequently these rules are not transformations. It would clearly be undesirable
to allow T-rules to leave traces in one language and not in the other.
Two questions may be asked at this point. First, what about other prop-
erties of wh -movement, which are supposed to follow from its status as a
syntactic transformation such, as for example, subjacency or structure-preser-
vation? Secondly, if w/i-movement is not a T-rule, what sort of rule is it?
As regards the first question, it is unclear in what sense multiple applica-
tion of w/i-movement as in (40) can be regarded as structure preserving.
(40) to komu co dal?

who whom what gave
Certainly, no independently motivated PS rules exist which could derive in

Comp a sequence of NPs, their number limited only by the number of NPs
which can occur within the related S.
As for subjacency, there are problems with it the discussion of which will
go beyond the scope of this paper. 9 It suffices to say that it would have to
be reduced to clause-boundedness in Polish, i.e., the binding categories would
have to be S and S. Consider the ungrammatical sentence (41). Suppose that
its surface structure is (42).
(41) *Kogo kto powiedzial, ze Janek spotkai?

whom who said that John met
(42) Kogoy kto2 s [ t 2 powiedzial $[ti ie s [Janek spotkai t t ]]]
If only S or S is binding, the ill-formedness of (41) cannot be accounted for.

(Its English counterpart is ruled out by the doubly filled Comp filter inapplic-
able in Polish as (40) shows).
That NP is not binding is shown by those examples where the modifier has
been separated from the head which were discussed above.
There are many exceptions to subjacency so defined. Consider for example
(43). It is perfectly grammatical.
(43) Kogo s[Janek chcial [s[spotkac t, ]]]

whom John wanted to-meet
'Whom did John want to meet?'
It seems, and I believe that a more careful analysis will support this claim,
that the explanation of the fact that wfc-movement obeys both the w/z-Island
and the Complex Noun Phrase Constraint requires reference to principles
other than subjacency.
At this point, I can only give a very tentative answer to the second crucial
question. Of the syntactic processes described by generative grammarians,
Polish w/z-movement seems to resemble most a 'stylistic' rearrangement rule.
Thus, it will be related to the rule or rules responsible for 'scrambling' of con-
stituents in the Polish sentences in (44).
(44) Janek dal ksiqik? Mary si.
John gave book to-Mary
Janek dal Marysi ksiqzkg.
Janek Marysi dal ksiqzkf.
Marysi Janek dal ksiqzkq.
Ksiqzk$ dal Janek Marysi.
etc.
There are several formal similarities between scrambling and w/z-movement.

They behave in the same way with respect to Resnic, Nesting Hypothesis, NP/
GOC, wh-Island and Complex NP Constraint as well as other constraints of
unclear nature which, for example, allow (45) but prohibit (46).
(45) nie b$d$ mowil dziewczynie.

?
'About your girl I will not speak.

'About which girl will I not speak?'
nie mowil twojej. ?
who I n 0 t q * a ^ o u t your
Of course, there are problems with this proposal. Contrary to the model of
grammar proposed in Chomsky - Lasnik 1977, 'stylistic' rules will have to
apply in Polish before semantic interpretation rules (and filters). Independent
evidence for ordering at least some interpretive rules after 'scrambling' is

shown in (47)(48). The scope of logical elements-negation and the quantifier
appears to be sensitive to 'stylistic' rearrangement.
(47) Wielu ludzi nie przyszio na przyjcie.

many people not come to party
'Many people did not come to the party.'
(48) Nie przyszio wielu ludzi na przyjcie.
not come many people to party
'Not many people came to the party.'
It seems that the overall theoretical cost of the analysis proposed here is
smaller than if it had been left to particular grammars to decide whether
movement transformations leave traces or not.
The important function of traces in English is that they allow all rules of
semantic interpretation, including LRIs, to apply at a single level of syntactic
derivation-surface structure. One may ask how such a unification of grammar
is possible in a language in which deep structure configurations of elements
are deformed by movements which do not leave traces and which have to
apply before at least some rules of interpretation? What may be a problem for
a positional language like English is not one for an inflectional language like
Polish. Case endings, which are not changed by stylistic movements, and not
traces, serve as 'depository'of the semantically relevant information about the
underlying syntactic function of an NP.
What follows from the above is that Polish should not allow the deletion
of -words which are carriers of that non-recoverable information (English
assures such a recoverability since the underlying syntactic function of a wh-
word in Comp is identical to the surface function of its trace). This prediction
appears to be fulfilled.
Unlike English, Polish relative pronouns cannot be deleted under any cir-
cumstances.
(49) A man (whom) John knows came.

