The Rise of Agreement A Formal Approach To The Syntax and Grammaticalization of Verbal Inflectio

The Rise of Agreement
<DOCINFO AUTHOR ""TITLE "The Rise of Agreement: A formal approach to the syntax and grammaticalization of verbal inection"SUBJECT "Linguistik Aktuell/Linguistics Today, Volume 81"KEYWORDS ""SIZE HEIGHT "240"WIDTH "160"VOFFSET "4">
Linguistik Aktuell/Linguistics Today
Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies
into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical
problems as these are currently discussed in syntax, semantics, morphology, phonology, and
systematic pragmatics with the aim to establish robust empirical generalizations within a
universalistic perspective.
Series Editors
Werner Abraham
University of Vienna
Elly van Gelderen
Arizona State University
Advisory Editorial Board
Cedric Boeckx
Harvard University
Guglielmo Cinque
University of Venice
Gnther Grewendorf
J.W. Goethe-University, Frankfurt
Liliane Haegeman
University of Lille, France
Hubert Haider
University of Salzburg
Christer Platzack
University of Lund
Ian Roberts
Cambridge University
Ken Sar
Rutgers University, New Brunswick NJ
Lisa deMena Travis
McGill University
Sten Vikner
University of Aarhus
C. Jan-Wouter Zwart
University of Groningen
Volume 81
The Rise of Agreement: A formal approach to the syntax and
grammaticalization of verbal inection
by Eric Fu
A formal approach to
the syntax and grammaticalization
of verbal inection
Eric Fu
Johann-Wolfgang-Goethe University
John Benjamins Publishing Company
Amsterdam / Philadelphia
The paper used in this publication meets the minimum requirements
8
TM
of American National Standard for Information Sciences Permanence
of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data
Fu, Eric.
The rise of agreement : a formal approach to the syntax and
grammaticalization of verbal inection / Eric Fu.
p. cm. (Linguistik Aktuell/Linguistics Today, issn 01660829 ; v. 81)
Includes bibliographical references and index.
1. Grammar, Comparative and general--Agreement. 2. Grammar,
Comparative and general--Verb. 3. Grammar, Comparative and general--
Grammaticalization. 4. Historical linguistics. 5. Grammar, Comparative
and general--Inection. I. Title. II. Series.
P299.A35 F87 2005
415 22--dc22 2005049336
isbn 90 272 2805 1 (Hb; alk. paper)
2005 John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microlm, or
any other means, without written permission from the publisher.
John Benjamins Publishing Co. P.O. Box 36224 1020 me Amsterdam The Netherlands
John Benjamins North America P.O. Box 27519 Philadelphia pa 19118-0519 usa
JB[v.20020404] Prn:18/07/2005; 14:55 F: LA81CO.tex / p.1 (46-100)
Table of contents
Acknowledgements ix
Notes for the reader and list of abbreviations xi
Chapter 1
Introduction 1
1.1 The grammaticalization of verbal agreement markers 2
1.2 Previous accounts 5
1.2.1 NP-detachment (Givn 1976) 6
1.2.1.1 The sufxing preference 7
1.2.1.2 The pioneering role of 1st and 2nd person 9
1.2.1.3 Incompatibility between verbal agreement
and overt subjects 10
1.2.2 Accessibility theory (Ariel 2000) 12
1.2.3 Reanalysis of focus shells (Simpson & Wu 2002) 13
1.2.4 Section summary 15
1.3 Outline of the Book 16
Chapter 2
Theoretical preliminaries 23
2.1 Introduction 23
2.2 Minimalist syntax 23
2.3 Distributed Morphology 28
2.4 Language change: A generative perspective 33
2.5 Grammaticalization 40
2.6 Summary 49
Chapter 3
The structural design of agreement 55
3.1 Introduction 55
3.2 The phrase-structural representation of agreement 56
3.2.1 Conceptual arguments 58
3.2.2 Evidence from word order facts 59
3.2.3 The morphological realization of agreement 61
Table of contents
3.3 The structural relation involved in feature matching 67
3.4 Syntactic approaches to word formation 75
3.5 Interim summary: The design of agreement 80
3.6 Toward a realizational theory of agreement 81
3.6.1 The structural representation of agreement morphemes 81
3.6.2 Feature matching under closest c-command 83
3.6.3 Agreement and word formation 89
3.7 Multiple agreement: Inected complementizers in Germanic 94
3.7.1 Previous accounts of complementizer agreement 98
3.7.1.1 In-to-C movement 98
3.7.1.2 AgrP within CP 101
3.7.1.3 Multiple Agree 104
3.7.2 A Late Insertion account of complementizer agreement 105
3.8 Conclusion 114
Chapter 4
The transition from pronoun to inectional marker 129
4.1 Introduction 129
4.2 Telling apart clitics and agreement markers 130
4.2.1 Syntactic criteria 130
4.2.2 Morphological criteria 135
4.3 Syntactic preconditions for the rise of agreement 139
4.4 Paths toward agreement I: In-oriented clitics 143
4.5 Paths toward agreement II: C-oriented clitics 146
4.6 Summary 151
Chapter 5
The reanalysis of C-oriented clitics 157
5.2 Bavarian 158
5.2.1 The diachronic development of Agr-on-C in Bavarian 162
5.2.1.1 2sg /-st/ 162
5.2.1.2 2pl /-ts/ 164
5.2.1.3 1pl /-ma/ in Lower Bavarian 165
5.2.2 Clitics, V2, and the rise of agreement 167
5.2.3 Developments in other German varieties 171
5.2.3.1 Cimbrian 172
5.2.3.2 Walser dialects 175
5.3 Rhaeto-Romance 179
5.3.1 A grammatical sketch of the Swiss Rhaeto-Romance dialects 179
Table of contents
5.3.1.1 Word order: V2 and SVO 180
5.3.1.2 The distribution of pronominal forms and pro-drop 183
5.3.2 Earlier grammaticalization processes affecting enclitic pronouns 185
5.3.2.1 1st person forms 185
5.3.2.2 2nd person forms 187
5.3.3 Clitic doubling and the rise of agreement 190
5.3.4 The reanalysis of emphatic doubling structures 193
5.3.4.1 Syntactic restrictions on Agr-on-C 199
5.4 Reanalysis of C-oriented clitics in non-V2 languages 203
5.4.1 Uto-Aztecan 203
5.4.2 Mongolian 209
5.5 Conclusion 214
Chapter 6
Morphological blocking and the rise of agreement 229
6.2 The Blocking Principle 231
6.2.1 The rise of new verbal agreement endings in Bavarian 232
6.2.1.1 The diachrony of 2sg /-st/ 235
6.2.2 Morphological blocking versus analogical leveling 237
6.2.3 On the status of morphological doublets 242
6.3 The pioneering role of 1st and 2nd person 247
6.4 French 249
6.5 Northern Italian dialects 256
6.5.1 Piattino 257
6.5.2 Vicentino 260
6.6 Rhaeto-Romance 269
6.7 Language loss and the grammaticalization of agreement markers 275
6.8 Grammaticalization and multiple agreement in Skou 279
6.8.1 Strategies of agreement marking in present-day Skou 279
6.8.2 The historical origin of multiple agreement marking in Skou 282
6.9 Conclusion 283
Chapter 7
Concluding summary 299
References 305
Index 325
JB[v.20020404] Prn:2/08/2005; 9:33 F: LA81AC.tex / p.1 (44-118)
Acknowledgements
This book is a revised version of my doctoral dissertation, submitted to the University
of Frankfurt in November 2004 and defended in February 2005. I would like to use
this opportunity to express my gratitude to a number of people who have contributed
in various ways to the completion of this work. First, I would like to thank my the-
sis supervisor Gnther Grewendorf for his condence and support over the years. He
originally attracted me to theoretical linguistics and without him, I probably would
never have discovered the intellectual challenge and pleasures of studying human lan-
guage. Likewise, I am grateful to my second supervisor Ian Roberts, whose work on
diachronic syntax and in particular the phenomenon of grammaticalization provided
the initial stimulus for this book. His ideas had a profound inuence on my work and I
beneted greatly fromhis suggestions and comments on earlier versions of this book. I
would also like to thank the other members of my thesis committee, Helen Leuninger,
Jrg Meibauer, and in particular Ede Zimmermann, who readily agreed to act as a
third supervisor when it became necessary.
Furthermore, I would like to express my gratitude to a number of people who pro-
vided comments, support and advice and reassured me that the rise of agreement is a
topic worth writing a book about: Artemis Alexiadou, Mira Ariel, Ermenegildo Bidese,
Theresa Biberauer, Adriana Castagna, Greville Corbett, Mark Donohue, Hans-Martin
Grtner, Germen de Haan, Eric Haeberli, Liliane Haegeman, Katharina Hartmann,
Ruth Kempson, Anthony Kroch, Tom McFadden, Cian and Melanie McLoughlin,
William Morris, Peter hl, Albert Ortmann, Cecilia Poletto, Henk van Riemsdijk,
Joachim Sabel, Wolfgang Schulze, Peter Sells, Halldr Sigursson, Andrew Simpson,
Markus Steinbach, Carola Trips, Helmut Wei, Gunther De Vogelaer, Hedde Zeijlstra,
and Jochen Zeller. Special thanks go to my colleagues, both linguist and non-linguist,
who accompanied the race to nish this work and who provided constant support and
encouragement during the whole of the last year: Patrick Brandt, Ortrud Bruchelt,
Ccile Meier, and Shin-Sook Kim. Furthermore, I am particularly grateful to Werner
Abraham and Elly van Gelderen, who not only provided detailed comments on this
work, but also gave me the opportunity to have it published in the series Linguistik Ak-
tuell. To be sure, this name list is not exhaustive, and I apologize to those who I should
have mentioned here, but failed to do so.
At this point, I would also like to give my warmest thanks to my parents and
my grandma who always supported and encouraged me without a second thought
in everything I have ever done.
JB[v.20020404] Prn:2/08/2005; 9:33 F: LA81AC.tex / p.2 (118-122)
Acknowledgements
Finally, I think I would never have completed this book it if it had not been for
the three As in my life, Anne, Adrian and Anton, whose love, support and friendship
provided constant comfort and enabled me to carry on when things didnt look that
bright. This work is dedicated to you.
JB[v.20020404] Prn:18/07/2005; 14:57 F: LA81NO.tex / p.1 (46-185)
Notes for the reader and list of abbreviations
In the glosses to the linguistic examples cited in this work, the following notational
conventions are employed to distinguish between different kinds of bound (inec-
tional) markers: discrete inectional formatives are separated by a hyphen, whereas a
clitic boundary is marked by =. Fused marking, that is, instances where a given form
realizes more than a single (inectional) feature/meaning, is indicated by a dot sepa-
rating the individual features/meanings. Further abbreviations used in the glosses to
the linguistic examples are listed below.
abs absolutive
acc accusative
agr agreement
agro object agreement
agrs subject agreement
aor aorist
ap antipassive
asp aspect
caus causative
cl noun class
clit clitic
comp complementizer
cond conditional
dat dative
emph emphatic particle
erg ergative
femfeminine
foc focus (marker)
fut future tense
imperf imperfect
indic indicative
in inection
masc masculine
modal modal verb
modprt modal particle
neg negation
neut neuter
nmlz nominalization
nom nominative
object object (marker)
part participle
pass passive
past past tense
perf perfect
pl plural
poss possessive
pron pronoun
prt particle
prev preverb
prog progressive
pstprt past participle
pstwit past witnessed
pres present tense
qprt question particle
qmarker question marker
rereexive pronoun
sg singular
subjnc subjunctive
t tense
wh wh-pronoun/wh-question marker
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.1 (45-120)
Chapter 1
Introduction
The goal of this book is to explore the diachronic development of subject-verb agree-
ment, focusing on the reanalysis of pronominal clitics as inectional formatives. It is
shown that pronominal clitics can evolve into agreement morphology in quite a variety
of different syntactic environments, contra common beliefs prevalent in the literature
on this topic (cf. e.g. Givn 1976; Siewierska 1999). However, it is a central claim of
this study that the apparent diversity is only supercial whereas a closer inspection re-
veals a well-dened set of underlying syntactic and morphological conditions which
are operative during language acquisition and shape the grammaticalization process
in question across languages. These abstract conditions are shown to derive from the
way verbal agreement is established in natural languages by a combination of syn-
tactic mechanisms which value the content of agreement heads and morphological
principles governing the spell-out of the resulting feature combinations. In this sce-
nario, the grammaticalization of inectional morphology is taken to reect a division
of labor between syntax and morphology in which the former imposes a set of nec-
essary conditions on the reanalysis of pronominal elements while the ultimate trigger
of this change is morphological in nature. More specically, it is demonstrated that
the reanalysis of subject clitics is triggered by morphological blocking effects which
operate during language acquisition and ensure that the most specic potential expo-
nent of a given inectional head is selected and stored in the Lexicon. This hypothesis
is motivated by the observation that cross-linguistically, the grammaticalization pro-
cess under discussion either establishes agreement in languages that previously lacked
agreement or serves to repair a defective paradigm (cf. Chapter 6 for details):
(1) New verbal agreement morphology arises historically only for those slots
of the agreement paradigm where the existing verbal inection is non-
distinctive.
The theoretical framework adopted in this book combines the recent version of mini-
malist syntax devised in Chomsky (2000, 2001a, 2001b) with a Late Insertion approach
to morphology (Distributed Morphology). Thus, it is assumed that syntactic heads
are provided with phonological exponents post-syntactically (cf. e.g. Halle & Marantz
1993, 1994; Noyer 1997).
This introductory chapter is organized as follows. Section 1.1 illustrates the basic
properties of the diachronic phenomenon under consideration. Section 1.2 provides
a critical discussion of previous analyses of the grammaticalization of verbal agree-
ment ranging from the seminal paper by Talmy Givn (1976) to a recent approach to
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.2 (120-207)
the topic by Simpson and Wu (2002). In Section 1.3, I give an outline of this book,
sketching out the main theoretical proposals, both synchronic and diachronic, that are
developed in the chapters to come.
Before introducing the subject matter of this study in some more detail, let me
establish some basic descriptive terminology that will be used throughout this work
(basically following Corbett 1998, 2003). The element which determines agreement is
referred to as the controller. The element which changes its formas a result of the agree-
ment relation is called the target. The properties in which controller and target exhibit
covariance (i.e., agreement) are called agreement features. The phonological realization
of agreement features is called the marker or exponent of the relevant agreement fea-
tures. Finally, the agreement relation is typically restricted to a certain local syntactic
environment, which is referred to as the domain of the agreement relation.
1
. The grammaticalization of verbal agreement markers
It is a long-standing observation in historical linguistics that verbal agreement mor-
phology develops from (originally independent) personal pronouns. This diachronic
relation was rst established for the person and number endings of the early Indo-
European languages by 19th century grammarians such as Bopp (1816), Mller
(1875), or Brugmann (1916) when early work in historical linguistics sought to es-
tablish the sounds and the lexical inventory of a (hypothetical) common ancestor of
the Indic, Germanic and Romance language families, now commonly referred to as
Proto-Indo-European (PIE). In the 20th century, the basic insight has been elaborated
in the work of Indo-Europeanists such as Meillet (1912), Kuryowicz (1964), Watkins
(1969), and Szemernyi (1989) (cf. W. Lehmann 1993 for an overview). The diachronic
relation between agreement morphemes and personal pronouns in Indo-European
can be detected from a supercial inspection of the shape of pronouns and agree-
ment morphemes.
2
Table 1 lists the forms that are usually reconstructed for the set of
personal pronouns (nominative and accusative) and the primary and secondary agree-
ment endings for the present active (and aorist), taken from Szemernyi (1989: 228,
248).
3
Table 1. Reconstructed pronouns and agreement sufxes of Proto-Indo-European.
Pronouns Verbal agreement
Nominative Accusative Primary Secondary
1sg *eg(h)om, *eg o *(e)me, *m e, *m em *-mi *-m
2sg *t u *tu *twe/*te, *tw e/*t e *tw em/*t em *-si *-s (<*-t)
3sg *-ti *-t
1pl *wei, *nsms *nes/*nos, *n es/*n os,*nsme *-mes *-mes
2pl *y us, *usms/uswes *wes/*wos; *w es/*w os *usme, *uswes *-tes *-tes
3pl *-nti *-nt
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.3 (207-277)
Chapter 1. Introduction
Table 2. Pronouns and absolutive/ergative agreement in Basque.
Pronouns Absolutive agreement Ergative agreement
absolutive ergative go have
1sg ni ni-k n-u d-itu-t I have them
2sg su su-k s-us d-itu-su you have them
1pl gu gu-k g-us d-itu-gu we have them
2pl se-k se-k s-us-e d-itu-sue you have them
The similarity between personal pronouns and agreement sufxes is particularly
clear in the 1st person forms (see Szemernyi 1989: 229 for the claim that 1sg *-m de-
veloped from the nominative form *eg(h)om consisting of a particle eg(h)- which was
prexed to the pronominal stem -om). At rst sight, no such similarity holds between
the respective 2nd person forms. However, it is possible to construct a pronominal
origin for the 2nd person agreement endings as well (cf. e.g. Szemernyi 1989: 228ff.,
359ff. for details).