Cztowiek *(ktorego) Janek zna przyszedl.
Chomsky (1977) argues that w/j-Movement with a subsequent obligatory dele-

tion of the w/i-word in Comp is applied in the derivation of constructions pre-
viously assumed to involve such rules as Tough Movement, Adjective Phrase
Deletion, Topicalization and Comparative Deletion. Interestingly, there are
good reasons to reject w/?-Movement analysis for the counterparts of many,
if not all, such structures in Polish. This is explainable if we assume that a) if
w/i-Movement had applied in these structures the -word would have to be
necessarily deleted (for some reason), b) the notion of the 'recoverability'of

functional information should be understood in the way suggested above, i.e.
the recoverability conditions are different for two languages.
Consider, for example, (50)-(51).
(50) John is easy s [to please]

Janekjest tatwy PP [do zadowolenia]
a man s [ t o do this work]
m^zczyzna PP [do wykongri tej pracy]
(51) John is too weak to beat Bill.
Janek jest za slaby by pobic Billa.
In (50) infinitival clause in English corresponds to Polish PP with the preposi-
tion do followed by a derived nominal. Such a PP, naturally, does not pro-
vide a domain for vWz-movement. In the case of Adj complements, a -
movement analysis cannot account for the fact that only the subject of the
embedded clause can be omitted in the surface structure.
(52) John is too strong to beat.

*Janek jest za mocny by pobic.
A detailed analysis of other 'covert' -movement structures would reveal,
I believe, similar problems.
3 Conclusions
In a number of cases, Polish counterparts of the transformationally derived
English constructions have been shown to possess properties which create
problems for a transformational analysis.
It is, of course, possible that in each particular case the problem can be
reduced to the way the parameters of Universal Grammar are fixed in the
two languages compared. Unfortunately, it is difficult to imagine a solution
of this sort which could account for the full range of facts considered and
which would not consist of a number of fragmentary, unrelated explanations.
If we accept the hypothesis as to non-transformational character of Polish,
such a unified account becomes possible.
What has been presented here is naturally only a tentative proposal. To
advance it beyond this initial stage would require, first of all, a theory of non-
transformational "stylistic" movements, within which we should be able to
account for the various properties of Polish w/2-fronting. Unfortunately, little
work has been done in this area, particularly important for the "free word
order" languages, within the EST framework.
Notes
1. The paper is a report on a work in progress. Some of the arguments merely signalled
here are developed in Zabrocki (1981) and in Zabrocki (in preparation).
2. The asymmetry of idioms argument is a reinterpretation of the one given in Chomsky
1980.
3. The weakness of the argument as presented here is that it fails to explain why there
are not (at least I could not find any) exclusively passive idioms among the adjectival,
base derived passives (cf. Wasow 1977).
4 .Zapomniec can take an accusative NP for an object but the meaning is different in
such a case:
zapomniec list forget a letter
zapomniec liicie - forget about a letter
The passive in 22a can have either of the two meanings.
5. The sentence is grammatical but with a reading different from that of 22b. It corre-
sponds rather to
zapomniatem twojq zaslugq
which means "I forgot what your merit was". The question has no ambiguity of the
passive in 22a, mentioned in note 4.
6. Cf. Fodor 1978 for a different formulation and an extensive discussion.
7. For numerous native speakers of Polish, the example in (36) sounds ill-formed as well.
As I see it, the reason for it is that the extraction from within NPs, although possible,
is clearly a "marked" process in the sense of Kster 1978. Consequently, the struc-
tures created by the rules of this sort are very unstable. Any increase in complexity
leads to processing difficulties and lowers their acceptability. It should not surprise
thus that (36), which involves two such movements of which one crosses a clause
boundary, sounds awkward. The point is that it is better than corresponding (37).
8. Of course, such a conclusion crucially depends on the assumption of the correctness
of trace theory. Notice that our analysis not only does not undermine trace theory
but indirectly supports it. The principles of the theory do not fail when applied to
Polish but are simply inapplicable in a predictable way.
9. Cf. Zabrocki (in preparation) for discussion.
References
Chomsky, N.
1977 "On wA-movement",Formal syntax, edited by P. Culicover et al. (New York:
Academic Press), 7 1 - 1 3 2 .
1979 "Lectures on core grammar and markedness", unpublished manuscript.
1980 Rules and Representations. Oxford: Basil Blackwell.
1979 "Filters and control",Linguistic inquiry 7 : 4 2 5 - 5 0 4 .
Culicover, P. - T. Wasow - A. Akmajian, eds
1977 Formal syntax (New York: Academic Press).
1978 "Parsing strategies and constraints on transformations", Linguistic inquiry
9:427-493.
Horn, G. .
1978 lexical interpretive approach to some problems in syntax (Bloomington:
Koster, J.
1978 "Conditions, empty nodes and markedness", Linguistic inquiry 9:551 - 5 9 3 .
Wasow, T.
1977 "Transformations and the lexicon",Formal syntax, edited by P. Culicover et
al. (New York: Academic Press), 3 2 7 - 3 6 0 .
Zabrocki, T.
1981 Lexical rules of semantic interpretation; Control and NP movement in Eng-
lish and Polish (Poznan: A. Mickiewicz University),
in preparation "On vWi-movement in Polish".
Index of names
Abelson, R. 47, 67 Buckingham, H. 3 8 7 , 4 0 3