4
Similar observations can be made in present-day languages such as
Basque where the verbal afxes signaling agreement with (absolutive and ergative) 1st
and 2nd person arguments bear a clear resemblance with the respective pronouns (cf.
e.g. Arregi 1999).
5
As can be seen from Table 2, the prexes of the absolutive paradigm are iden-
tical with the onset of the corresponding absolutive pronouns. Apart from the form
for 1sg, the sufxes marking agreement with ergative subjects are even more simi-
lar to the relevant pronouns, since they preserve even the vowel of the 2nd person
and 1pl pronouns. The study of non-European languages and language families led
to the discovery that such similarities can be observed cross-linguistically, leading to
the insight that in general, pronouns are the primary source of bound person/number
markers (cf. e.g. Humboldt 1822, 1836; Meinhof 1936; Greenberg 1966b, 1978; Givn
1971, 1976; Lehmann 1988, 2002; Hopper & Traugott 1993; Corbett 1995; Siewierska
1999, 2004). For example, Moscati et al. (1969) note that in the Semitic languages,
there are systematic similarities between agreement markers and personal pronouns.
6
Similar observations can be made in various Bantu languages such as Swahili (Givn
1976) or Chichewa (Bresnan & Mchombo 1987). Andersen (1988) demonstrates that
in Pari, a Western Nilotic language spoken in the Sudan, the pronominal origin of the
proclitic agreement markers is still transparent in the present-day language. The Cau-
casian languages represent a particularly interesting case since the rise of new (person)
agreement markers goes hand in hand with a change from an absolutive/ergative to
nominative/accusative alignment (cf. e.g. Harris 1994; Schulze 1998). Poppe (1954),
(1960), and Comrie (1980) observe that a number of Mongolian languages (including
Buryat, Kalmyk, Dagur and Moghol) have grammaticalized verbal agreement sufxes
from pronouns quite recently.
7
This seems also to be true of several Tibeto-Burman
languages (such as Mao Naga, Tangut, Mikir, and Sgaw Karen, cf. LaPolla 1994;
Siewierska 2004). Burenhult (2002) shows that the preverbal subject agreement mark-
ers found in the Mon-Khmer language Jahai clearly resemble the relevant pronominal
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.4 (277-337)
forms. The change from clitic pronouns to agreement markers can also be observed
in many Australian (Hale 1973; Dixon 1980, 2002; Nordlinger 1998) and Papuan lan-
guages (Donohue 2002). In the Austronesian family, relevant examples come from
Kisar (Blood 1992), a set of languages spoken on Sulawesi (Taratn, Uma, and Tukang
Besi, cf. Cysouw 2003b), and the Oceanic language Asumboa (mentioned in Siewier-
ska 2004). Steele (1977, 1995) claims that second position clitics have developed into
prexal agreement markers in a number of Uto-Aztecan languages. Related develop-
ments can be observed in many other native American languages (cf. Mithun 1991 for
an overview and a set of interesting generalizations).
It is usually assumed that the historical development of verbal agreement mark-
ers from previously free pronouns universally follows the grammaticalization path
sketched in (2) (cf. e.g. Lehmann 1988: 59ff., 2002; Corbett 1995; Hopper & Traugott
1993; Siewierska 2004: 262).
8
(2) independent pronoun weak pronoun clitic pronoun afxal (agglu-
tinative) agreement marker fused agreement marker
Due to phonological erosion, a formerly independent referential pronoun turns into
a phonologically weak clitic which needs to attach to a lexical host (typically the -
nite verb). At some point, this pronominal element undergoes even further reduction,
loses its referential potential and is reanalyzed as an obligatory part of the verbal inec-
tion which redundantly marks the featural content of an argument DP.
9
Subsequently,
the resulting agglutinative agreement marker may be subject to further phonological
erosion and undergo fusion with other inectional markers such as tense before it
eventually disappears completely, restoring the stage that marked the beginning of the
grammaticalization path in (2). Note that the cyclic character of this change is a typi-
cal trait of grammaticalization processes (the most prominent being Jespersens cycle).
The last step of this historical development, that is, the loss of agreement marking,
can be observed in many Germanic languages, most notably English and Mainland
Scandinavian.
The observation that the rise of agreement markers follows a universal historical
pathway raises the question of whether it is possible to attribute the continuing re-
duction of pronouns to deeper factors. The present work adheres to the idea that the
facts described by (2) are to be explained by the assumption that grammaticalization
universally involves the reanalysis of lexical material as the phonological realization
of functional categories/heads (Roberts & Roussou 2003). By assumption, the lat-
ter are inherently prosodically defective. Thus, phonological deciency is viewed as
a precondition for the reanalysis as a functional head. In addition, due to the inher-
ent prosodical weakness of functional categories, further phonological reduction is
promoted after the reanalysis has taken place (see Chapter 2, Section 4 for discussion).
In many cases, the investigation of the historical processes in question is com-
plicated by the fact that the agreement markers are of considerable antiquity. Thus,
the relevant changes often predate the earliest written records of a language. This is
particularly clear in the case of a reconstructed proto-language such as Proto-Indo-
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.5 (337-391)
European, but also holds for cases such as Basque where the historical relation between
pronouns and agreement markers is still transparent in the present-day language (in
contrast to e.g. the Germanic family where the inherited agreement formatives bear
no resemblance to the present-day pronouns). Therefore, instances where the relevant
processes took place quite recently or can actually be witnessed as an ongoing pro-
cess are of special interest. Accordingly, the data sample discussed in this book comes
primarily from languages where the relevant changes are either still underway or have
taken place during the recorded history of the language, focusing on non-standard
varieties of Romance and Germanic.
A well-known case in point is Non-Standard French, where subject clitics have de-
veloped into a new form of verbal agreement marking (cf. Kuen 1957; Wartburg 1970;
Ashby 1977; Harris 1978; Lambrecht 1981; Auger 1993, 1994; Gerlach 2002, among
others). The historical development of subject clitics into markers of verbal agree-
ment is also attested in various Northern Italian and Rhaeto-Romance dialects (cf.
e.g. Meyer-Lbke 1894; Rohlfs 1949; Spiess 1956; Kuen 1957; Renzi & Vanelli 1983;
Rizzi 1986; Linder 1987; Brandi & Cordin 1989; Haiman 1991; Haiman & Beninc
1992; Poletto 1995, 1997, 1999, 2000; Roberts & Roussou 2003; Tortora 2003). Sim-
ilar developments can be observed in a number of Germanic dialects, in particular
Bavarian, where subject enclitics which attach to the nite verb in inversion contexts
have evolved into new forms of (verbal) inection (for Bavarian cf. Pfalz 1918; Lessiak
1963; Altmann 1984; Bayer 1984; Kollmer 1987; Wiesinger 1989; Wei 1998, among
others).
10
Summing up, it appears that there is some consensus concerning the general path
that is taken by the diachronic development of verbal agreement morphology from
former pronouns. However, nothing has been said so far about syntactic aspects and
possible triggers of the change in question. These questions are dealt with in the next
section which considers a set of previous explanations for the grammaticalization of
agreement markers.
. Previous accounts
In this section, I briey discuss three scenarios which have been proposed in the liter-
ature to explain the reanalysis of pronominal elements as verbal agreement markers.
The major part of this section is devoted to the most inuential of these proposals, the
so-called NP-detachment hypothesis offered in Givn (1976). In addition, I discuss an
approach in terms of Accessibility theory (Ariel 2000) and a recent proposal by Simp-
son and Wu (2002) which attributes the grammaticalization of agreement markers to
the reanalysis of focus constructions as agreement congurations. Since the analysis
developed in this book is partially based on the model of grammaticalization envis-
aged in Roberts and Roussou (2003), their account of the development of agreement
markers will be dealt with in some more detail in Chapter 2 which introduces the key
assumptions I will make use of.
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.6 (391-461)
.. NP-detachment (Givn 1976)
Current thinking about syntactic aspects of the historical development of predicate-
argument agreement is perhaps most substantially inuenced by the work of Talmy
Givn, most notably Givn (1971), (1976). Accordingly, Givns hypotheses are dis-
cussed in some more detail here. Givn claims that cross-linguistically, agreement
markers arise historically from the reanalysis of resumptive pronouns which relate
anaphorically to a left or right dislocated topic (so-called NP-detachment):
(3) a. [Joe]
i
,
topic
he
i
scored a goal.
pronoun
Joe
subject
he-scored a goal.
agr
b. He
i
scored a goal,
pronoun
[Joe]
i
.
topic
he-scored a goal,
agr
Joe.
subject
This change goes hand in hand with another major reanalysis. Due to an over-use of
the stylistically marked sentences involving NP-detachment (Givn 1976: 154 calls this
process de-marking), the former dislocation structures are reinterpreted as the neu-
tral syntax, with the former topic becoming the new subject. In other words, Givn
assumes that the diachronic source of grammatical agreement is some form of topic-
agreement.
11
The afxation of the former (free) resumptive pronoun to the verb is
attributed to the fact that unstressed pronouns often cliticize to the verb, which fa-
cilitates a reanalysis as part of the verbal inection. A similar process is taken to be
the source of object agreement, with topicalized objects being reanalyzed as residing
in their base (or case) position and resumptive clitics reanalyzed as markers of object
agreement.
An often discussed example of this scenario comes from Non-Standard French,
where clitics have developed into obligatory elements which can be doubled by full
phrases for reasons of emphasis (cf. e.g. Wartburg 1970; Ashby 1977). The relevant
sentences preferably receive a basic, non-dislocated interpretation. Therefore, the sub-
ject clitics are probably better analyzed as clitic/prexal agreement markers on the verb
(see Chapter 6 for more arguments in favor of this analysis):
(4) a. (Moi)
me
je
clit.1sg
porte
carry
la
the
table.
table
I carry the table.
#
As for me, I carry the table.
b. Moi
me
*(je)
clit.1sg
porte
carry
la
the
table.
table
(Gerlach 2002: 224)
The relation between the above examples and their diachronic origin as topic left dislo-
cation structures is still obvious: on the surface, they are identical to the corresponding
topic left dislocation construction of standard French. We can therefore conclude that
topic left dislocation structures are apparently a viable syntactic environment for the
historical development of subject-verb agreement. This scenario predicts that across
languages, the realization of agreement is connected to the topicality of the agreement
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.7 (461-548)
controller (since only topics can evolve into agreement controllers). And indeed, it
appears that we do nd traces of this particular historical origin of agreement mark-
ers. For example, if it is assumed that of all the verbs arguments, the subject/agent is
the most likely topic in a given utterance (cf. Givn 1976), Givns analysis provides
an explanation for the observation that cross-linguistically, subject agreement is the
most frequent instance of syntactic agreement. Furthermore, the fact that in many lan-
guages (Givn mentions Mandarin, Kinyarwanda and Malagasy as examples), subjects
must be either denite or generic is taken to follow from the origin of these subjects
as former topics, which are generally subject to similar restrictions (either denite or
generic, but never referential-indenite; Givn stresses that even in languages that tol-
erate indenite subjects such as English, subjects tend to be denite and referential).
In addition, it appears that in many languages, the possibility of object agreement is
determined by the discourse status of the object: in general, object agreement is more
likely to be triggered by topicalized DPs while focused objects often fail to control
agreement (cf. Siewierska 2004).
However, at a closer look, it appears that Givns analysis is not compatible with
another set of cross-linguistic generalizations on the rise and realization of agreement.
This suggests that there must exist alternative scenarios where pronouns can evolve
into agreement markers, in contrast to Givns strong claim that the reanalysis of re-
sumptive pronouns in NP-detachment constructions constitutes the only diachronic
path to grammatical agreement (cf. e.g. Siewierska 1999: 240 for a related statement):
One overriding theme and claim of this paper is that verb agreement
paradigms always arise from anaphoric pronoun paradigms. (Givn 1976: 180)
... The sufxing preference
Another claim put forward in Givn (1971, 1976) is that the position of agreement
afxes relative to the verb stem reects the position of the pronominal elements prior
to the reanalysis as agreement markers. Given that subjects precede the verb in the vast
majority of the worlds languages (i.e., in both SVO and SOV word order types), this
should lead us to expect that subject agreement is much more frequently realized by
prexes than by sufxes. However, it has often been observed that cross-linguistically,
there is a general tendency for inectional markers to be realized as sufxes on the
verb, sometimes referred to as the sufxing preference (cf. Sapir 1921; Greenberg 1963;
Cutler, Hawkins, & Gilligan 1985; Hawkins & Gilligan 1988; Bybee et al. 1990; Siewier-
ska & Bakker 1996; Julien 2002).
12
Table 3 illustrates that the effects of the sufxing
preference can also be observed if only agreement markers are taken into account
(combining gures from Tables 2, 4, 7 and 8 from Bybee et al. 1990).
13
From Table 3, we can see that in all word order types, sufxation is the most fre-
quent morphological means for marking agreement on the verb. At a closer look, it
becomes clear that the cross-linguistic sufxing preference is due to a set of inde-
pendent factors. First, in SOV languages, the vast majority of agreement markers is
realized as bound postverbal forms: prexal agreement markers are found in only six
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.8 (548-591)
Table 3. Position of agreement markers in 71 languages, by word order.
Word order
Position of Agr V-initial V-medial V-nal Total
Agr V 0 27 (10%) 10 (4%) 37 (6%)
Agr+V 11 (19%) 103 (37%) 80 (31%) 194 (32%)
V Agr 12 (20%) 10 (4%) 0 22 (4%)
V+Agr 36 (61%) 137 (49%) 171 (65%) 344 (58%)
of the 32 SOV languages examined by Bybee and her colleagues. Second, instead of ex-
hibiting the mirror image of SOV languages, SVO languages employ roughly as many
postverbal agreement markers as preverbal agreement markers, whereas verb-initial
languages seem to pattern with SOV languages. In particular, there are many SVO lan-
guages that show only sufxal instead of prexal verbal agreement markers (e.g. the
Romance family).
14
These ndings contradict the prediction that all languages where
subjects precede the verb should develop prexal agreement morphology. Moreover,
note that the very existence of agreement sufxes is problematic, given the fact that
unstressed clitic pronouns, which are generally taken to be the source of agreement af-
xes, occur very rarely in post-verbal position (apart from verb-initial languages and
inversion contexts). In contrast, they often tend to occupy a preverbal position rather
high up in the clause, to the left of In (as in Romance) or in the C-domain (e.g., the
Wackernagel clitics found in many Slavic languages).
The problem raised by SOV languages is also discussed by Givn (1976: 180ff.) in
connection with the sufxal agreement markers of (Proto-) Indo-European, which is
usually reconstructed as an SOV language. Givn suggests that the agreement sufxes
arose as a by-product of two major reanalyses that took place in early stages of Indo-
European and started from a periphrastic (perfective) participial-stative construction
which consisted of a non-agreeing participle followed by an inected auxiliary that car-
ried a prexal agreement marker (originally arising from the reanalysis of a pronoun):
(5) subject
i
... [[ participle ] agr
i
+aux]
In the course of time, the auxiliary cliticized to the participle. As a result, the prexal
agreement marker was sandwiched between the verb and its host. Eventually, the
inected auxiliary was reanalyzed as sufxal agreement, leading from a periphrastic
to a synthetic verb form. In an additional reanalysis, the passive construction turned
into an active past tense form. Givn mentions other examples where this complex
sequence of change events has apparently taken place (the sufxal perfect conjugation
of Semitic and certain past tense forms in Aramaic). Furthermore, as is well-known, a
somewhat similar development can be observed in Romance, where an inected form
of habere cliticized onto the non-nite main verb, giving rise to the synthetic future
(Meillet 1912; Benveniste 1968; Roberts 1993b; Roberts & Roussou 2003). Thus, we
cannot a priori exclude the possibility that this development took place in early Indo-
European (still, there are open questions, e.g. concerning the origin of the agreement
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.9 (591-644)
prexes on the auxiliary). However, since the person endings are already present in the
earliest records and are reconstructed for Proto-Indo-European as well, a closer under-
standing of the changes that actually gave rise to them presumably cannot be achieved
and any account must remain speculative.
15
Crucially, however, it is rather obvious
that the development envisaged by Givn cannot account for all instances of sufxal
agreement in SOV languages. Moreover, it cannot account for sufxal agreement in
SVO languages (at least without further assumptions), since the auxiliary precedes the
main verb in periphrastic constructions. Therefore, we must explore alternative paths
that may lead to sufxal agreement in order to account for its predominance across
languages (cf. Chapter 4 and Chapter 5 below for some discussion).