Ahlqvist, A. 202, 203 Bujas, 1 . 1 0 9 , 1 1 5
Aissen, J. 147, 168 van Buren, P. 415
Akmajian, A. 2 4 9 , 4 2 5 , 4 2 6 , 4 2 8 , 4 4 3 Burt, M. 3 9 6 , 3 9 9 , 4 0 4 , 4 1 3 , 4 1 6
Alatis, J . E . 2 0 3 , 3 4 8 , 3 6 3
Allan, K. 1 5 0 , 1 6 8 Campbell, M. A. 280
Allen, C. 17, 279, 280 Canale, M. 3 8 0 , 4 0 3
Allen, S. 141 Cancino, H. 3 9 5 , 4 0 5
Allerton, D . J . 141 Carpenter, P. A. 47, 66
Althaus, H.-P. 330 Carroll, J. B. 2 0 2 , 2 0 3
Anderson, S. 249, 280 Carroll, J. M. 3 8 4 , 4 0 4
Andrassy, V. 1 0 8 , 1 1 5 Carstensen, B. 1 1 0 , 1 1 5
d'Anglejan, A. 3 9 5 , 4 0 3 Carterette, E . C . 4 0 4
Aprcsjan, J. D. 168 Casagrande, J . B . 2 8 9 , 2 9 9
Ard, J. 4 1 3 , 4 1 4 , 4 1 6 Catford, J. C. 3 2 0 , 3 2 8
Arndt, . 331 Chafe, W. 4 2 0 , 4 2 6 , 4 2 7 , 4 2 8
Chernov, G. V. 284, 299
Babby, L. 161, 168 Chesterman, A. 6 0 , 6 5 , 3 1 9 , 3 2 8
Bach, E. 2 2 0 , 2 4 9 Chitoran, D. 1 1 0 , 1 1 5 , 1 1 6
Bach, K. 3 5 1 , 3 6 3 Chomsky, C. 4 1 2 , 4 1 6
Bailey, R. W. 4 1 3 , 4 1 6 Chomsky, N. 1 7 , 4 6 , 174, 175, 1 7 6 , 1 7 7 ,
Baltin, . 264, 280 184, 205, 219, 220, 242, 243, 244,
Barik, H . A . 284, 299 245, 247, 249, 260, 263, 266, 279,
Barrera-Vidal, A. 320, 328 280, 334, 347, 388, 4 0 4 , 409, 4 1 2 ,
Bauer, I. 108, 115 415,416,436,440,441,443
de Beaugrande, R. 65 Christ, H. 104
Becka, J . V . 308, 311 Clark, . V. 47, 66, 384, 385, 404
Bellugi, U. 347 Clark, . H. 47, 66, 384, 385, 4 0 4
Belnap, N. 3 5 3 , 3 6 3 Clyne, M. 327, 328
Beneke. J. 69, 103 Cogen, C. 4 0 5
Berens, F . J . 105 Cole, P. 1 6 8 , 1 6 9
Berman, A. 220 Cole, R. A. 405
Bever. T. G. 384, 4 0 4 Comrie, B. 4 1 3 , 4 1 4 , 4 1 6
Bialystok, E. 3 9 4 , 3 9 5 , 4 0 3 Corder, S.P. 2 9 1 , 2 9 9 , 3 0 0 , 3 1 7 , 3 2 8
Bilimic, J. 1 0 8 , 1 1 5 Coste, D. 328
Blank, M. A. 386 Coulmas, F. 104
Bock, J. K. 4 7 , 6 7 Coulthard, . 70, 73, 103, 105
Bolinger, D. 3 0 3 , 3 1 1 Crystal, D. 70, 103
Borkin, A. 4 3 , 4 4 Culicover, P. 249, 259, 281, 4 1 3 , 414,
Borsely, R. 1 1 8 , 4 3 , 2 7 4 , 2 7 8 , 2 7 9 , 2 8 0 417, 4 4 3
Bowen. S. D. 3 1 2 , 3 4 7 , 3 4 8 Cummins, J. 394, 4 0 4
Breivik. L. E. 1 5 0 , 1 5 1 . 1 5 3 , 1 6 7 . 1 6 8
Bresnan, J. W. 1 , 2 , 4 , 5 , 6 . 8 , 9 . 12, 13, Dahl, . 5 3 , 4 2 6 , 4 2 8
14, 16, 17, 249, 262, 279. 280 Danes, F. 46, 66, 4 2 3 , 4 2 8 , 4 2 9
Brislin, R.W. 300 Danesi, . 329
Brown. R. 3 4 7 . 4 1 3 , 4 1 6 Davidson, D. 4 2 9
446 Index of names
Davis, S. . 3 8 3 , 4 0 4 Foley, W. . 1 2 0 , 1 4 1
Davison, A. 280 Ford, . 17
Davy, D. 70, 103 Foreman, D. 82, 104
Dezs<5, L. 1 1 0 , 1 1 5 Foss, D . J . 3 8 6 , 4 0 4 , 4 0 5
van Dijk, T. 6 6 , 4 2 1 , 4 2 9 Foulke, E. 300
Dimitrijevic, N. 327, 328 Franck, D. 119