16
... The pioneering role of 1st and 2nd person
If it is assumed that verbal agreement marking develops from the reanalysis of resump-
tive pronouns in NP-detachment structures, we predict that new agreement markers
should develop initially for 3rd person, given that the vast majority of left- or right-
dislocated elements are 3rd person DPs. Accordingly, the development of 1st and
2nd person agreement should result from a later development, presumably driven by
analogy.
17
However, this prediction is not borne out by the facts.
Across the worlds languages, we can observe that the manifestation of verbal
agreement is sensitive to person distinctions. More specically, there are many lan-
guages where verbs agree with 1st and 2nd person subjects, but fail to do so with
3rd person subjects (cf. e.g. Bybee 1985; Mithun 1991; Cysouw 2003a; Siewierska
2004). Relevant examples come from languages such as Turkish (no verbal agreement
for 3sg, Kornlt 1990), the Nilotic language Turkana (Dimmendaal 1983), the Mon-
golian language Buryat (Poppe 1960; Comrie 1980), Pashto (no number agreement
with 3rd person subjects, MacKenzie 1990), the Australian language Wambaya (no 3sg
form, Nordlinger 1998), a number of Tibeto-Burman languages (LaPolla 1992), and
many native languages of North America (e.g. the Yuman and Siouan families, Mithun
1991).
18
The observation that verbal agreement marking for 1st and 2nd person sub-
jects is much more common than for 3rd person subjects is also corroborated by
quantitative studies. Bybee (1985) notes that 54%of the languages in her sample which
manifest agreement on the verb do not mark agreement with 3rd person arguments.
Moreover, in those languages which exhibit a full paradigmof person agreement mark-
ers, it can often be shown that the grammaticalization of 1st and 2nd person agreement
markers predated the grammaticalization of 3rd person forms, in contrast to what is
predicted by Givns analysis. Good examples of this chronological difference are pro-
vided by many native American languages (cf. Mithun 1991 for an overview).
19
We
can therefore conclude that the generalization in (6) describes a characteristic of the
grammaticalization of bound person agreement markers across languages.
20
(6) Sequence of the development of person agreement marking
1st and 2nd person markers become bound to the verb before 3rd person
markers.
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.10 (644-697)
Similar person restrictions can be observed in languages where the grammaticalization
of new verbal agreement markers took place quite recently and which are discussed in
more detail in this book such as Bavarian, Non-Standard French, and various Swiss
Rhaeto-Romance dialects. Therefore, we can conclude that we must look for alterna-
tive syntactic scenarios apart from NP-detachment which may license the diachronic
development of verbal agreement markers, at least for those languages that have failed
to develop 3rd person agreement markers historically.
... Incompatibility between verbal agreement and overt subjects
Some languages exhibit curious restrictions on the co-occurrence of overt subjects
and full-edged verbal agreement. A well-known instance are the Celtic languages. In
Welsh, for example, the verb agrees in person and number with pronominal subjects,
but fails to do so with full subject DPs (cf. e.g. Roberts 1999b; Siewierska 2004).
(7) Welsh
a. Gwel-sant
see-3pl.past
(hwy)
they
y
the
ferch.
girl
They saw the girl.
b. *Gwel-sant
see-3pl.past
y
the
plant
children
y
the
ferch.
girl
The children saw the girl.
c. Gwel-odd
see-3sg.past
y
the
bachgen/bechgyn
boy/boys
y
the
ferch.
girl
The boy/boys saw the girl. (Siewierska 2004: 152)
The examples in (7) show that the manifestation of verbal agreement depends on the
choice of subject. While the verb agrees in person and number with a pronominal
subject, as in (7a), it must carry default 3sg agreement if the subject is realized by
a full DP, as shown in (7b, c). Similar facts can be observed in Modern Irish where
even the co-occurrence of pronominal subjects and full-edged verbal agreement is
prohibited:
21
(8) Irish
a. *Chuirf-inn
put-cond.1sg
m
I
isteach
in
ar
on
an
the
phost
job
sin.
that
b. Chuirf-inn
put-cond.1sg
isteach
in
ar
on
an
the
phost
job
sin.
that
I would apply for that job. (McCloskey & Hale 1984: 491)
According to McCloskey and Hale (1984), in sentences with rich agreement, the sub-
ject position is occupied by a referential pro which is accessible for other grammatical
processes such as coordination (note that in Irish, the verb agrees only with the rst of
two conjoined DPs):
22
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.11 (697-771)
(9) D
if
mbeinn-se
be-cond.1sg
pro agus
and
tusa
you
ann.
there
If you and I were there.
A similar phenomenon can be observed in the Austronesian language Chamorro
(Chung 1998: 30f.). In Chamorro, the presence of person agreement on the verb re-
quires that a pronominal subject must be realized as pro. In (10), the verb carries
merely number agreement and an overt pronoun may be optionally present. In con-
trast, the presence of an afx which signals both person and number agreement
requires that the (pronominal) subject be realized as pro, as shown in (11):
(10) Ginin
imperf
man-ma-akki
pl-ap-steal.prog
(siha)
they
kamuti
sweet.potato
kda
every
dia.
day
They used to steal sweet potatoes every day.
(11) Ha-fahan
3sg-buy
pro/*gui
he
i
the
lepblu.
book
He bought the book.
As noted by Ariel (2000), the distribution of agreement and overt subjects exhibited
by languages such as Irish, Welsh or Chamorro is not expected on the assumption that
verbal agreement develops from NP-detachment, since this reanalysis necessarily in-
volves the presence of an overt subject DP which developed from a former dislocated
topic that regularly co-occurred with a resumptive pronoun. For languages where the
presence of verbal agreement is incompatible with the presence of overt subjects, we
are therefore led to conclude that the relevant agreement morphology does not origi-
nate from resumptive pronouns that relate to a dislocated topic, but must have evolved
in another context.
Summing up, the previous discussion has shown that the strong claim advocated
by Givn (NP-detachment is the only context where verbal agreement can evolve)
is contradicted by a set of cross-linguistic observations on the realization of subject-
verb agreement. First, it has been argued that Givns analysis is not compatible with
the so-called sufxing preference, the fact that in many languages, person/number
agreement is preferably realized by verbal sufxes, independent of other properties
such as word order. A second problem has been shown to arise from the observation
that in many instances, 1st and 2nd person agreement markers evolve prior to ver-
bal afxes signaling agreement with 3rd person subjects. Finally, I have argued that
Givns hypothesis is challenged by the existence of languages where full verbal agree-
ment in person and number is not compatible with overt subjects. These problems are
taken to indicate that there must exist other diachronic paths that lead to predicate-
argument agreement. In the following, I discuss two alternative proposals concerning
the historical development of agreement markers.
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.12 (771-834)
.. Accessibility theory (Ariel 2000)
The pioneering role of 1st and 2nd person has inspired an alternative approach to the
rise of agreement which is based on Accessibility theory (Ariel 1998, 2000). The ba-
sic idea behind Accessibility theory (Ariel 1990) is that the speaker chooses between
different types of referring expressions such as full DPs, pronouns, clitics, or agree-
ment markers (called accessibility markers) according to the salience (referred to as
mental accessibility) of the referred element in a given discourse context. In other
words, there is by assumption a close relationship between the salience/accessibility
and the morphological encoding of a discourse referent. The higher the accessibility
of a discourse referent, the higher the accessibility marker chosen. Lower accessibil-
ity markers such as proper names or full DPs are more informative and are typically
chosen when the referred element is introduced for the rst time into the discourse or
has a low degree of salience. Markers of higher accessibility include (weak) pronouns,
clitics, agreement markers and zero (i.e., pro). These are less informative and often
phonologically reduced or decient.
Based on these assumptions, Ariel (2000) claims that the development of agree-
ment markers from pronouns is not merely the result of phonological reduction,
but also driven by the speakers desire to use appropriate (reduced) markers to re-
fer to highly accessible/salient discourse referents. Note that in contrast to the NP-
detachment analysis, an approach in terms of Accessibility theory does not require the
presence of an additional (topicalized) DP. Rather, pronouns may develop directly
into verbal inections:
(12) a. i. You scored two goals. (free pronoun)
ii. You-scored two goals. (clitic pronoun)
iii. pro 2sg-scored two goals. (agreement marker)
b. You 2sg-scored two goals. (addition of an overt subject)
The transition from free pronoun to bound agreement marker illustrated in (12a.i
iii) is attributed to the pressure to develop an appropriate morphological encoding
of highly accessible discourse referents. According to Ariel, an independent argument
may then be added in contexts involving lower accessibility where the reduced marker
is not appropriate anymore, giving rise to new instances of grammatical agreement
with the reduced form losing its referential potential. This account predicts that agree-
ment markers should developinitially for 1st and 2nd person arguments since these are
always highly accessible as necessary discourse participants. In contrast, 3rd persons
are non-participants and therefore not as highly accessible as the speaker and hearer.
Accordingly, agreement for 3rd person forms should develop later, if at all. Similarly,
the requirement that reduced markers develop only for highly accessible referents hin-
ders the development of object agreement, since objects are inherently less accessible
referents than subjects (cf. Siewierska 2004: 265f.).
Thus, an approach in terms of Accessibility theory avoids the problems arising in
connection with the special role of 1st and 2nd person forms. However, it is associ-
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.13 (834-849)
ated with a different set of limitations since the development of 3rd person agreement
and object agreement seems to fall out of the scope of this proposal. Furthermore, it
does not provide a satisfying explanation for the presence of an overt controller in an
agreement conguration, that is, the development from stage (12a.iii) to (12b): if ver-
bal agreement is assumed to arise historically as a marker of high accessibility, then
the presence of an additional overt subject typically a marker of low accessibility
is not expected. Ariels suggestion that the presence of an overt controller may re-
sult from the need to introduce an intermediate stage of accessibility marking (which
combines high and low accessibility markers) undermines the whole concept of acces-
sibility marking (the further reduction of the high accessibility marker in question is
accompanied by the introduction of a lower accessibility marker). In addition, the as-
sumption that the reduced marker loses its referential potential in this development is
contradicted by pro-drop languages such as Italian where overt subjects can co-occur
with verbal agreement (note that Ariel assumes that agreement is referential in pro-
drop languages). Thus, we can conclude that Accessibility theory does not provide a
satisfying account of the historical development of agreement either. See Chapter 6 for
an analysis of the pioneering role of 1st and 2nd person which does not make use of
functionalist notions such as discourse salience/accessibility.
.. Reanalysis of focus shells (Simpson & Wu 2002)
In a recent paper, Simpson and Wu (2002) suggest that in general, syntactic con-
cord/agreement develops out of focus constructions which have lost their em-
phatic function. In most instances (e.g., the development of negative concord),
this corresponds to the following phrase-structural congurations, where (13) is
reanalyzed as (14).
(13)
XP1
X1
FocP
spec Foc
X1
0
Foc
0
ZP
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.14 (849-918)
(14)
XP1
X1
AgrP
spec Agr
X1
0
Agr
0
ZP
By assumption, FocPs are originally selected by some higher functional head. The
morphological material contained in SpecFocP (or Foc
0
) serves to repeat and thereby
emphasize the semantic content of the selecting functional head. Over time, the fo-
cus interpretation of the selected FocP decays, and the FocP is reanalyzed as an AgrP,
the content of which merely redundantly signals the feature content of the selecting
head. Later on, the content of the AgrP may come to be reanalyzed as the real, for-
merly selecting functional category, giving rise to Jespersens cycle effects. The rise
of subject-verb agreement is attributed to a related scenario. The initial stage of this
change involves a subject pronoun originally base-generated in SpecP which turns
into a clitic and is reanalyzed as an instantiation of
0
. For reasons of emphasis, the
clitic may select a FocP that contains a reinforcing full DP (in SpecFocP) with match-
ing -features. In turn, the additional full DP may raise up to the structural subject
position (SpecTP) while the clitic (which is in need of a lexical host) attracts the nite
verb to
0
. When this construction loses its stylistic force, the clitic is eventually rean-
alyzed as a mere agreement morpheme that is part of the nite verb, with the formerly
reinforcing subject becoming the real subject. The nal change involves a transition
from (15a) to (15b).
(15) a. [
TP
John
i
[
T
[
P
[
V+he
clit
[
FocP
t
i
... [
VP
... t
V
...]]]]]]
b. [
TP
John
i
[
T
[
P
t
i
[
V+he
Agr
[
VP
... t
V
...]]]]]
This analysis is related to the traditional idea that verbal agreement develops from a
doubling conguration where a full DP/pronoun is added to reinforce a weak or clitic
subject pronoun which cannot bear stress or convey emphasis. Eventually, the clitic is
reanalyzed as an agreement marker and the originally reinforcing DP becomes the true
subject (cf. e.g. Kuen 1957; Wartburg 1970; Ashby 1977 on the history of the Romance
languages, in particular French). Simpson and Wus formal rendition of this intuition
is appealing since it provides us with a unied account of the historical development of
agreement/concord phenomena. In addition, Simpson and Wu (p. 308) point out that
their proposal can account for the sufxing preference if it is assumed that the verb
will always left-adjoin to the light verb containing the former clitic (cf. Kayne 1994).
However, note that according to this analysis, new subject-verb agreement mor-
phology develops quite low in the clause structure. This raises two problems. First,
the relevant functional head,
0
, is standardly associated with object agreement, not
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.15 (918-966)
with subject agreement (cf. e.g. Chomsky 1995). Second, in most languages, we can
observe that clitic pronouns tend to occur in a rather high position, either adjacent to
C or to the left of T/In. Again, this observation conicts with the idea that clitics
located in
0
are reanalyzed as agreement markers. Furthermore, it has been observed
that in a number of languages, the realization of agreement is subject to restrictions
connected with the topic/focus distinction. In general, it appears that topical argu-
ments are likely to trigger agreement while focused ones fail to do so (cf. Siewierska
2004). If these restrictions are taken to reect the context where verbal agreement de-
veloped historically (which is quite plausible), this asymmetry is exactly the opposite of
what is predicted by Simpson and Wus proposal. Thus, at least there must exist alter-
native syntactic environments apart from the reanalysis of focus constructions where
subject-verb agreement may develop historically.
.. Section summary
In this section, I have discussed three previous approaches to the development of
agreement from pronominal clitics, arguing that each scenario accounts only for a
subset of the relevant phenomena. More specically, it has been demonstrated that an
analysis in terms of NP-detachment can account for the development of new agree-
ment markers in languages such as Non-Standard French, but is at odds with the
sufxing preference, the pioneering role of 1st and 2nd person forms, and the incom-
patibility between overt subjects and full-edged agreement exhibited by some Celtic
languages, for example. In contrast, an Accessibility-based account fares better with
respect to the special role of 1st and 2nd person, but requires special assumptions to
account for 3rd person agreement and the presence of an overt subject in agreement
congurations. Finally, the assumption that verbal agreement evolves out of former
focus constructions covers instances where the reinforcing full DP/pronoun in clitic
doubling congurations is reanalyzed as the subject of the clause, but falsely predicts
that cross-linguistically, agreement is more likely to be triggered by focal than by top-
ical constituents (apart from other, more technical problems). Thus, it appears that
NP-detachment, direct reanalysis of pronouns (i.e., Accessibility theory) or reanaly-
sis of focus shells are possible pathways towards verbal agreement, but crucially, none
of the relevant approaches can claim to constitute the only historical scenario where
agreement markers may develop from pronouns. In this book, I claim that the search
for such a unique scenario is in fact misguided. Rather, it is argued that the reanaly-
sis of pronominal elements is subject to a set of abstract syntactic and morphological
conditions which may be fullled in a variety of different syntactic environments (cf.
Chapters 4 and 5).
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.16 (966-1015)
. Outline of the Book
This book aims at developing a formal account of the grammaticalization of verbal
agreement morphology, focusing on the reanalysis of (C-oriented) subject clitics as
inectional markers.
23
Note that this work presupposes familiarity with the Princi-
ples and Parameters approach to syntax, that is, the theoretical apparatus laid out for
example in Chomsky (1981, 1982, 1986a, 1986b), and Chomsky and Lasnik (1993),
including X-bar theory and properties of the individual modules of Government and
Binding theory. Relevant theoretical concepts which are of more recent origin are
introduced and discussed in Chapter 2, including key properties of minimalist syn-
tax and Distributed Morphology. In the following, I provide an overview over the
organization of the book.
Chapter 2 introduces the basic theoretical assumptions adopted in this work. As
already hinted at above, a diachronic investigation into the historical development of
agreement morphology touches on several disciplines of linguistics, involving syntax,
morphology and the theory of language change in equal parts. Accordingly, I pro-
vide an outline of the current version of minimalist syntax (Chomsky 2000, 2001a,
2001b), an introduction into Distributed Morphology, and an overview of generative
approaches to language change, focusing on the phenomenon of grammaticalization
which is analyzed as a historical process that provides new phonological exponents for
a set of (universally present) functional heads (cf. Roberts & Roussou 2003).