Dingwall, W. 3 4 9 , 3 6 3 Frser, B. 356, 358, 359, 361, 362, 363
Di Pietro, R. 320, 327, 328, 349, 363 Frazier, L. 63, 66
Dirven, R. 3 2 3 , 3 2 5 , 3 2 8 Freedle, R . O . 4 7 , 6 6
Dittmar, N. 3 9 4 , 4 0 4 Freihoff, R. 3 9 7 , 4 0 5
Downing, P. 1 2 1 , 1 4 1 Fried, V. 4 2 9
Dressler, W. U. 1 7 4 , 1 8 4 , 1 8 6 Friedman, . P. 4 0 4
Dulay, H. 3 9 6 , 3 9 9 , 4 0 4 , 4 1 3 , 4 1 6 Fujimura, O. 1 8 5 , 3 7 6
Duncan, S. 47, 66, 103, 104
Duskova, L. 1 6 7 , 1 6 8 , 3 0 6 , 3 1 1 Gass, S. 4 1 3 , 4 1 4 , 4 1 6
Dyta, S. 1 9 - 4 4 Gazdar, G. 363
Gee, J . P . 1 6 7 , 1 6 8
Ebneter, R. 320, 328 Geis, M. L. 279, 280
Eckman, F . R . 3 3 3 , 3 4 6 , 3 4 7 , 4 1 5 , 4 1 6 Gerver, D. 284, 300
E d m o n d s o n , W. 69, 70, 71, 75, 103, 104, Gingras, R. C. 3 9 5 , 4 0 5 , 4 0 7
356, 363 Givon, T. 47, 66
v a n E k , J. 3 9 7 , 4 0 4 Ghring, H. 3 2 6 , 3 2 7 , 3 2 9
Eliasson, S. 172, 173, 174, 185, 375, 376, Goldman-Eisler, F. 284, 300
377,406 Goody, . . 4 7 , 6 6
Ellis, A. W. 3 8 8 , 3 9 1 , 4 0 4 Goyvaerts, D. L. 1 7 4 , 1 8 5
Emonds, J . E . 2 5 1 , 2 5 4 , 2 5 5 , 2 5 8 , 2 6 3 , Graczyk, I. 309, 311
265, 277, 279, 280 Green, G. 1 5 6 , 1 6 9
Engels, L.K. 3 1 7 , 3 2 5 , 3 2 9 Greenbaum, S. 66, 281
Enkvist, . . 4 5 - 6 7 , 291, 3 0 0 , 4 2 2 , 428, Greenberg, J. H. 5 1 , 6 6
429 Gregersen, . 168
E r d m a n n , P. 1 5 0 , 1 6 7 , 1 6 8 Grice, . P. 4 2 0 , 4 2 9
Esser, J. 1 1 6 , 3 2 0 , 3 2 9 Grimes, J. 66
Grimshaw, J. 1 , 2 , 4 , 5 , 6 , 8 , 9 , 1 2 , 1 3 ,
Faerch, C. 6 9 - 1 0 5 , 3 2 7 , 3 2 9 , 3 9 4 , 4 0 1 , 14, 16, 17, 18
404 Grzegorek, M. 1 4 3 - 6 9 , 2 5 0 , 3 0 2 , 3 1 2
Farkas, D. 17, 18, 168 Gussmann, E. 1 7 1 - 1 8 6 , 3 6 6 , 3 7 1 , 3 7 5 ,
Felix, S.W. 4 0 1 , 4 0 4 376
Ferguson, C. A. 3 9 4 , 4 0 7 Gutknecht, C. 330
Filipovic, R. 1 0 7 - 1 1 7 , 3 2 0 , 3 2 3 , 3 2 5 ,
329, 330 Haarmann, . 203
Fillmore, C. 103, 104, 119 41, 309, 311, Haberland, . 3 5 0 , 3 6 3 , 4 0 4
429 Hkkinen, . 405
Fillmore, L. W. 119, 394, 404 Hakuta, . 3 9 5 , 4 0 5
Finch, A. 3 9 2 , 4 0 6 Halle, . 174, 175, 176, 177, 184, 249,
Findler, N . V . 4 7 , 6 6 373,376
Fink, S. 320, 329 Halliday, . . . 4 6 , 6 6 , 119, 123, 141,
Finocchiaro, M. 4 0 4 144, 169, 302, 303, 304, 311, 326,
Firbas, J. 5 3 , 1 6 8 , 4 2 3 , 4 2 4 , 4 2 5 , 4 2 6 329,426
Fischer-J0rgensen, E. 1 7 3 , 1 7 4 , 1 8 5 Hammerberg, . 66
Fisiak, J. 110, 116, 173, 175, 1 8 5 , 2 5 0 , Hankamer, J. 18
314, 329, 349, 350, 363, 375, 376, Harman, G. 3 6 4 , 4 2 9
379,404 Harnish, R . M . 3 5 1 , 3 6 3
Fiske, D.W. 4 7 , 6 6 Hartmann, R. R. . 3 1 9 , 3 2 4 , 3 2 9
Flores d'Arcais, G. B. 383, 384, 404, 406 Harwood, D . A . 405
Index of names 447
Hasan, R. 46, 66 Karpf, A. 3 2 4 , 3 3 0