As already mentioned, an analysis of the diachronic development of verbal agree-
ment necessarily builds on a theory of the syntactic and morphological mechanisms
which establish agreement in natural languages. In Chapter 3, I therefore develop a
synchronic approach to predicate-argument agreement which is based on the follow-
ing assumptions. First, agreement morphemes do not head their own projection in
the syntax. Rather, they are parasitic on other functional heads with which they may
combine (via adjunction) either before or after the syntactic derivation. Agr-heads
present in the syntactic derivation are valued by an appropriate controller under closest
c-command (i.e., Agree), while post-syntactically inserted Agr-heads (so-called disso-
ciated Agr-morphemes, cf. Embick 1997) are licensed under structural adjacency with
an Agr-morpheme that has been valued in the syntax. Based on the latter mechanism,
a novel analysis of complementizer agreement in Germanic is proposed, arguing that
this particular form of multiple agreement results from the post-syntactic insertion of
a dissociated Agr-morpheme onto C.
Chapter 4 explores the diachronic consequences of this particular approach to
verbal agreement. In particular, I propose a set of restrictions on the reanalysis of clitic
pronouns as exponents of agreement morphemes which derive from the theory of
verbal agreement proposed in the previous chapter. Furthermore, it is claimed that
the reanalysis of subject clitics may extend the pro-drop properties of a language by
creating newinstances of referential pro. Alternatively, subject clitics may be reanalyzed
as agreement markers in clitic doubling structures, with the relevant -role assigned to
the former double. In addition, I argue that the grammaticalization process in question
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.17 (1015-1071)
must give rise to structural simplication, in the sense that the structures resulting
from the reanalysis of clitics must involve less complex (syntactic) derivations. Based
on these assumptions, a set of historical scenarios is presented where the reanalysis of
clitic pronouns as agreement markers may be triggered.
In Chapter 5, I take a closer look at a subset of these scenarios, focusing on the
reanalysis of C-oriented clitics. It is shown that the latter change is shaped by inde-
pendent syntactic properties of the languages under consideration, arguing that the
V2 property facilitated the development of new agreement markers in German and
Rhaeto-Romance varieties. In addition, I discuss a set of Uto-Aztecan and Mongo-
lian languages which lack the V2 property but still exhibit a historical development
in which C-oriented clitics turned into verbal agreement markers. Furthermore, it is
demonstrated that the grammaticalization of new agreement formatives often pro-
ceeds via an initial stage where a clitic is reanalyzed as a dissociated agreement mor-
pheme which is inserted and valued in the morphological component, giving rise to
instances of multiple agreement (such as complementizer agreement in Germanic).
In Chapter 6, I discuss a set of morphological factors that shape the grammatical-
ization of agreement markers, focusing on the observation that across languages, the
development of new agreement formatives is triggered only in contexts where the ex-
isting verbal agreement morphology is non-distinctive. This observation is accounted
for by the assumption that the learner scans the input for the most specic exponent
of a given inectional morpheme. More technically speaking, it is argued that the ac-
quisition of inectional morphology is guided by blocking effects which ensure that
new verbal agreement formatives are more specic (i.e., realize a greater subset of the
-features contained in the relevant Agr-head) than existing morphology. Relevant ex-
amples come from a number of German varieties, Non-Standard French, and various
Northern Italian and Rhaeto-Romance dialects. In addition, I take a look at Ameri-
can Russian and the Papuan language Skou. Moreover, the account proposed in this
section is shown to provide a new explanation for the pioneering role of 1st and 2nd
person in the development of predicate-argument agreement if it is assumed that 3rd
person forms are inherently underspecied for person features.
Notes
. Note that the choice of the terms controller and target already implies that the structural rela-
tion between these two elements is asymmetrical in nature, in the sense that the target changes
its morphological form in order to match (inherent) feature specications of the controller (cf.
Lehmann 1993; Corbett 1998; Chomsky 2002, among many others). Furthermore, there is a
clear intuition that the semantic interpretation of agreement features is linked to properties of
the controller, rather than to properties of the target. In current syntactic theorizing, this asym-
metry is generally modeled by the notion that agreement features are interpretable (i.e., legible
by grammar-external modules such as the conceptual system) on nouns, but uninterpretable on
verbs, adjectives etc. (Chomsky 1995, 2000, 2001a, 2001b). As Chomsky (2002: 112) puts it: It
looks as if there is some real truth to the traditional idea that verbs agree with nouns and not
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.18 (1071-1152)
:8 The Rise of Agreement
conversely. The thing that is agreeing, presumably the verb, the adjective, the article, and so on,
they all seem to have uninterpretable features, features that are not independently interpreted
by the outside systems.
i. A more indirect indication of the diachronic origin of verbal agreement markers is the fact
that cross-linguistically, agreement morphemes mark a subset of the grammatical distinctions
marked by personal pronouns (cf. Barlow 1992; Lehmann 1993; Corbett 1998; Ariel 2000). The
set of -features marked by verbal agreement is usually assumed to be conned to [person],
[number], [gender] and [noun class] (the set of languages which exhibit agreement in noun class
includes Bantu languages such as Swahili or Chichewa, cf. Givn 1976; Bresnan & Mchombo
1987; Krifka 1995, and many Caucasian languages, cf. e.g. Schulze 1998; Harris & Smeets 2003).
. The primary endings appear in the present whereas the secondary endings are used in the
aorist and imperfect (e.g. in Sanskrit and the earliest Greek records). Historically, the primary
endings are derived via afxation of a deictic -i to the secondary endings. In other words, in
contrast to the somewhat misleading terminology, the secondary, unextended endings are in
fact older than the primary, extended endings (Szemernyi 1989; W. Lehmann 1993).
|. Note that the 3rd person forms did not develop from personal pronouns (as becomes clear
fromtable 1, no 3rd person pronouns are reconstructed for PIE). Traditionally, it is assumed that
the 3sg ending -t developed from the demonstrative *to, whereas the 3pl sufx *-nt is related to
the participial (i.e. originally derivational) afx *-nt- (Szemernyi 1989: 361).
,. Arregi (1999) claims that the pronouns and agreement markers of Basque are not only histor-
ically, but also synchronically related. For example, he decomposes the (absolutive) 2pl pronoun
sue into a set of formatives including s- 2nd person, -u- nominal stem and -e plural. Under
this analysis, the similarity between pronouns and agreement becomes even more intimate. Ar-
regi then suggests that pronouns and verbal agreement actually involve the same lexical items,
which are underspecied for the kind of stem (verbal or nominal) they attach to.
o. Recently, Ariel (1998, 2000) has observed that clitics are currently evolving into a new agree-
ment paradigm in the future tense of Colloquial Hebrew.
. In Dagur Mongolian, for example, the agreement sufxes are still fully identical to the
pronominal forms (cf. Martin 1961).
8. Note that (2) actually conates two distinctions. On the one hand, it represents the different
morphological stages which can be identied in this grammaticalization process (such as free
vs. bound form). On the other hand, it includes information about the changing syntactico-
semantic properties of the element in question, that is, the transition froma referential pronoun
(with a deictic and/or anaphoric potential) to an agreement marker which does not have any
referential potential on its own, but merely reects the -set of an argument.
. The fact that there are a few languages (mostly Creoles) which apparently exhibit non-bound
agreement markers suggests that the phonological reduction involved in this process does not
necessarily result in bound agreement formatives (cf. Bybee et al. 1990; Julien 2002).
:o. An instance of the change under discussion which has received some attention in historical
studies on Germanic is the development of the 2sg ending -st, which is usually attributed to the
cliticization of the 2sg pronoun -t(hu) to the inherited 2sg sufx -s in inversion contexts (for
Old High German cf. e.g. Brinkmann 1931; Braune 1950; Paul 1952; Lhr 1984; for Old English
e.g. Wright 1925).
::. See Bresnan and Mchombo (1987) for an application of Givns proposals in an LFG anal-
ysis of verbal agreement marking in Chichewa (Bantu). According to Bresnan and Mchombo,
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.19 (1152-1206)
Chichewa exhibits two types of agreement which developed historically via the incorporation
of resumptive pronouns. First, there is true grammatical agreement where an NP bears an ar-
gument relation to the verb, while its -features are redundantly expressed by verbal afxes.
Second, there is the possibility of so-called anaphoric agreement where the apparent Agr-
morpheme is an incorporated pronominal argument, while the coreferential DP is analyzed
as an adjunct that occupies an A-position (either topic or focus). By assumption, anaphoric
agreement preserves the historical origin of these constructions more faithfully than grammat-
ical agreement which is analyzed as a more recent development. Based on this distinction, it is
argued that in Chichewa, the subject agreement marker can function either as anaphoric or as
grammatical agreement, whereas the object marker can solely be used for anaphoric agreement,
which leads to a set of subject/object asymmetries in this language.
. Note that the term sufxing preference actually involves two separate generalizations: rst, of
all morphological means to realize verbal inection, sufxing is the most frequent one. Second,
postverbal inectional markers show a much greater tendency to appear as bound elements (on
the verb) than preverbal markers.
. Note that Table 3 distinguishes between free (e.g. Agr V) and bound (Agr+V) forms. The
gures refer to individual agreement markers in the languages under consideration (i.e., 1sg,
2sg etc. are counted separately).
. Furthermore, it becomes clear that there is a general tendency for agreement markers to be
afxed to the verb. Apart from the verb-initial languages in Bybee et al.s sample where all pre-
verbal markers are bound to the verb, this tendency is generally stronger for postverbal markers
(particularly in SOV languages where it is 100%).
. In fact most researchers agree today that even the exact origin of the number markers in
Indo-European (which came into existence via a later development) eludes explanation, al-
though it seems to be clear that they bear some relation to the oblique forms of the personal
pronouns (see Table 1 above) (cf. e.g. W. Lehmann 1993).
. Based on a survey involving 402 languages, Siewierska (2004: 165) claims that the position of
person agreement afxes in fact does not reect the sufxing preference as clearly as other inec-
tional markers: The existence of a preference for sufxes over prexes receives support fromthe
ordering of tense, aspect and modality afxes and from the existence of languages which are ex-
clusively sufxing as well as from the lack of languages which are exclusively prexing (see, e.g.,
Hawkins & Gilligan 1988). However, it nds only very weak support from the location of per-
son agreement afxes. Among the languages in the sample the markers of a[gent] agreement,
p[atient] agreement and possessor agreement are marginally more often sufxes than prexes,
but the difference is only of 1 to 3 per cent. However, even if agreement markers are not pri-
marily sufxing, there is still a mismatch between the number of languages where the source of
agreement (subject clitics) precedes the verb and the number of languages where agreement is
realized by sufxes, which is not expected under Givns assumptions.
. Moreover, note that in many languages left dislocation of 1st and 2nd person forms is
extremely marked, if not ungrammatical:
(i) ??I, I will read the book on clitics.
Similar observations hold with respect to left dislocation of personal pronouns in general, which
again seems to indicate that in these contexts, verbal agreement could only develop via analogy.
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.20 (1206-1283)
(ii) *Er,
he
der
pron
hat
has
das
the
Buch
book
gestohlen.
stolen
As for him, he stole the book.
. Note that many languages lack 3rd person personal pronouns. Instead, 3rd persons are re-
ferred to by full DPs, demonstratives (which may be a historical source for pronouns and verbal
agreement, cf. e.g. Siewierska 2004), or by nothing at all. In particular the use of demonstratives
instead of explicit 3rd person personal pronouns is quite widespread across the languages of the
world. Examples include Hindi/Urdu (Kachru 1990), Korean (Shin-Sook Kim, personal com-
munication), Mongolian (Poppe 1954, 1960), Pashto (MacKenzie 1990), Sanskrit (W. Lehmann
1993), and many native American languages (Mithun 1991). Given these facts, it is likely that
at least in some instances, the lack of 3rd person agreement can be attributed to the lack of an
appropriate lexical source for the grammaticalization process in question.
. Mithun (1991) observes that in the Athabaskan languages (spoken e.g. in Alaska, Western
Canada, Oregon, and California), 1st and 2nd person subject prexes occur very close to the
verb stem, while 3rd person subject markers occupy a position at the left edge of the verbal
inection, separated from the 1st and 2nd person markers by mode and aspect prexes. This
fact is then taken to indicate that the 1st and 2nd person markers were grammaticalized earlier
than the 3rd person markers. A similar point can be made with respect to number marking
in these languages. In general, person and number markers are fused forms in the case of 1st
and 2nd person subjects, which indicates that they are of considerable antiquity. In contrast,
number marking for 3rd person subjects occupies a separate slot in the inectional make-up of
the verb or has often barely been grammaticalized. Particularly revealing is the shape of transitive
pronominal afxes in a number of languages where afxes for agent and patient are adjacent
(e.g. Lakhota, Caddo, Maricopa). In general, afxes that mark 1st and 2nd persons show a high
degree of fusion, in contrast to those including 3rd persons (Mithun 1991: 87). In other words,
a 1st person subject acting on a 2nd person object is marked by a fused 1st/2nd person marker,
whereas a 1st person acting on a 3rd person often involves two separate person markers. Again,
the advanced phonological reduction of 1st/2nd afxes can be taken to indicate their special
antiquity.
. Note that the facts described by (6) may be blurred by the subsequent erosion of the verbal
agreement system, which may lead to such exceptional cases like English, where only 3rd person
subjects are marked on the verb (via the 3sg.pres.indic ending -s). See Chapter 6 for a discussion
of further apparent exceptions where the development of 3rd person markers seem to take place
parallel to or even predate the development of 1st and 2nd person markers.
. In Modern Irish, an overt subject requires the use of the so-called analytic form of the verb,
which is unmarked for person and number, whereas the fully inected synthetic form is not
allowed in these contexts. Note that rich person-number marking on verbs is more or less re-
stricted to the conditional, whereas other moods/tenses show less agreement morphology (cf.
McCloskey & Hale 1984: 492).
. Other possible analyses include incorporation of a subject pronoun into the verb (as dis-
cussed by McClosky & Hale 1984: 527f.) or checking of the EPP by rich verbal agreement with
pronominal properties (cf. Alexiadou & Anagnostopoulou 1998).
. Alternative sources of agreement morphology are not taken into consideration or only
briey commented on in this book (cf. Siewierska 2004 for an overview). These may include
the reanalysis of other inectional afxes as agreement markers (cf. Chafe 1977 on the 3rd per-
son subject prexes in Iroquoian), the incorporation of inected auxiliaries into the main verb
JB[v.20020404] Prn:18/07/2005; 15:02 F: LA8101.tex / p.21 (1283-1288)
(cf. Hetzron 1976 on Cushitic, Haas 1977 on Muskogean, Roberts 1993b, Roberts and Roussou
2003 on Romance) or analogical changes in which agreement markers originally conned to
a certain tense/mood or verb class gradually gain a wider distribution (cf. e.g. Lhr 1984 who
claims that Germanic bi-st 2sg be developed in analogy to the relevant agreement ending found
on the preterite-presents). Note that at least the latter two scenarios have nothing to say about
the origin of the agreement morphology which is transferred to other verbs (either via fusion or
analogy). However, it is in fact quite likely that this inection is the result of a previous process
in which pronouns were reanalyzed as agreement markers. Hence, it appears that at least some
of the alternative scenarios can be traced back to the reanalysis of pronouns as well.
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.1 (45-121)
Chapter 2
Theoretical preliminaries
. Introduction
This chapter serves to lay the theoretical foundations for the synchronic and di-
achronic analyses that are developed in the chapters to come. Since agreement is a
phenomenon that involves syntax and morphology in equal parts, I will give an outline
of both the syntactic and morphological mechanisms that participate in the realiza-
tion of predicate-argument agreement and the way these mechanisms interact with
each other. In addition, a formal account of the historical rise of agreement has to be
founded on a theory of language change and in particular the phenomenon of gram-
maticalization. Accordingly, this chapter introduces generative approaches to language
change (and variation) and grammaticalization in some more detail.
The chapter is organized as follows. Section 2.2 presents an outline of the current
version of minimalist syntax devised in Chomsky (2000, 2001a, 2001b). Section 2.3
provides an introduction into Distributed Morphology (Halle & Marantz 1993), the
morphological framework adopted in this work. In Section 2.4, I give an outline of the
generative approach to syntactic change in terms of a change in parameter values. Fi-
nally, Section 2.5 focuses on the phenomenon of grammaticalization, arguing that this
kind of change provides new phonological exponents for a closed class of functional
heads (Roberts & Roussou 2003).
. Minimalist syntax
This section serves to introduce the core properties of the syntactic framework adopted
in this work (the version of minimalism proposed in Chomsky 2000, 2001a, 2001b),
focusing on the aspects of the theory that are relevant for the implementation of
agreement.