Hatch, E. 3 9 4 , 3 9 5 , 4 0 5 Kasper, G. 6 9 - 1 0 5 , 358, 360, 361, 362,
Hegediis, J. 1 1 0 , 1 1 6 363,401,403,404
Helbig, G. 3 1 9 , 3 2 2 , 3 2 3 , 3 2 9 Katz, J . J . 3 5 1 , 3 6 3
Hendrick, R. 2 5 1 , 2 5 6 , 2 5 7 , 2 5 8 , 2 5 9 , Keenan, E. 4 1 3 , 4 1 4 , 4 1 6
260, 261, 262, 263, 264, 265, 266, Keller, E. 69, 104
268,269,277,278,279,280 Keller-Cohen, D. 399, 402, 405
Heny, F. 17 Kellerman, E. 100, 104, 333, 335, 346,
Hickey, R. 1 8 7 - 2 0 3 348,415,416
Hirschbhler, P. 2, 9, 12, 17, 18 Kenstowicz, M. 1 7 4 , 1 8 5
Hirvonen, T. 4 0 6 , 4 0 7 , 4 0 8 Kettemann, . 324, 330
Hjelmslev, L. 3 0 7 , 3 1 1 Kimball, J. P. 154, 155, 167, 169
Hollien, H. 3 8 7 , 4 0 3 Kiparsky, P. 173, 174, 181, 185, 249,
Van Hoof, M. 300 280, 373, 376
Hooper, J. B. 1 7 4 , 1 8 5 Kirkwood, H.W. 1 5 0 , 1 6 9
Hopper, P. J. 66 Kisseberth, C. 1 7 4 , 1 8 5
Horn, G. 1 7 , 2 0 5 - 5 0 , 2 6 4 , 2 6 6 , 2 7 9 , Klima, . S. 2 5 2 , 2 5 3 , 2 5 4 , 2 8 0
280,436,437,444 Kohonen, V. 5 1 , 6 7 , 4 2 8 , 4 2 9
Hornstein, N. 4 1 0 , 4 1 1 , 4 1 6 Knig, E. 1 1 2 , 1 1 6
House, J. 7 0 , 7 1 , 7 5 , 9 9 , 103, 1 0 4 , 3 5 8 , Kopczynski, A. 2 8 3 - 3 0 0 , 3 0 3 , 3 1 1
360, 361, 362, 363 Kortmann, E. 1 3 2 , 1 4 1
Hudson, D. 17 Koster, J. 436, 437, 443, 444
Hbler, A. 116 Krashen, S. D. 3 9 3 , 3 9 4 , 3 9 5 , 3 9 9 , 4 0 0 ,
Hllen, W. 3 1 9 , 3 2 0 , 3 2 9 401,402,405,413,416
Humboldt, W. von 59 Kranjcevic, S. 1 0 8 , 1 1 7
Hnig, W. 328 Kruppa, U. 323, 330
Krzeszowski, T. P. 184, 185, 202, 203,
Ivir, V. 114, 1 1 6 , 3 0 1 , 3 0 3 , 3 1 1 285, 300, 3 0 1 - 3 1 2 , 319, 320, 321,
330, 349, 360, 3 6 3 , 4 1 9 , 4 2 9
Jackendoff, R. S. 17, 18, 251, 252, 253, Khlwein, W. 1 0 4 , 3 1 3 - 3 1
254, 256, 258, 260, 261, 263, 265, Kuno, S. 13, 1 8 , 5 3
266, 269, 273, 277, 278, 279, 280,
4 2 5 , 4 2 6 , 428 Lado, R. 301, 312, 323, 336, 347, 348
Jackson, H. 320, 322, 329 Lakoff, G. 120, 141, 3 9 9 , 4 0 5 , 4 2 1 , 4 2 9
Jackson, K. L. 315, 331, 348 Lambert, K. 363
Jacobsen, W. M. 1 7 , 1 6 8 Lambrecht, K. 119
Jacobson, S. 5 1 , 6 6 Lamendella, J. 394, 395, 405, 407
Jger, G. 1 0 5 , 3 2 1 , 3 2 9 Lamminmki, R. 386, 387, 405
Jakobovits, L. A. 429 Langendoen, T. 429
James, C. 304, 311, 314, 329, 333, 346, Larsen-Freeman, D. 363
347, 380, 405 Lashley, K. S. 3 8 8 , 4 0 5
Janicki, K. 3 2 7 , 3 2 9 Laskowski, R. 1 7 5 , 1 8 4 , 1 8 5
Jaworska, E. 1 7 , 2 5 1 - 8 1 Lasnik, H. 260, 279, 412, 416, 440, 4 4 3
Jeffres, L. A. 405 Lass, R. 1 7 5 , 1 8 5
Jenkins, L. 1 4 6 , 1 6 9 Lea. W. A. 3 8 4 , 4 0 5
John, A. P. 315, 316, 330 Leben, W. 1 7 4 , 1 8 6
Johnson, D.E. 1 9 , 4 4 Lee, W. R. 314, 322, 323, 327, 330, 343,
Just, M. A. 47, 66 347, 348
Leech, G. 66, 103, 104, 281, 328
Kade, O. 312 Lehiste, I. 1 1 3 , 1 1 7
Kaisse, E. 18 Lehmann, W. P. 51
Kllgren, G. 4 6 , 6 7 Lehtonen, J. 3 6 4 , 3 8 0 , 3 8 3 , 3 8 4 , 3 8 5 ,
Kando, T. 3 5 5 , 3 6 3 386, 388, 389, 390, 391, 392, 394,
Karlsson, F. 3 8 9 , 4 0 5 396, 398, 400, 403, 405, 406, 407
448 Index of names
Levelt, W. J. . 3 8 3 , 4 0 4 , 4 0 6 Ohmann, R. 47, 67