Perhaps the most obvious difference between minimalism and previous versions
of the Principles and Parameters approach concerns the elimination of the represen-
tational levels D-structure and S-structure. Compare (1a), the so-called T-model (or
Y-model) assumed in work such as Chomsky (1981), (1986a), with (1b), the grammar
model proposed in Chomsky (1993) and subsequent work:
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.2 (121-183)
(1) a. b. Lexicon
D-structure
S-structure
PF LF
Move-
Move-
Lexicon
syntactic derivation
Spell-out
PF LF
The minimalist grammar model in (1b) involves only two (instead of four) levels of
representation (see Section 3 for a slightly revised form of (1b)). Accordingly, prin-
ciples and constraints that formerly applied to the levels D-structure or S-structure
are reformulated as conditions on derivations or as wellformedness conditions on the
structure of the remaining levels of representation PF and LF which constitute the
interface levels to other cognitive systems (bare output conditions).
Syntactic structures are built by the operations Merge and Move. The basic
structure-building operation Merge forms larger units by combining two elements X,
Y either drawn from the lexicon or constructed previously during the syntactic deriva-
tion. The syntactic category of the resulting structure is determined by properties of
either X or Y. The second operation that forms new phrase-markers is syntactic move-
ment, that is, the operation Move. Following Chomsky (1995), the operation Move can
be dened as follows:
(2) Given the phrase marker with syntactic objects K and , Move targets K,
raises , and merges with Kto formthe newcategory with the constituents
, K.
Thus, the basic structure-building operation Merge is also part of the more com-
plex operation Move. In minimalist syntax, it is generally assumed that the syntactic
derivation is morphologically driven. More precisely, syntactic operations (Move) are
triggered by the need to eliminate certain morphological features which cannot be
interpreted at the interfaces to other cognitive systems (PF, LF). Failure to delete non-
interpretable features prior to the interface level where they are visible causes the
syntactic derivation to crash. These features are also referred to as formal features,
since they are associated with formal properties of grammar such as tense, case and
agreement. In earlier versions of the minimalist program (Chomsky 1993, 1995), it is
assumed that functional heads host formal features such as [Nominative], [Past], and
-features (e.g. [person], [number], [gender]) which are deleted by entering into a
checking relation with identical features on substantial lexical categories such as N, V
or A. The latter are combined with inectional afxes in the lexicon and are inserted
fully inected. Checking relations are established by either head movement that ad-
joins a lower head to a higher functional head or XP-movement into the specier of
a functional head. Overt movement is triggered by strong non-interpretable features
which are visible at PF, whereas weak non-interpretable features can be eliminated
by covert movement operations that take place between Spell-out and LF. Economy
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.3 (183-235)
Chapter 2. Theoretical preliminaries
principles such as Procrastinate ensure that movement and feature checking must be
postponed to the covert syntax if possible. In general, considerations of derivational
economy play an important role in the implementation of syntactic operations. For
example, in contrast to the nature of the operation Move which applies freely (if not
ruled out by independent principles), syntactic movement is considered a Last Resort
operation that may be triggered only by the need to check off non-interpretable fea-
tures. Locality is attributed to the workings of economy principles as well. Conditions
such as Shortest Move/the Minimal Link Condition or Merge over Move require that at
any given point in the syntactic derivation, the computational system must carry out
the most economic operation, preferring less complex over more complex derivations.
In Chomsky (1993, 1995), syntactic agreement is established via a specier-head
conguration between an argument (the agreement controller) and the predicate (the
agreement target). Thus, (overt) agreement checking is parasitic on A-movement to
the specier of a functional head (Chomsky 1993: AgrS/AgrO; Chomsky 1995: T/)
which also attracts the inected verb. By assumption, overt movement is triggered by
strong categorial features (e.g. [*D], [*V], where the star indicates feature strength)
which are part of the set of formal features of a given functional head and must be
eliminated prior to Spell-out.
1
The minimalist analysis of agreement is completely revised in Chomsky (2000,
2001a, 2001b), where a rather different model of the computational system and its
basic components is developed.
2
First, on this view, the syntactic operations which
are necessary to check off (or, rather, value) the set of non-interpretable/non-valued
-features located in T/ are triggered by the -set itself (and not by the presence of
non-interpretable categorial [D]/[N] or [V] features on T/). Second, feature check-
ing/valuing is dissociated from (overt) DP-movement into the specier of the relevant
functional head. Instead, the set of non-interpretable -features located in T/ may
access the interpretable -set of an argument which stays in situ. The latter operation
is called Agree which establishes a relation (agreement, Case checking) between an LI
[lexical item] and a feature F in some restricted search space (its domain) (Chom-
sky 2000: 101). A feature set that starts an Agree operation is referred to as a probe.
A probe seeks to establish a relation with another set of matching features which is
called the goal. In the case at hand, the unvalued -set in T/ acts as a probe that seeks
a set of matching interpretable -features (the goal) with which it can establish agree-
ment. Similar to Chomsky (1993, 1995), -features are assumed to be interpretable
on DP/NP, but uninterpretable on T/. More precisely, Chomsky (2000, 2001a) pro-
poses that the value of the -features on T/ is not specied. When the set of unvalued
-features succeeds in establishing an Agree relation with the -set of a nominal argu-
ment, the value of the latter is assigned to the previously unvalued -set. Afterwards,
the uninterpretable -set on T/ is marked for deletion (ultimate deletion is post-
poned to the point where the derivation reaches Spell-out). The core properties of the
probe/goal mechanism can be summarized as follows (Chomsky 2000: 122):
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.4 (235-295)
(3) Matching is a relation that holds of a probe P and a goal G. Not every match-
ing pair induces Agree. To do so, G must (at least) be in the domain D(P) of P
and satisfy locality conditions. The simplest assumptions for the probe-goal
system are:
a. Matching is feature identity.
b. D(P) is the sister of P.
c. Locality reduces to closest c-command.
In this framework, the signicance of structural Case is considerably diminished. By
assumption, Case is not part of the feature make-up of the probes T/. Therefore,
it cannot trigger checking operations, since Agree requires feature matching between
probe and goal. Still, Chomsky assumes that there is an intricate relation between
structural Case and -feature checking. More precisely, an uninterpretable Case fea-
ture renders the goal active, that is, accessible for the operations Agree and Move. The
establishment of an Agree relation then leads to the erasure of uninterpretable fea-
tures in both the probe () and the goal (Case).
3
In other words, an uninterpretable
-set (the probe) can only be marked for deletion by a goal with an unchecked Case
feature. Accordingly, a phrase with a checked/deleted Case feature cannot undergo fur-
ther A-movement and is frozen in place (Chomsky 2000: 123).
4
Thus, we must add
the following statement to the characterization of the probe/goal mechanism in (3)
above (Chomsky 2000: 123):
(4) The operations Agree and Move require a goal that is both local and active.
By assumption, the activating Case feature can only be deleted by a probe with a
complete set of -features. For example, the defective T head (T
def
) of non-control
innitivals hosts only a [person] feature. Movement of the subject to SpecTP deletes
the uninterpretable [person] feature, but does not lead to deletion of the subjects
structural Case feature. Accordingly, the subject of a raising innitival can be targeted
by further syntactic operations (Agree and Move triggered by a probe in the matrix
clause). In a similar vein, the reduced feature content of an expletive (presumably only
[person] as well) fails to value and delete the uninterpretable -set of nondefective
T which may then establish long-distance agreement with an associate DP (in situ)
via Agree:
(5) There were [three men] declared guilty.
Agree
Overt phrasal movement is parasitic on a previously established Agree relation. It is
triggered by the requirement that a given functional head must project a specier. This
is implemented by the assumption that functional heads may host an EPP feature. The
EPP feature is assumed to be universally present on T, while its presence on C (wh-
movement) and (object shift) is subject to parametrization. Thus, Move is a complex
operation which consists of the basic operations Agree and Merge:
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.5 (295-357)
(6) Move of , targeting , has three components (Chomsky 2000: 135):
a. A probe P in the label L of locates the closest matching G in its domain.
b. A feature G of the label containing G selects a phrase as a candidate for
pied-piping.
c. is merged to a category K.
As in previous work (Chomsky 1995), it is assumed that the more basic operation
Merge preempts Move (Merge over Move). In addition to the reformulation of syn-
tactic movement and feature checking, Chomsky (2000) devises a new concept of
locality that is based on the hypothesis that the syntactic computation proceeds via
certain derivational stages which are called phases. By assumption, CP and P consti-
tute phases, but not TP or VP.
5
This yields a strong form of cyclicity: neither may the
head of a phase trigger operations after is completed, nor can operations triggered
by higher heads access elements inside a completed phase apart from its head and
speciers. These conditions can be stated as follows (Chomsky 2000: 107f.):
(7) The head of a phase is inert after the phase is completed, triggering no
further operations.
(8) Phase Impenetrability Condition (PIC)
In phase with head H, the domain of H is not accessible to operations
outside , only H and its edge are accessible to such operations.
Accordingly, material contained in a phase must move to the left edge of before it
can enter into an Agree or Move operation triggered by a higher probe. For example,
the PIC requires that A-movement must target the edge of every phase (CP, P) that
lies between the foot and the head of the resulting movement chain. From the PIC, it
follows that the left edge of a phase is accessible to syntactic operations until the next
higher phase is completed.
Furthermore, phases are associated with Spell-out. That, is at every phase level, the
complement of the phase head is sent to the interface levels PF and LF after the phase
has been completed (so-called cyclic Spell-out). The phase head and its speciers
are sent to the interfaces with the complement of the next higher phase head. For
example, the VP complement of is subject to Spell-out after P has been completed,
while and its speciers are spelled-out together with the TP complement of the next
higher C head. This entails that there are no covert syntactic operations that take
place after Spell-out (i.e., there is no LF-movement). Rather, all syntactic operations
(Merge, Agree, Move) must take place prior to Spell-out, independent of whether they
have visible effects at PF or not.
To summarize, in the present section I gave an outline of the syntactic frame-
work adopted in this book. Importantly, an analysis of agreement in terms of Agree no
longer requires the creation of a spec-head conguration during the syntactic deriva-
tion. In Chapter 3, I will present further evidence that this approach is on the right
track, drawing on evidence from long-distance agreement in the Nakh-Dagestanian
language Tsez. Note that the basic assumptions laid out in the present section will be
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.6 (357-400)
slightly modied in the chapters to come. In Chapter 3 (Section 3.6.2) it is argued
that Case and agreement licensing may be dissociated (i.e., checked off by different
functional heads). This requires some modications. First, Case checking does not
necessarily require the existence of an unvalued -set in the probe. Second, a DP with
a Case feature marked for deletion may still be accessible for a further Agree oper-
ation that values the uninterpretable -features of a higher Agr-head (see below for
discussion).
In the introduction to this chapter, I noted that agreement is a phenomenon on
the borderline between syntax and morphology. Accordingly, the next section presents
a set of basic assumptions on the role and workings of morphology in the realization
of agreement and on the nature of the interface between syntax and morphology.
. Distributed Morphology
The analysis of agreement presented in this book is framed in a realizational model
of grammar where word building operations are distributed over several components
of grammar (Distributed Morphology, henceforth DM; Halle & Marantz 1993, 1994).
DMassumes that the morphological component operates post-syntactically, interpret-
ing the output of the syntactic derivation. In other words, the morphological com-
ponent mediates between the syntactic and the phonological modules of grammar.
Accordingly, the structural design of the grammar looks like (9).
6
(9) Lexicon ( ) morphosyntactic/semantic features
Syntactic derivation
Spell-out
Morphology LF
PF
In this model of grammar, the syntactic operations Merge and Move operate on bun-
dles of morphosyntactic features that constitute syntactic terminal nodes (i.e., heads).
The syntactic terminal nodes are referred to as morphemes.
7
In the morphological
component, the terminal nodes are associated with phonological exponents in a pro-
cess called Vocabulary Insertion. The idea that phonological content is added after
syntax is also known as Late Insertion.
8
The information that links phonological ex-
ponents with morphosyntactic features (i.e., insertion contexts) is stored in individual
Vocabulary items. For example, the English verbal inection 3sg.pres.indic. /-z/ is as-
sociated with the following Vocabulary item (which can be read as an insertion rule:
the phonological exponent /-z/ is inserted in the context [3, sg, pres., indic.]):
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.7 (400-470)
(10) [3, sg, pres., indic.] /-z/
The insertion procedure requires that the feature specication of the Vocabulary item
is nondistinct from the features of the insertion site (i.e., a certain morpheme). Usu-
ally, this requirement is met by several items, which then enter into a competition.
The item that realizes the greatest subset of features is chosen for insertion. In the ex-
ample at hand, the availability of the Vocabulary item in (10) blocks the insertion of
the less specied exponent -, which is found in all other contexts, representing the
elsewhere case (cf. Kiparsky 1973; Aronoff 1976; Anderson 1986 for the use of else-
where principles in phonology/morphology). Thus, the paradigm exhibited in (11) is
captured by positing merely two Vocabulary items (listed in (12)) for the realization of
In in English.
(11) a. I/you/we/you-pl/they sleep.
b. He/she/it sleep-s.
c. *He/she/it sleep.
(12) a. [3, sg, pres., indic.] /-z/
b. elsewhere -
This approach implies that Vocabulary items may be underspecied for the feature
complexes they realize. In (12), the exponent - is maximally underspecied since
it is not linked to any feature specication at all (the elsewhere case). The basic prop-
erties of the insertion procedure are captured by Halles (1997: 428) Subset Principle
(not to be confused with the Subset Principle of Wexler & Manzini 1987):
9
(13) The Subset Principle
The phonological exponent of a Vocabulary item is inserted into a morpheme
in the terminal string if the item matches all of a subset of the grammatical
features specied in the terminal morpheme. Insertion does not take place if
the Vocabulary item contains features not present in the morpheme. Where
several Vocabulary items meet the conditions for insertion, the item match-
ing the greatest number of features specied in the terminal morpheme must
be chosen.
Most work in DMacknowledges a systematic difference between lexical and functional
categories (Halle 1990: concrete vs. abstract morphemes; Harley & Noyer 1999:
l-morphemes vs. f-morphemes). Functional morphemes constitute a closed class
and contain features relevant for the syntactic computation (e.g., [plural], [+past],
[+wh]). In addition, they presumably carry semantic features associated with a cer-
tain set of logical meanings (cf. Section 2.5 below). A property specic to functional
morphemes is that their spell-out is deterministic, that is, the feature content of a given
functional morpheme serves to determine a unique phonological realization. Thus, the
insertion of Vocabulary items realizing functional morphemes is guided by the Subset
Principle which ensures that the most specic candidate wins out over its competitors.
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.8 (470-531)
In contrast, the spell-out of open class lexical morphemes is non-deterministic
and not constrained by a competition between compatible Vocabulary items. Thus, in
a context (i.e., a certain lexical morpheme) which requires the insertion of a nomi-
nal head, any Vocabulary item that is compatible with this specication can be chosen
for insertion (dog, cat, house, ball etc.). In other words, for lexical morphemes, there
is a choice in spell-out which does not exist for functional morphemes and which
I refer to as free-choice spell-out.
10
This fact is reected by the shape of insertion
rules/Vocabulary items for lexical categories which are identical for all members of
a given lexical category, cf.
11
(14) a. [
DP
D [
NP
___ ]] /dog/
b. [
DP
D [
NP
___ ]] /kt/
etc.
Prior to Vocabulary Insertion, a set of morphological operations may apply to the
output of the syntactic component, changing the content and (hierarchical) structure
of morphemes. The most important of these are the insertion of so-called dissociated
morphemes, (morphological) Merger or Fusion, Fission and Impoverishment. In the fol-
lowing, these mechanisms are only briey introduced; they are discussed in more detail
when they have a concrete bearing on the issues dealt with in this book.
The constituent structure of morphemes derived in the syntax can be modied
by the post-syntactic insertion of (functional) dissociated morphemes which may at-
tach to other functional morphemes. Following Embick (1997), these morphemes are
called dissociated, since they are not present in the syntactic derivation and merely
reect properties expressed by structural congurations in the syntax proper. Within
DM, this mechanism is often used to account for case and agreement phenomena. For
example, Marantz (1992), Halle and Marantz (1993), Halle (1997) analyze subject-
verb agreement in terms of the post-syntactic adjunction of an [Agr] morpheme to
T (cf. Embick 1997 for a detailed discussion of the insertion of dissociated mor-
phemes). See Chapter 3 for an analysis of complementizer agreement in Germanic
in terms of dissociated Agr-morphemes. In Chapters 46, it is argued that dissoci-
ated Agr-morphemes play an important role in the historical rise of new agreement
morphology as an intermediate step on the grammaticalization path from clitics to
syntactic agreement markers.
Another morphological operation that may change the constituent structure gen-
erated by the syntax is Morphological Merger of terminal nodes that do not form a
constituent in the syntactic output (cf. Marantz 1984, 1988). For our purposes, the
most important instance of Merger are cases of apparent syntactic lowering, that is,
syntactic heads that are not joined together via head movement, but still are spelled
out as a unit. A prominent example of this type of Merger is afx-hopping in En-
glish, the post-syntactic afxation of Tense/Agr to the main verb, which on standard
assumptions fails to undergo overt head movement in English (cf. Bobaljik 1995, 2002,
2003).