Levi, J. 18 van Oirsouw, R. 415
Li, C. 5 4 , 6 7 , 4 2 8 , 4 2 9 Olesky, W. 309, 312, 3 4 9 - 364
Lieb, H.-H. 3 2 5 , 3 3 0 van Oosten, J. 145, 147, 148, 155, 167,
Lightfoot, D. 4 1 0 , 4 1 1 , 4 1 6 169
Lindholm, J. 280 Osgood, C. E. 4 7 . 6 7 , 4 2 4 , 4 2 9
Lindsay, P. H. 3 8 4 , 4 0 6 stman, J.-O. 119
Lipinska, M. see Grzegorek, M.
Liston, J. L. 3 0 2 , 3 1 2 Palmberg, R. 4 0 6
Longacre, R. E. 67 Palmer, F. R. 3 3 9 , 3 4 8
Lyons, J. 7 4 , 1 0 4 , 3 1 1 , 3 8 1 , 3 8 2 , 3 9 2 , Pankhurst, J. 415
406 Partee, B.H. 4 2 9
Peranteau, P. M. 18
Perlmutter, D. M. 4 3 , 4 4
McCawley, J. 1 3 0 , 2 0 2 , 2 0 3
Pfeiffer, O . E . 186
Macek, D. 112, 117 Phares, G. 18
Mcintosh, A. 311 Piaget, J. 4 1 5
McNeill, D. 4 1 0 , 4 1 7 Piepho, E. 104
Madden, C. 4 1 3 , 4 1 6 Pike, K. L. 323
Maling, J. 2 7 9 , 2 8 0
Pinker, S. 4 1 0 , 4 1 1 , 4 1 3 , 4 1 5 , 4 1 7
Mares, F . V . 186 Pisarek, W. 2 7 9 , 2 8 0
Markkanen, R. 3 1 9 , 3 2 0 , 330
Platzack, C. 6 3 , 6 7
Marshall, J. C. 4 0 4 , 4 0 6
Politzer, R. L. 3 0 2 , 3 1 2
Marslen-Wilson, W. D. 384, 387, 388, 4 0 6 Postal, P.M. 1 9 , 3 0 , 4 4 , 2 7 9 , 2 8 0
Martin, J. W. 3 1 2 , 3 4 7 , 3 4 8 Pullum, G. K. 43, 44, 145, 169, 279, 280
Marton, W. 184, 186, 285, 300, 314, 319,
320, 323, 330 Quirk, R. 2 5 3 , 2 6 2 , 281
Mascar, J. 3 7 3 , 3 7 6
Mathesius, V. 54, 422, 423, 4 2 9 Raabc, H. 116
Matisoff, J. A. 1 2 9 , 1 4 1 Raasch, A. 104
d e M a t t e i s , M. 3 1 6 , 3 3 0 Radden, G. 328
Matthei, E. 4 1 1 , 4 1 7 Rando, E. 1 5 4 , 1 6 6 , 1 6 9
Mazalin, D. 108 Rsnen, A. 4 0 6 , 4 0 7 , 4 0 8
Meisel, J. 3 9 4 , 3 9 8 , 3 9 9 , 4 0 6
Rehbein, J. 69, 104
Mel'uk, I . A . 168
Reighard, J. 17
Mey, J. L. 3 5 0 , 3 6 3
Richards, J . C . 3 1 5 , 3 3 0 , 4 0 3
Miller, G. 249
Riley, P. 350, 363
Milsark, G. L. 146, 147, 166, 167, 168,
Ringbom, H. 1 6 8 , 3 9 9 , 4 0 1 , 4 0 6 , 4 2 9
169
Rintell, E. 359, 362, 363
Mitchell, T. F. 3 4 7 , 3 4 8
Rischel, J. 185
Moravcsik, E. 141
Roos, E. 324, 330
Morgan, J. L. 6, 18
Rosenberg, S. 47, 67
Morton, J. 3 8 8 , 4 0 4 , 4 0 6
Moulton, W. 3 4 9 , 3 6 3 Ross, J. R. 9, 13, 18, 43, 44, 55, 147,
Mukattash, L. 3 3 3 - 3 4 8 169, 265, 281
Munby, J. 3 9 7 , 4 0 6 Roulet, E. 3 0 0 , 3 2 8
Rubach, J. 1 8 3 , 1 8 4 , 1 8 6 , 3 6 5 - 7 7
Ruder, K. F. 3 9 2 , 4 0 6
Nakic, A. 1 0 8 , 1 1 7 Rlker, K. 3 0 3 , 3 1 2
Napoli, D. J. 1 5 4 , 1 6 6 , 1 6 9 Rusiecki, J. 311, 312, 349, 363
Nemser, W. 3 2 5 , 4 1 4 , 4 1 7
Neubert, A. 312 Sacks, H. 73
Nickel, G. 109, 116, 117, 1 8 5 , 2 0 3 , 300, Sadock, J. 168, 169, 353, 364
329, 330, 338, 345, 347, 348 Sajavaara, K. 349, 364, 3 7 9 - 4 0 8
Norman, D. A. 3 8 4 , 4 0 6 Sapir, E. 325
Index of names 449
Schachter, J. 333, 335, 343, 344, 348 T h o m p s o n , S. A. 54, 66, 4 0 5