12
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.9 (531-582)
Arelated operation is Fusion which may create a mismatch between the number of
underlying morphemes and the number of inserted Vocabulary items, in the sense that
two (or more) syntactic nodes are fused into a single terminal node which is then real-
ized by a single phonological exponent. Again, the verbal inection of English is a case
in point: it is commonly assumed that Agr and T fuse into a single morpheme prior to
Vocabulary Insertion (Halle & Marantz 1993; Halle 1997). Instances of Fusion are tra-
ditionally referred as cumulative exponence and may involve quite a number of different
inectional features as in the verbal inections of Latin (cf. Matthews 1991: 233): in a
1st person singular present indicative active form like am- o I love, the single exponent
- o serves to simultaneously realize ve inectional features, namely person, number,
tense, mood, and voice.
13
Similar to Fusion, Fission (Halle &Marantz 1993; Halle 1997; Noyer 1997) also dis-
turbs the isomorphism between syntax and morphology, albeit with contrary effects:
while Fusion leads to the amalgamation of two separate syntactic terminals, Fission re-
sults in a situation where a single syntactic terminal node is realized by more than one
Vocabulary item. The concept of Fission is related to the notion that the insertion of
Vocabulary items discharges the inectional features present in the morpheme (Noyer
1997). In standard cases, the insertion procedure stops after a phonological exponent
is inserted, even if this exponent discharges only a subset of the inectional features
present in the morpheme. However, when a morpheme is marked for undergoing Fis-
sion, the inectional features that are not discharged (or matched, Halle 1997) by the
rst insertion operation are copied into an additional morpheme which is generated
by the insertion procedure. Subsequently, this additional morpheme is then itself sub-
ject to Vocabulary Insertion. Typical examples of Fission come from languages where
agreement is marked by a combination of prexes and sufxes, as in many Afro-Asiatic
languages (cf. Noyer 1997 for extensive discussion).
14
This is illustrated by the follow-
ing example from Tamazight Berber where the agreement morpheme splits into three
positions of exponence (Noyer 1997: 89) which are realized by successive Fission (of
a single Agr-morpheme) and insertion of the Vocabulary items in (16):
(15) t-dawa-n-t
2-cure-pl-fem
you (pl, fem) cured (Noyer 1997: 89)
(16) a. [2] /t-/
b. [pl] /-n/
c. [fem] /-t/
Due to the fact that morphemes created by Fission contain only a subset of the fea-
tures contained in the original morpheme, Fission often results in the insertion of less
marked or elsewhere Vocabulary items, that is, Fission extends the domain of less
marked exponents (Halle 1997: 432).
The insertion of less marked exponents can also follow from context-sensitive Im-
poverishment rules which delete morphosyntactic features from morphemes prior to
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.10 (582-642)
Vocabulary Insertion (cf. Bonet 1991 for the original proposal). As a result, Vocabulary
items that require the presence of those features cannot be inserted and a less specied
exponent must be used to realize the morpheme affected by Impoverishment (which
Halle & Marantz 1994 call the Retreat to the General Case).
An illustrative example of the workings of Impoverishment is provided by Sauer-
land (1996). It is a well-known fact that the Germanic languages show different
paradigms for the inection of (attributive) adjectives. The choice of the weak in-
ectional ending on adjectives is determined by the presence of inection on the
determiner: in cases where a denite determiner is present, a weak ending must ap-
pear on the adjective (de-r gut-e Mann the good man-nom, de-m gut-en Mann the
good man-dat). However, if no determiner is present, the strong adjectival inection
must be chosen (mit gut-em Gewissen with good conscience-dat). In the presence of
an indenite article, a mixed paradigm is observed the strong adjectival inection
appears if the article shows no agreement/inection (ein gut-er Mann a good man-
nom), while the weak adjectival inection shows up if the indenite article carries
agreement/inection (ein-em gut-en Mann a good man-dat). Interestingly, Sauer-
land observes that the endings used in the weak paradigm are in fact a subset of the
strong endings, namely the least marked (or default) endings of the strong inection.
He therefore concludes that the weak paradigm is the result of an impoverishment
rule that deletes inectional features in the presence of an inected determiner (in
most languages [gender] is deleted; the picture is more complex in languages with
rich inection such as German and Icelandic, see Sauerland 1996 for details). This in
turn leads to the insertion of less specied forms that constitute the most unmarked,
underspecied Vocabulary items of the strong paradigm.
15
From the above discussion, it has become clear that DMposits a strongly syntactic
approach to the formation of words: morphosyntactic primitives are combined via the
syntactic operations Merge and Move and then (eventually) spelled-out by phonolog-
ical exponents. This model seems to entail (or at least favor) a very complex clause
structure where there is ideally a one-to-one correspondence between morphosyntac-
tic features and terminal nodes (cf. e.g. Sigursson 2000, 2001, 2002; Poletto 2000 for
the idea that features such as person and number head separate projections), that is,
there are separate projecting nodes for individual inectional categories such as per-
son, number, gender etc. However, the idea that the syntactic component manipulates
morphosyntactic primitives does not necessarily require that each individual feature
heads its own projection (cf. Harley & Noyer 1999). First, the possibility to insert (dis-
sociated) morphemes post-syntactically entails that not every morpheme (and thus
feature) enters the syntactic computation as a projecting head. Second, if we agree on
the existence of a purely morphological operation such as Fission, then it follows that
the syntax must operate (at least sometimes) on bundles of morphosyntactic features
which can then be split up into several morphemes by post-syntactic morphological
operations. Note that Fission may give the (false) impression that this split of inec-
tional features/heads is located in the syntax. Thus, we can maintain our assumption
that there is a small core inventory of functional nodes (C, T, , D) which is universally
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.11 (642-709)
present in each human language. Wider arrays of functional categories are possible
(such as e.g. split-CP, Rizzi 1997), but only triggered if there is relevant evidence in the
input data during language acquisition. Still, morphological operations such as Fission
and the insertion of dissociated morphemes may give rise to the (wrong) impression
of a richer system of functional heads present in the syntax.
After establishing the main traits of the theoretical framework adopted in this
book, the remainder of this chapter presents a set of assumptions on the theory of lan-
guage change (and variation) and in particular the phenomenon of grammaticaliza-
tion that lay the foundations for the diachronic analyses developedin the Chapters 46.
. Language change: A generative perspective
In the last twenty years, the investigation of diachronic phenomena has developed
into a productive and well-established sub-discipline of generative linguistics (for an
overview of more recent work cf. the volumes edited by Battye & Roberts 1995; Beck-
man 1995; van Kemenade & Vincent 1997; Pintzuk, Tsoulas, & Warner 2000, Lightfoot
2002). Following the pioneering work of scholars such as David Lightfoot, Anthony
Kroch and Ian Roberts, there is by now general agreement on the notion that the study
of language change can provide important insights into the properties of Universal
Grammar that cannot be gained from a purely synchronic perspective. This convic-
tion stems from the insight that the triggers for language change are located in the
workings of language acquisition, that is, the interaction between innate principles of
grammar and the linguistic evidence the child encounters, reviving ideas that were al-
ready present in the work of neo-grammarians such as Hermann Paul (e.g., Paul 1880).
From this point of view, language change reects the limits to attainable grammars
(Lightfoot 1991: 172), in the sense that language change reveals how certain grammar-
specic choices which give rise to individual languages are selected on the basis of the
evidence available to the learner.
In the Principles and Parameters framework, (syntactic) variation between indi-
vidual languages is attributed to different settings for a limited number of parameters
that are associated with invariable principles of Universal Grammar (UG) (Chomsky
1981, 1986a). The task of acquiring a given grammar then consists of lling in the gaps
left open by the principles of UG, that is, detecting the parameter settings which are
reected by the linguistic input the child is confronted with. The stimulus that serves
to set a given parameter one way or other is usually called trigger experience or cue (see
Lightfoot 1991, 1999 for discussion). Apart from accounting for synchronic paramet-
ric differences between different individual languages (such as e.g. present-day English
and Italian), this approach can also be applied to diachronic parametric differences
between different historical stages of a single language (such as e.g. Old English and
present-day English). Accordingly, syntactic change is often conceived of as a change
that affects the parameter settings for a given language.
16
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.12 (709-754)
Due to the fact that a single parameter typically determines a whole complex of
(syntactic) properties, a single parametric change manifests itself in a variety of com-
plex effects and at times dramatic changes at the syntactic surface. In other words,
parametric change is often abrupt and catastrophic in nature (Lightfoot 1991, 1999).
An instructive example of the workings of parametric change is given in Roberts
(1993a) who shows that in the history of French, three apparently independent changes
(the loss of V2, null subjects, and so-called simple inversion in interrogatives) can
actually be analyzed as reexes of a single parametric change that affected the mecha-
nism of nominative case assignment. Furthermore, Roberts links the syntactic change
in question with morphophonological changes that led to the loss of distinctive inec-
tional endings. Thus, morphological properties of individual lexical items are assumed
to play an important causal role for setting off changes that affect the syntactic com-
ponent of the grammar (see below for details).
This assumption is in line with recent approaches to parametrization which take
parameters to be lexical, that is, they are associated with properties of individual lex-
ical items (the Lexical Parametrization Hypothesis, cf. Borer 1984; Wexler & Manzini
1987). In the late 1980s, the Lexical Parametrization Hypothesis converged with an-
other theoretical development within the framework of Principles and Parameters
theory that aimed at implementing the traditional distinction between lexical and
grammatical categories directly into the structure of the clause. Following work by,
for example, Fukui (1986), Chomsky (1986b), and Abney (1987), grammatical cat-
egories like determiners, conjunctions and inections were associated with a closed
class of functional heads that project their featural content in accordance with univer-
sal principles of phrase structure (X-bar theory or Bare Phrase Structure). Initially, in
the verbal/clausal domain only two functional heads were distinguished: C (conjunc-
tions, clause type, subordination) and In (verbal inection, niteness, nominative
case assignment). In later work by Pollock (1989), Belletti (1990), Chomsky (1991),
the category In is split further into distinctive AgrS (subject agreement, nominative
case assignment), T (tense, niteness), and AgrO(object agreement, accusative case as-
signment) heads. In the 1990s, this line of research has led to a sudden increase of the
number of functional categories (cf. e.g. Cinque 1999). By now, most researchers agree
on the existence of a universal inventory of core functional categories that consists of
the elements C (clause type, subordination), T (tense, subject-verb agreement, nomi-
native assignment), (voice, transitivity, accusative assignment, object agreement) and
D (nominal inection, deniteness) (cf. Chomsky 1995, 2000, 2001a, 2001b). Note,
however, that many scholars today argue for the existence of more elaborate systems
that include a ne-grained structure of inectional/functional heads in the Inand C
domain (cf. Rizzi 1997; Cinque 1999, among others).
At least with respect to the set of core functional categories it is assumed that
they are universally present in natural languages as basic building blocks or skeleton
of clause structure and that they trigger syntactic operations in order to license their
(abstract) morphological content (cf. Chomsky 1991, 1995, 2000). On these assump-
tions, it seems natural to analyze parametric variation between individual languages
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.13 (754-822)
in terms of varying (morphological) feature specications of a (universally) given set
of functional categories (cf. e.g. Chomsky 1991, 1995; Ouhalla 1991):
If substantive elements (verbs, nouns, etc.) are drawn from an invariant universal
vocabulary, then only functional elements will be parametrized.
(Chomsky 1991: 419)
In this approach to parametrization, the computational system (C
HL
), that is, the
mechanisms that build up syntactic hierarchical structures by combining individual
lexical items, is universal and therefore cross-linguistically and diachronically invari-
able (Chomsky 2000, 2001a, 2001b). Thus, strictly speaking, there is no such thing
as syntactic change: the properties of the syntactic component of grammar remain
constant over time (cf. Hale 1996, 1998; Longobardi 2001; Keenan 2002). Appar-
ent syntactic change (and synchronic differences between languages) results from
changes affecting the feature content of a closed class of functional categories such
as C, T, and D (e.g., via phonological erosion or grammaticalization processes). It
is expected that a single change in the featural properties of these core functional cat-
egories results in a complex of (at times dramatic) distinct changes on the syntactic
surface, which is a hallmark of parametric change (cf. Lightfoot 1979, 1991, 1999).
Moreover, if overt inectional morphology is taken to reect the (abstract) fea-
ture content of functional categories, it is possible to construe a correlation between
parametric change and morphological change, that is, to provide a principled explana-
tion for the traditional observation that changes affecting the inectional morphology
of a given language often go hand in hand with syntactic change (cf. e.g. Paul 1880;
Sapir 1921). In the generative literature on language change, this line of research has
proven to be very productive, covering for example the historical impact of the loss
of verbal inection on the availability of verb movement (for the history of English
cf. Roberts 1993a; for the Scandinavian languages cf. Platzack 1988 on Swedish) and
pro-drop (for the history of French cf. Roberts 1993a; Vance 1997; for Swedish Falk
1993; for English Allen 1995; Haeberli 1999), or the relation between the loss of nom-
inal inections (i.e., case) and changes affecting argument order and the rise of ECM
constructions (for the history of English cf. Lightfoot 1979, 1991, 1999; van Kemenade
1987; Roberts 1997; Kiparsky 1997; Haeberli 1999). However, while the loss of inec-
tional morphology (and its syntactic consequences) has received quite some attention,
the related question of how inectional morphology arises historically has remained
largely unadressed in generative works, although the latter constitutes a major eld of
interest in traditional and typological approaches to language change (see Section 2.5
for discussion).
Still, the generative approach to language change has not gone unchallenged. In
particular the concept of parametric change seems to be in conict with two widely-
held views on language change and acquisition. First, it is a traditional observation that
language change is a gradual process. A given change may go on over many centuries
before it is completed (cf. e.g. the change from OV to VO in the history of English,
Lightfoot 1979, 1991, 1999; Kemenade 1987; Roberts 1993a; Pintzuk 1999; Trips 2002;
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.14 (822-885)
Fu & Trips 2002). At rst sight, this trait of language change does not seem to be
compatible with the abrupt character of parametric change. Another problem arises
from the assumption widely held in generative linguistics that language acquisition
proceeds in a awless fashion. In other words, most researchers (working in the gen-
erative paradigm) agree that children always succeed in acquiring the target grammar
that generates the linguistic data they are exposed to, even if this data is apparently
awed and insufcient (sometimes called the the logical problem of language ac-
quisition, cf. Chomsky 1986a for discussion). This leads to the logical problem of
language change as Niyogi and Berwick (1998) choose to call it:
After all, if all children successfully attain the grammars of their parents and they
continue to do this generation after generation, then the linguistic composition of
every generation would look exactly like the linguistic compositionof the previous
generation and languages would not change with time. Yet they do.
(Niyogi & Berwick 1998: 192f.)
Thus, more has to be said to reconcile the idea of parameter change with (i) the ap-
parent gradualness of language change and (ii) the apparent perfection of language
acquisition. To answer the latter problem, it is usually assumed that for some reason,
the trigger experience that resulted in a given parameter setting in the parent grammar
has become obscure or ambiguous in the output of this grammar (cf. Lightfoot 1979,
1991) due to factors such as language contact, (morpho-) phonological erosion or syn-
tactic reanalyses that blur the evidence for certain parametric choices in the linguistic
input. Of course, this raises a number of further questions concerning the nature of
the trigger experience (see below for some discussion). Note furthermore that ambi-
guity of trigger experience is only a necessary condition for a change to take place.
It is still left unclear what actually motivates parametric change. Here, a widespread
line of thinking assumes that principles of economy take over and ultimately decide
if the trigger experience contained in the input is truly ambiguous (Clark & Roberts
1993; Roberts 1993b; Roberts & Roussou 2003, among others). In general, these econ-
omy considerations come in two varieties: rst, some researchers assume that there
are marked and unmarked (or default) parameter values and that the learner assigns a
given parameter the unmarked value if no decision can be made based on the evidence
available in the input (cf. Wexler & Culicover 1980; Berwick 1985 for discussion).
A related idea lies behind the Subset Principle (Berwick 1985: 37; Wexler & Manzini
1987: 61):
17
(17) The Subset Principle (OGrady 1997: 283)
The acquisition device selects the most restrictive parametric value consistent
with experience.
Thus, if there are two possible settings A and B for a given parameter, and A generates
a subset of the sentences generated by B, the learner will acquire the more restrictive
setting A in the absence of clear trigger experience. In this sense, A can be viewed as
the unmarked setting.
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.15 (885-925)
Alternatively, the decision in question is assumed to be sensitive to the notion of
derivational/representational economy (e.g., a Least Effort Strategy, Clark & Roberts
1993), in the sense that the learner assigns a given input string the most economical
representation/derivation that is compatible with the data. If the manifestation of a
given marked parameter value is ambiguous in the input, the learner will choose the
parameter value that guarantees the most economical representation/derivation. These
considerations are usually employed to account for the diachronic loss of movement
operations (if the data is compatible with a non-movement analysis), as for example
the loss of V-to-I or object movement (for Case licensing) in the history of English (cf.
e.g. Clark & Roberts 1993; Roberts 1993a, 1995, 1997, 1999a). Most recently, Roberts
and Roussou (2003) argue that the kind of structural simplication associated with
the loss of movement operations is a key feature of grammaticalization processes, in
the sense that an element previously moved to the vicinity of a functional head X is
reanalyzed as the exponent of X (see this chapter, Section 2.5 and Chapters 4 and 5 for
discussion).
A different approach is presented in Lightfoot (1999) who argues for a cue-based
theory of language acquisition, originally developed by Dresher and Kaye (1990) and
Dresher (1999) for the acquisition of phonological properties such as stress patterns.
The basic assumption is that UG contains not only a set of parameters, but also speci-
es for each parameter a cue that serves to switch the parameter one way or other (cf.
Fodor 1998 for a related approach). According to Lightfoot, a cue is not directly present
in the input, but rather part of the abstract mental representations/structures derived
by parsing the input (i.e., a piece of I-language in the sense of Chomsky 1986a). For
example, Lightfoot assumes that the cue for the V2 property is
SpecCP
[XP], XP an arbi-
trary phrasal category that occupies SpecCP. This form of abstract trigger experience
is contained in a (incomplete) structural description of an input string, but not in the
input itself. In other words, a cue can only be detected if the learner has assigned a
given input string a structure. If the learner detects a cue that is attested robustly in
these (initially incomplete) parses, this will activate a given parameter or syntactic op-
eration in the learners grammar.
18
In other words, cues are points of variation the
absence/presence of which determines differences between individual grammars. Lan-
guage change results either if a given syntactic operation fails to be cued or if it starts
to be cued, in contrast to the target grammar. Thus, instead of assuming the existence
of binary parameters, it is perhaps more appropriate to speak of basic or extended
parametric choices, where the cues for the latter are present only in certain grammars.
From this point of view, language change is a contingent and unpredictable process
that depends on the availability of certain cues in the linguistic input the learner is ex-
posed to. Often, a given parametric choice is associated with a set of cues that may be
syntactic or morphological in nature. For example, it is usually assumed that the cue
for V-to-I movement consists of (i) syntactic evidence that the verb has moved out of
VP (i.e., the conguration
In
[V
n
]) and (ii) morphological evidence, namely the pres-
ence of rich verbal agreement morphology.
19
In the context of the present study, then,
the question arises which cues are involved in the acquisition of agreement morphol-
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.16 (925-995)
ogy. In Chapter 4 below, I will discuss a set of morphological and syntactic signals that
help the learner to decide whether a given element should be categorized as a piece of
verbal inection or a pronominal clitic.
Crucially, the expression of cues interacts in intriguing ways with core syntactic
properties of grammar. Thus, we should expect that the categorial reanalysis of a pro-
noun may at times provoke catastrophic changes in other parts of the grammar (see
Chapters 4 and 5 for discussion and relevant examples). This is in line with Light-
foots tenet that the learner does not try to match the input (in contrast to most other
approaches, cf. e.g. Clark & Roberts 1993; Gibson & Wexler 1994), but rather scans
structures for cues that trigger certain parametric choices, without regard to the nal
result (Lightfoot 1999: 149), that is, whether the new grammar will differ from the
target grammar or not. The latter aspect ts in well with the view that language change
is an abrupt process that may at times lead to catastrophic consequences, but seems to
be at odds with the often noted apparently gradual nature of language change, which
brings us back to the rst problem noted above.
It has been argued that the often gradual character of language change represents
a serious problem for the concept of parametric change which predicts that languages
should change rather abruptly (cf. Hopper & Traugott 1993; Harris & Campbell 1995
for discussion). In what follows, however, it is shown that the impression of gradualness
arises from two factors which traditional approaches failed to recognize and that we
can therefore maintain the concept of parametric change. First, I demonstrate that
the impression of gradualness disappears under a sufciently constrained notion of
grammar (and grammar change) as the proper object of linguistic study. Second,
it is claimed (following the pioneering work by Kroch 1989) that language change
typically proceeds via a stage of internal bilingualism, where generations of speakers
have command over more than one internalized grammar giving rise to a degree of
linguistic variation which gradually disappears when one grammar wins out over the
other over time.
In generative linguistics, there is general agreement that the object of fruitful lin-
guistic study is a special notion of language, namely the linguistic knowledge of an
idealized speaker/hearer, cf. Chomsky (1965: 3) for a classic statement:
Linguistic theory is concerned primarily with an ideal speaker-listener, in a com-
pletely homogenous speech-community, who knows its language perfectly and is
unaffected by such grammatically irrelevant conditions as memory limitations,
distractions, shifts of attention and interest, and errors (randomor characteristic)
in applying his knowledge of the language in actual performance.
On this conception of language, the proper object of linguistic study is generally re-
ferred to as linguistic competence or simply grammar. Accordingly, the proper object of
the linguistic study of language change must be dened as a change between (individ-
ual) grammars, which I will refer to as grammar change (cf. Hale 1996; Lightfoot
1999). It is fairly clear that in this sense, language change is necessarily an abrupt
phenomenon, namely a clearly identiable difference between a grammar A and a
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.17 (995-1049)
grammar B. Hence, the impression of gradualness is in fact a misperception that arises
from confusing the actual change process with the (highly gradual) diffusion of this
change within a population/linguistic community.
20
The actual change is not a soci-
olinguistic phenomenon, but rather the result of cognitive processes that determine the
process of language acquisition, resulting in a grammar in the mind of the individual
speaker that differs from the target grammar. Only under this restricted interpretation
of language change as grammar change we can hope that the inspection of language
change can reveal something about the structure and the workings of the language
faculty/UG. Accordingly, this study focuses on instances of grammar change, that is,
given the subject matter of this work, cases where the categorial status of a pronominal
element changes in the new grammar acquired by the learner.
The issue at hand is also taken up in the work of Anthony Kroch, who tries to
reconcile a formal approach in terms of parametric change with the apparent gradual-
ness of language change. It is a well-known observation that language change usually
proceeds via a (intermediate) stage where old and new (i.e., changed) linguistic forms
co-exist side by side, leading to a degree of variation which is not encountered in sta-
ble linguistic communities. Furthermore, it has been noted that the loss of the old
form and its eventual replacement by the new form is a gradual process which at times
may extend over many centuries.
21
In a series of publications (cf. e.g. Kroch 1989, 1994,
2001), Kroch has developed a formal account of these observations which is based on
the Principles and Parameters framework. Here, the notion of grammar competition
represents the core concept of an integrated theory of language change and variation.
The basic idea of Krochs approach is that parametric change must always proceed via
a stage where the speaker (or, a generation of speakers) of a language X has access to
more than one internalized grammar (sometimes referred to as an instance of internal
diglossia). The grammars in question may differ in a number of parametrical choices,
giving rise to a wider range of linguistic variation. However, blocking effects imposed
by UG (see Kroch 1994 for details; cf. Aronoff 1976 on blocking effects on morpho-
logical doublets) restrict the co-existence of grammars that differ only minimally with
respect to a set of parameter doublets (i.e., co-existing competing values for a sin-
gle parameter), thereby warranting that one grammar will eventually win out over its
competitors, which completes the change in question.
We can therefore conclude that the apparent gradualness of language change does
not impose a challenge on the concept of parametric change. Rather, change between
grammars is necessarily abrupt but may involve a stage where speakers have command
over more than one internalized grammar, giving rise to an unusual degree of lin-
guistic variation, and, over time, the impression that language change is gradual when
one parametrical choice wins out over the other. The next section discusses the phe-
nomenon of grammaticalization, the prime example of an allegedly gradual change,
and shows how this diachronic process can be subsumed under the model of language
change outlined in the present section.
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.18 (1049-1102)
. Grammaticalization
It is an important insight of 19th century linguists such as Bopp (1816), Schlegel
(1818), Humboldt (1822), (1836), Mller (1875), Paul (1880), Gabelentz (1891), or
Brugmann (1916) that grammatical categories like determiners, inections, conjunc-
tions or auxiliaries evolve historically fromformerly (free) substantial lexical categories
such as nouns or verbs. The termgrammaticalization is rst coined in the work of Meil-
let (1912, Levolution des formes grammaticales, republished 1965: 133) to describe
this type of change. A more recent characterization of grammaticalization denes it as
[...] the process whereby lexical items and constructions come in certain linguis-
tic contexts to serve grammatical functions, and, once grammaticalized, continue
to develop new grammatical functions. Thus nouns and verbs may change over
time into grammatical elements such as case markers, sentence connectives, and
auxiliaries. (Hopper & Traugott 1993: i)
Fromthe 1980s on, the study of grammaticalization processes has become a main focus
of descriptive diachronic/typological linguistics, leading to a wealth of new data and
a set of generalizations on the course and dening properties of grammaticalization
(major publications include Heine & Reh 1984; Traugott & Heine 1991; Heine, Claudi,
& Hnemeyer 1991; Hopper & Traugott 1993; Ramat & Hopper 1998; Heine & Kuteva
2002; Wischer & Diewald 2002; Lehmann 2002). Recently, the issue is taken up in
generative work as well, cf. for instance Roberts (1993b), von Fintel (1995), Newmeyer
(1998, 2001), Roberts and Roussou (1999, 2003), Lightfoot (1999), Abraham (2003),
(2004), and van Gelderen (2004a, b).
Traditional approaches stress that grammaticalization is a highly gradual process,
changing slowly but surely the syntactic category of a lexical item along universal his-
torical paths, called grammaticalization clines or paths (cf. Hopper & Traugott 1993;
Heine & Kuteva 2002; Lehmann 2002 for overviews). A typical example is the cline
that characterizes the development of agreement markers (see Chapter 1, example (2),
repeated here for convenience):
(18) independent pronoun weak pronoun clitic pronoun afxal (agglu-
tinative) agreement marker fused agreement marker
Often, different stages of a grammaticalization cline are taken to represent a contin-
uum between, for example, pronouns and agreement markers.
22
Accordingly, it is
often claimed that syntactic categories are not discrete units, but rather fuzzy theo-
retical entities.
Note that these properties conict with standard generative assumptions and
the notion of grammar (or parametric) change introduced in the previous section.
In particular, the gradual character of grammaticalization processes and the appar-
ent existence of universal pathways of change seem to be incompatible with the idea
that grammar change proceeds in the non-deterministic, bumpy fashion envisaged
in the work of David Lightfoot, for example. In what follows, however, we will see
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.19 (1102-1175)
that it is possible to reconcile these apparently contradictory views if certain theoret-
ical assumptions on language change in general and grammaticalization processes in
particular are adopted (cf. Hale 1996; Lightfoot 1999).
This study adheres to the idea that syntactic categories represent discrete units. Ac-
cordingly, the apparent gradual nature of grammaticalization processes is considered
a supercial impression that results from the factors discussed in the previous section,
that is, mixing up the notions of grammar change and the (sociolinguistic) diffusion of
a change and failing to recognize the role of grammar competition in grammar change.
As noted above, the existence of more than one internalized grammar may generate an
unusual degree of variation in the E-language of an individual. If this type of variation
concerns the syntactic category of a given element (e.g., a clitic in grammar A and an
agreement marker in grammar B), then this will create the supercial impression of a
highly gradual change process involving indiscrete or fuzzy category boundaries.
The study of grammaticalization processes has led to the identication of another
set of properties which are characteristic of this type of change such as gradual loss
of phonological and semantic content, a development from free into bound forms, an
increasingly xed position etc. In this book, I maintain that these seemingly distinct
properties or processes can receive a uniform formal explanation on the assumption
that grammaticalization is to be analyzed as the reanalysis of substantial lexical ele-
ments as exponents of (higher) functional categories/heads (basically following von
Fintel 1995; Roberts & Roussou 1999, 2003). This change may also affect functional
elements, which then turn into other functional elements, as in the case of clitics (pre-
sumably realizations of D
0
) that turn into agreement markers (cf. Roberts & Roussou
2003). Below, it is shown that on this assumption, many characteristics of grammat-
icalization processes can be directly attributed to properties of functional categories.
For example, if phonological/prosodic deciency is a universal property of functional
categories, then a given lexical element must undergo a certain degree of phonological
reduction before it can be reanalyzed as the realization of a functional head.
23
In the
following, this is illustrated with a selection of typical properties of grammaticalization
processes (cf. Hopper & Traugott 1993; Lehmann 2002 for detailed overviews).
It is a well-known observation that grammaticalization processes are character-
ized by the gradual loss of phonological and semantic substance, often referred to as
phonological attrition and semantic bleaching, respectively. The reduction of phonolog-
ical substance can be characterized as the loss of segments or of marked phonological
features (cf. the reduction of Latin ille to French le). The process of phonological re-
duction is usually accompanied by a loss of prosodic independence which proceeds
along the following pathway (cf. Hopper & Traugott 1993: 7)
(19) content item > grammatical word > clitic > inectional afx > phoneme >
It has often been observed that functional elements such as complementizers, de-
terminers, or auxiliaries (and, of course, bound elements such as inections) are
prosodically/phonologically decient, since they cannot bear stress and are preferably
monosyllabic/moraic elements in contrast to lexical categories (cf. Kenstowicz 1994
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.20 (1175-1229)
on English; Vogel 1999 on Italian; Roberts & Roussou 2003).
24
Thus, the constant loss
of phonological content is expected on the assumption that grammaticalization is to
be analyzed as the transition from lexical elements to functional categories. It seems
plausible to assume that the absence of stress (via an intermediate clitic stage) and a
reduced segmental make up is a necessary precondition for the reanalysis as a func-
tional head to take place (cf. Roberts & Roussou 2003 for discussion). Moreover, after
the reanalysis is completed, it is predicted that the element is prone to undergo fur-
ther phonological reduction due to the decient phonological nature of functional
categories, accounting for the pathway described by (19).
More or less directly related to the general property of phonological attrition is
the notion of coalescence which refers to the fact that grammaticalization typically in-
volves the development of bound forms from formerly free forms, cf. (19). Again,
this observation can be attributed to the general phonological deciency of functional
categories, which are often realized by afxes or clitic elements which require the pres-
ence of a (lexical) host they can attach to. Note that the existence of free inectional
markers (as in many Creole languages for example) indicates that the development
of bound forms is not a necessary consequence of the reanalysis of lexical material
as functional heads, but rather a tendency, caused by the phonological deciency of
functional categories.
Whereas phonological attrition is a rather straightforward process, the exact
meaning of the notion semantic bleaching cannot be pinned down that easily. In-
tuitively, it designates the erosion of substantial lexical meaning. For example, nouns
evolving into nominalizing afxes (or, in our case, pronouns evolving into agreement
markers) lose their referential (and descriptive) capacity (such as German -heit < Old
High German heit, Gothic haidus person, nature, form, rank), while full verbs evolv-
ing into auxiliaries or tense markers lose their thematic (and predicational) properties
(cf. Old English sceal, full verb subject has to pay an amount of money or has to re-
turn something to somebody > (root) modal > future marker, Lehmann 2002: 114).
Following proposals by von Fintel (1995) and Roberts & Roussou (2003), these seem-
ingly distinct semantic changes can be subsumed under a unied account that is based
on the following two insights: First, the loss of semantic content characteristic of
grammaticalization does not affect the semantic properties of a given lexical item in
a random fashion. Rather, the semantic erosion always concerns a certain subset of
semantic properties: verbs lose their argument structure (or, their thematic proper-
ties), nouns their referential (and descriptive) content, prepositions their capacity to
designate spatial relations etc. This can be subsumed under the notion that gram-
maticalization involves the loss of predicative or descriptive content (see Roberts
& Roussou 2003 for examples and extensive discussion). Second, another set of se-
mantic properties is preserved or added to the formerly substantial lexical category
as a result of the grammaticalization process. This set of semantic properties can be
termed logical content. Examples include the retention of modal content in the devel-
opment of modals/auxiliaries from formerly full verbs (cf. the case of Old English sceal
noted above) or the addition of the feature [deniteness] in the case of determiners
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.21 (1229-1297)
evolving from nouns or numerals. Thus, the semantic changes typical of grammatical-
ization are not adequately described by the very broad term semantic bleaching, since
they show a set of clearly structured properties and, even more important, the reten-
tion or even addition of logical semantic content. According to von Fintel (1995), the
distinctive feature that differentiates logical from predicative or descriptive content is
permutation invariance, which can be characterized as follows:
The intuition is that logicality means being insensitive to specic facts about the
world. For example, the quantier all expresses a purely mathematical relation-
ship between two sets of individuals (the subset relation). Its semantics would not
be affected if we switched a couple of individuals while keeping the cardinality of
the two sets constant. There couldnt be a logical item all blonde because it would
be sensitive to more than numerical relations. (von Fintel 1995: 179)
More generally, logical content is associated with quanticational properties. Inter-
estingly, quanticational properties can be shown to be a characteristic of functional
categories, cf. the following passage taken from Roberts and Roussou (2003: 223):
Many functional elements clearly have logical meanings, in a sense that seems very
close to that dened above. This is clearly true of quanticational elements in DP
(occupying D, Q or Num). It is true also of modal elements, to the extent that
these quantify over possible worlds. It is also true of negation, as this just denes
a complement relation between sets. It may be true of Tense and Aspect, to the
extent that these notions can be construed as quantication over times or events.