Schachter, P. 429 T h o r s e n . N . 185
Schnk, R. 4 7 , 4 8 , 6 7 Todrys, K.W. 1 7 , 1 6 8
Schenkein, J. 47, 67 Trabold-Szkoda, K. 278
Schiefelbusch, R. L. 404 Tran-Thi-Chau 347, 348
Schouten.M. E.H. 3 3 3 , 3 4 8 Tyler, L. K. 387, 4 0 6
Schumann, J. 3 9 4 , 3 9 5 , 4 0 7
Schwartz, A. 2 7 1 , 2 8 1 Upshur, J . A . 3 2 3 , 3 3 1
Schwarze, C. 3 2 0 , 3 3 1 Urbariczyk, S. 2 7 9 , 2 8 1
Schwitalla, J. 103, 105 Ure, J. L. 2 8 9 , 3 0 0
Searle, J . R . 169
Seidenberg, . 4 0 6 vanValin, R. D. 1 2 0 , 1 4 1
Seliger, H.W. 3 9 4 , 4 0 7 Vater, H. 319, 320, 331
Selinker, L. 3 9 4 , 4 0 7 Vennemann, T. 51
Sgall, P. 53 Viereck, W. 324, 330
Shannon, C. 3 8 2 , 4 0 7 de Villiers, J. 4 1 3 , 4 1 7
Sharwood Smith, M. 4 0 1 , 4 0 7 , 4 0 9 - 17 de Villiers, P. 4 1 3 , 4 1 7
Sheldon, A. 4 1 3 , 4 1 4 , 4 1 7
Sigurd, B. 4 6 , 6 7
Wachtel, T. 350, 364
Sinclair, J. 70, 105
Walters, J. 356, 358, 359, 362, 363, 364
Singh, R. 17
Wasow, T. 249, 4 3 2 , 433, 443, 444
Skuttnabb-Kangas, T. 404
Waterson, N. 394, 4 0 7
Slama-Cazacu, T. 110, 117, 324, 331
Weaver, W. 3 8 2 , 4 0 7
Slobin, D. I. 380, 381, 382, 385, 387,
Weil, H. 54
399,407
Weinstock, J. 1 8 5 , 3 7 6
Smith, N. 1 7 , 3 8 1 , 4 0 7
Weisgerber, L. 325
Snow, C. 3 9 4 , 4 0 7
Spalatin, L. 111, 117, 301, 302, 303, 312 Weist, R. 278
Spillner, B. 320, 321, 328, 331 Weller, F. R. 331
Stalnaker, R. 364 Welsh, A. 3 8 8 , 4 0 6
Stein, M.J. 280 Werth, J. 141
Steinberg, D. D. 4 2 9 Westerlund, A. 57
Stemmer, B. 75, 104 Wexler, K. 2 5 9 , 2 8 1 , 4 1 3 , 4 1 4 , 4 1 7
Stephanides, . 110 Weydt, H. 104
Stockwell, R. P. 305, 312, 347, 348, 424, White, L. 4 1 0 , 4 1 1 , 4 1 2 , 4 1 3 , 4 1 7
429 Whitfield, F. J. 311
Strauss, J. 328 Whitman, R. L. 315, 331, 343, 347, 348
Strevens, P. D. 6 9 , 1 0 5 , 3 1 1 Whorf, B. L. 49, 54, 325
Studemund, M. 203 Widdowson, H. 7 1 , 1 0 5 , 2 9 0 , 3 0 0 , 3 4 0 ,
Svartvik, J. 66, 103, 104, 281, 300, 328 348, 393, 3 9 7 , 4 0 8
Swain, M. 3 8 0 , 4 0 3 Wierlacher, A. 329
Szwedek, A. 1 6 0 , 3 4 9 , 3 6 4 , 4 1 9 - 4 2 9 Wigger, A. 202, 203
Wilkins, D. A. 397, 4 0 8
Taba-Warner, S. 69, 104 Wilson, D. 3 8 1 , 4 0 7
Takala, S. 3 9 7 , 4 0 5 Wilss, W. 322, 329, 330
Tannen, D. 141 Wode, H. 394, 395, 398, 399, 4 0 8
Tanenhaus, . K. 4 0 4 Woolford, E. 1 , 1 8
Tarone, E. 3 9 2 , 4 0 7
Tavakolian, S. L. 417 Zabrocki, T. 250 319, 331, 4 3 1 - 4 3
Teleman, U. 4 0 4 Zimmer, K . E . 1 2 1 , 1 4 1
Thome, G. 3 2 9 , 3 3 0 Zonneveld, W. 1 7 4 , 1 8 6
Thompson, H. 3 9 9 , 4 0 5 Zwicky, A. 173, 186

(Trends in Linguistics. State-Of-The-Art Report - 22.) Jacek Fisiak-Contrastive Linguistics - Prospects and Problems-Mouton Publishers

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

(Trends in Linguistics. State-Of-The-Art Report - 22.) Jacek Fisiak-Contrastive Linguistics - Prospects and Problems-Mouton Publishers

Hochgeladen von

Copyright:

Verfügbare Formate

Contrastive Linguistics

Papers prepared for the

Library of Congress Cataloging in Publication Data

(Trends in linguistics. Studies and monographs; 22)

Poznaii, May 1981 Jacek Fisiak

at the fourth International Conference of Contrastive Projects held at Charzy-

Professor Jacek Fisiak Adam Mickiewicz University, Poznaii

Docent Karol Janicki Adam Mickiewicz University, Poznad

Free relatives in Polish and English

(1) I'll buy what you 're selling.

what you are selling PRO

with embedded questions. On the other, there is an analysis on which they

you are selling what

As I will illustrate below, Bresnan and Grimshaw's arguments are directed

(6) Co powiesz, b$dzie uzyte przeciwko tobie.

Finally, there are free relatives introduced by a wh-word and a following

'Whichever work John found, he had to give it up imme-

(19) Whatever books she has marked up with her notes.

(20)J What books she has f!sn t , 1 certain.

(21) Jakiekolwiek szeroko

(23) Ten, kto umyi r$ce, moze zaczqc jesc.

(25) Tam, gdzie Jan mieszka, jest bardzo cieplo.

(28) Jan widziai zlodzieja, co obrabowal bank.

(34) Maria kocha m$iczyzn$, ktorego ty podziwiasz.

Thus, if Polish free relatives involve vv/j-movement, it is only natural that

(41) Jan jest inteligentniejszy, niz myslelismy, te jest.

(48) a. Ten student, ktory zdat wszystkie egzaminy, wyjezdza

b. *Ten, ktory zdat wszystkie egzaminy, student wyjezdza

(49) Ktore Jan namalowat obrazy?

Similarly, finite complements are always introduced by a complementizer.

(58) a. M^zczyzna, ktorego widziaies, jest bratem Manx.

(60) a. Cokolwiek tu slyszales, musisz tym zapomniec.

Notice now that (64) is ungrammatical.

This is no problem for a Hirschbhler analysis since on such an analysis swoich

(66) kimkolwiek Jan rozmawia, Maria go lubi.

relatives are isomorphic with embedded questions. We saw earlier, however,

(70) ?Can whether you are right or not matter?

These sentences contain embedded questions. Similar sentences containing

(72) Can what you want be on the table?

(76) Czy widziales tam kogokolwiek?

(77) Ktokolwiek moze tarn wejsc.

(79) Jezeli kiedykolwiek bgdziesz w Poznaniu, zadzwon do mnie.

(80) *Z kimkolwiek Jan rozmawia, Maria lubi.

It is clear, then, that a superficially headless analysis is quite compatible with

(83) I'll live in whatever town you live in.

A note on inversion/conjunct ascension

In this paper, I will attempt to give a tentative account of the syntax of a

1.1 Inversion and conjunct ascension in arc pair grammar

construction involving an immigrant arc whose foreign predecessor is a Con

1.2.1 Evidence for inversion in Polish

(4) Po przyjsciu do domu przypomniaio mi si?, ze . . .

(6) Po przyjsciu do domu przypomnialem sobie, ze . . .

(8) *Po przyjsciu do domu mama data mi kolacjf.

1.2.2 Evidence for inversion in English

(13) Making a fool of herself in public amused Rita.

1.2.3 Evidence for conjunct ascension in Polish

(16) Marek J0I3 poszli do kina.

1.2.4 Evidence for conjunct ascension in English

(17) John and Mary went to the movies.

Nonetheless, there seems to exist a slightly stronger piece of evidence to

2.0 Other constructions involved

Below, I will attempt to provide evidence for the well-formedness of each

2.1 Evidence for 21 advancement