Alternatively, Tense may be an ordering predicate (Stowell 1996), another kind
of logical relation (see above). Complementizers, to the extent that they may be
factive or irrealis, are connected to modality.
25
In addition, functional meanings have high semantic types, in contrast to lexical mean-
ings, which have low semantic types. To take an example from von Fintel (1995: 184),
an adjective (of the type <<e,t>,<e,t>>) must assume a more complex semantic type if
it develops into a determiner (<<e,t>,<<e,t>,t>>). Again, these observations suggest
that an analysis of grammaticalization as the reanalysis of lexical elements as func-
tional categories is on the right track. If functional categories are characterized by (and
limited to) a certain kind of semantic content (i.e., logical meanings and high seman-
tic types, cf. von Fintel 1995), then it is expected that the transition from lexical to
functional category requires the loss of those semantic properties that are not com-
patible with the universal make up of functional categories, that is, non-logical (i.e.,
predicative or descriptive) content, and the retention or addition of logical content.
Exactly these types of semantic changes are a characteristic trait of grammaticalization
processes.
26
As von Fintel (1995: 184) puts it:
When a lexical item wants to become a functional item, it needs to become
permutation-invariant, it needs to shed any reference to particular entities, prop-
erties, or situations in the world. The itemwill also have to assume a high semantic
type. Lastly, it will have to conform to whatever semantic universals apply to the
particular functional category it wants to belong to.
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.22 (1297-1345)
Apart from the loss of thematic properties, grammaticalization also results in in-
creasingly restricted selectional properties (i.e., a simplied subcategorization frame),
which typically involves a decrease of structural scope (Lehmann 2002), that is, a
development from syntactic selection towards morphological selection. Traditionally,
this characteristic is referred to as condensation. A good example of this process is the
development of the Romance synthetic future, which started as a main verb (habere)
that took a non-nite clause as its complement, passed through a stage as a future-
indicating auxiliary selecting a VP and nally became an inectional sufx with its
(morphological) subcategorization frame reduced to a verb stem.
Roberts and Roussou (1999) argue that this development can also be attributed
to the universal prole of functional categories, on the assumption that only func-
tional categories can select for lexical categories, whereas lexical categories cannot
select for other lexical categories. In other words, it seems that grammaticalization
always proceeds in an upwards fashion, that is, exponents of hierarchically lower lex-
ical categories are reanalyzed as exponents of higher functional heads. According to
Roberts and Roussou (2003), this is a key property of grammaticalization processes in
general: Successive upwards reanalysis along the functional hierarchy is thus how we
dene grammaticalization paths. (p. 202). The idea that categorial reanalysis always
leads to a new exponent of higher functional head is captured by the following struc-
tural conguration (Roberts &Roussou 2003: 200), where a lower element Y turns into
the realization of a higher functional head X:
(20) XP
... YP
Y
Y = X
... (YP does not have to be the complement of X)
Roberts and Roussou (2003) explain this particular trait of grammaticalization pro-
cesses by assuming that grammaticalization is typically an instance of a Move > Merge
reanalysis, in which an element moved into the vicinity of a functional head F is rean-
alyzed as the phonological realization of F (i.e., a movement dependency is converted
into an instance of Merge).
27
To take a concrete example, the development of verbal
agreement markers is analyzed as a two-step process in which a pronominal element
previously moved to SpecAgrSP is rst reanalyzed as a subject clitic which is merged
as an instantiation of AgrS (Roberts & Roussou 2003: 178):
28
(21) [
AgrSP
DP
i
[
AgrS
V ] [
TP
. . . [
VP
t
i
... [
AgrSP
[
AgrS
D+V ] ...
In a subsequent change, D may undergo further reduction and develop into an agree-
ment afx which is part of V:
(22) [
AgrSP
[
AgrS
D+V
j
]... t
j
...] [
AgrSP
[
AgrS
V
j
]... t
j
... ]
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.23 (1345-1404)
In the resulting structure, the content of AgrS is licensed by moving the inected verb
to this position (in Roberts and Roussous terms, this change involves a transition from
AgrS*
Merge+Move
to AgrS*
Move
).
While I agree with the gist of this particular approach to grammaticalization, there
are still several points where the present work deviates from Roberts and Roussous
specic proposals. For example, I do not share the assumption that subject clitics are
merged as agreement heads, since this assumption implies that inectional heads may
bear a -role (Roberts & Roussou 2003: 183), a consequence which I do not nd par-
ticularly attractive. Instead, I adopt the common idea that subject clitics receive their
-role in the VP (or, rather SpecP) and move up into the inectional domain in order
to be licensed. Furthermore, for reasons that will become clear in Chapter 3, I assume
that agreement morphemes do not head their own projection in the syntax. Moreover,
in Chapters 4 and 5, it is argued that there are cases which contradict the claim that
the grammaticalization of agreement markers always proceeds in an upwards fashion.
More specically, it is shown that in a set of German and Rhaeto-Romance dialects,
new agreement markers develop rst as the realization of an agreement morpheme
conned to the C-domain and spread later to lower verbal positions (i.e., T
0
).
Obligatorication and xation are two other traits of grammaticalization processes
that can be directly related to syntactic properties of functional categories. The term
obligatorication refers to the observationthat lexical material subject to grammatical-
ization becomes more and more obligatory in a certain syntactic context. For example,
in the often documented development from demonstratives to determiners (as in Ger-
manic), the demonstrative is initially restricted to contexts where it adds an indexical
meaning to the noun it modies. In contrast, the resulting determiners are always
obligatory.
29
This follows if functional categories are analyzed as building blocks of
syntax which are universally present (either with phonological content or zero). Thus,
if D
0
has a PF-realization //, then // will show up in all NPs/DPs since D
0
is a
necessary component of a nominal expression (see von Fintel 1995 for a semantic
motivation of the universal presence of D
0
in noun phrases).
The term xation refers to the fact that grammaticalized elements tend to occur
in a xed position. This can be neatly illustrated with the development of agree-
ment markers. Whereas free pronouns can occupy a variety of different positions
(IP-internal, fronted to clause-initial position etc.), clitics are usually conned to cer-
tain structural positions (e.g., the Wackernagel position) or certain hosts (most often
the nite verb), but still may at times switch positions inside the clitic complex. Finally,
agreement markers typically show no variation at all, occupying xed positions/slots
in the verbal inection. Again, this is expected on the assumptions (i) that the or-
der/hierarchical position of functional categories is universally xed (Kayne 1994;
Cinque 1999; Chomsky 2000) and (ii) that the relative order of inectional afxes is
determined by the hierarchical structure of head complexes generated by successive-
cyclic adjunction of (inectional) functional heads (i.e., the Mirror Principle, Baker
1988). In other words, if grammaticalization involves the reanalysis of lexical elements
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.24 (1404-1466)
as exponents of functional heads, it is expected that this change leads to a more and
more xed position of the grammaticalized material.
Another property of grammaticalization which can be related to universal mor-
phological differences between lexical and functional categories is paradigmatization,
the development of an open class lexical element into an element that belongs to a
closed class of functional words or is integrated into a morphological paradigm typ-
ically exhibiting a high degree of formal, functional and semantic homogeneity. In
recent generative approaches to the interface between syntax and morphology, it is
usually assumed that paradigmaticity is associated with functional categories. The rel-
evant difference between lexical and functional categories becomes particularly clear
in Late Insertion models such as Distributed Morphology (cf. Section 2.3 above). In
contrast to substantial lexical heads (or l-morphemes), which are subject to free-choice
spell-out, different Vocabulary items belonging to the same paradigm (note that the
notion of paradigm is only a descriptive artifact here) compete for the insertion into
functional heads (or f -morphemes; see Section 2.3 above for details). Items belonging
to the same paradigm form a natural class due to the fact that they serve to real-
ize an identical subset of inectional features. The item realizing the greatest number
of inectional features is chosen for insertion. On these assumptions, the notion of
paradigm is directly related to the insertion procedure for functional heads which dif-
fers signicantly from the way Vocabulary Insertion proceeds in the case of lexical
heads. Thus, increasing paradigmaticity is an expected outcome of the reanalysis of
lexical as functional categories (cf. Chapter 6 for some discussion).
In brief, it seems that an analysis of grammaticalization as the transition from
lexical material to functional categories is on the right track since it provides a unied
explanation for basic characteristics of this type of change, attributing them to the
universal prole of functional categories. However, the notion transition from lexical
material to functional categories is in fact rather vague. Some clarications are in
order, concerning, for example, whether it is really correct to say that a lexical category
X changes into a functional category Y or whether grammaticalization can create new
functional categories absent in the target grammar.
It is fairly clear that the grammaticalization of a given lexical element as a func-
tional head does not lead to the disappearance of this lexical element. For example,
the development of the Romance synthetic future from inected forms of habere was
not accompanied by the loss of the verb habere. Similarly, the grammaticalization of
the complementizer dass/that from demonstratives in Germanic did not lead to the
loss of the demonstratives. Instead, the lexical sources continue to exist in the present-
day languages. Therefore, the often encountered statement that a lexical category X
becomes a functional head is actually misleading. For instance, in the case of the Ger-
manic declarative complementizer, the lexical itemthat [D
0
] did not turn into a lexical
item that [C
0
]. Rather, a new lexical item that [C
0
] came into existence, with that [D
0
]
still part of the lexicon. This innovation was presumably rst restricted to syntactic
contexts where the relevant evidence was ambiguous, that is, which were compatible
with an analysis either as complementizer or as demonstrative. In all other contexts,
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.25 (1466-1538)
that was still perceived as a demonstrative. Note that these considerations seem to sug-
gest that grammaticalization can in fact create new functional categories. In turn, it is
argued that this is not the case, at least with respect to the core functional categories
C, T, and D.
In current generative approaches to clause structure there is general agreement
upon the existence of a set of core functional categories, which, following Chomsky
(1995, 2000), are usually taken to be C, T, and D (see above). More precisely, it is
assumed that these categories represent the skeleton of clause structure, which is uni-
versally present in all natural languages (cf. Rizzi 1997 and in particular Cinque 1999
for more elaborate arrays of functional categories). Support for this assumption comes
from the consideration that functional categories are needed as a glue to combine
lexical categories in natural languages. Morphosyntactically, this glue is represented
by inections, case-endings and complementizers. Semantically, functional categories
are required to interpret combinations of lexical items with low semantic types: [...]
semantically we need functional meanings as a kind of glue holding together the low-
type meanings of content morphemes. (von Fintel 1995: 183). These considerations
have an important impact on formal theories of grammaticalization (note that this has
already been hinted at above in the discussion of the characteristics of grammatical-
ization processes): if the existence of a set of (core) functional categories is an innate
property, readily given by UG, then grammaticalization must be conceived of as the
(diachronic) lexicalization of a universal inventory of functional categories/heads.
30
If combined with the assumptions on the syntax-morphology interface laid out
in Section 2.3 above, this basic idea can be made more precise. Recall that the mor-
phological component operates post-syntactically, that is, lexical insertion takes place
post-syntactically, realizing syntactic terminal nodes, instead of feeding the syntactic
derivation as traditionally assumed. Then, the term grammaticalization describes a
change where a certain set of phonological features that serves to realize a lexical head
in the input string generated by the target grammar is reanalyzed by the learner as
the spell-out of a (adjacent) functional head. In other words, there is no change in
syntactic category at all (in the sense that category change means a change from,
say, V to T), but at least in the lexicon only a change in the set of phonologi-
cal features associated with a given functional head. In combination with the views
adopted directly above, then, grammaticalization does not create new functional cate-
gories; rather, it provides newlexical realizations for a universal inventory of functional
categories/heads (cf. also Roberts & Roussou 2003 on this aspect). Thus, grammatical-
ization clines describe a historical development where a certain set of phonological
features serves to spell-out different syntactic heads in distinct (diachronically related)
grammars, each time reecting the different properties of the underlying syntactic
terminals.
However, grammaticalization does not fully reduce to the redistribution of phono-
logical matrices. For various reasons, it may lead to at time severe syntactic changes.
For example, syntactic change may result from a reanalysis of the syntactic structure
involved: if a (auxiliary) verb moving to a functional head is reanalyzed as a piece of
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.26 (1538-1602)
inection associated with this functional head, then this reanalysis will lead to the loss
of verb movement in all contexts where the reanalysis takes place (cf. Roberts 1993b on
the synthetic future in Romance; Roberts & Roussou 2003 for a variety of further ex-
amples). In addition, on the assumption that the set of universally present functional
categories includes only C, T, and D, while wider arrays may occur in individual
languages, grammaticalization may enrich the functional inventory via adding pe-
ripheral categories to the set of functional categories present in the grammar (cf. e.g.
Wu 2004 on the grammaticalization of a functional category Aspect in the history of
Chinese). Finally, grammaticalization may lead to the addition of lexical meanings to
a given functional head, in the sense that the meaning of a lexical category is com-
posed with a functional meaning to yield a new, more complex functional meaning.
(von Fintel 1995: 184). For example, the reanalysis of an adjective many as a D-head
creates a new functional head (a quantier) which combines the existential force of the
formerly phonologically empty determiner with the plural meaning of many (cf. von
Fintel 1995: 185):
(23)
DP DP
NP NP
AP N
D D
many
$ $ many
a. b.
Summing up, this section has shown that an analysis of grammaticalization as the
transition from lexical elements into exponents of (formerly present, but phonologi-
cally empty) functional categories proves to be a promising avenue of research. It has
been demonstrated that this basic assumption provides a unied explanation for a
number of surface characteristics of grammaticalization phenomena, attributing the
latter to universal properties of a closed class of functional categories. In addition, we
have seen that the apparently gradual character of grammaticalization processes does
not represent a special property of this type of language change that sets it apart from
other types of changes and requires non-standard assumptions such as fuzzy syntac-
tic categories. Rather, the notions of grammar change and grammar competition are
sufcient to reconcile the gradual nature of grammaticalization with standard gener-
ative assumptions on language change and the prole of syntactic categories. Finally,
making use of a Late Insertion model of grammar, I have argued for an analysis of
grammaticalization in terms of a redistribution of phonological features which never-
theless may affect the syntax of a givenlanguage via adding new(peripheral) functional
categories, enriching the semantic content of existing categories or causing the loss of
movement operations. In Chapters 46 it is shown how the diachronic rise of agree-
ment morphology (and its syntactic effects) can be integrated into this overall picture
of grammaticalization.
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.27 (1602-1626)
. Summary
In this chapter, I have introduced the main theoretical notions that I will assume
throughout this book. In Section 2.2 I provided an introduction to the most recent ver-
sion of minimalist syntax, the probe/goal system developed in Chomsky (2000, 2001a,
2001b), which makes available a new approach to agreement in which the uninter-
pretable -set of T/ (the probe) accesses an agreement controller (the goal) under
closest c-command (via the operation Agree). Agreement can therefore be established
independent of (overt) NP-movement into the specier of the relevant functional
head. Section 2.3 gave an outline of Distributed Morphology and the ensuing model
of the syntax-morphology interface where syntactic terminal nodes are provided with
phonological exponents after the syntactic derivation. Section 2.4 illustrated the gen-
erative perspective on language change, according to which differences between his-
torical stages of a single language are to be analyzed parallel to parametric variation
between different individual languages. Hence, syntactic change is interpreted as an
abrupt change in parameter settings (associated with a closed class of functional cate-
gories) that is triggered during language acquisition. In addition, it was shown that this
model can be reconciled with the apparent gradualness of language change if certain
assumptions on the proper object of linguistic study (grammar instead of language)
and the course of parametric change (involving grammar competition) are adopted.
Finally, in Section 2.5, I argued that grammaticalization has to be analyzed as a histor-
ical process where (formerly phonologically empty) functional categories are provided
with a phonological realization via a reanalysis of exponents previously realizing lex-
ical heads. It was demonstrated that this model provides a unied explanation for a
number of surface characteristics of grammaticalization phenomena, attributing the
latter to universal properties of functional categories. In addition, I gave an outline of
how the historical processes in question can be modeled in a realizational model of
grammar like Distributed Morphology.
In the next chapter, I will develop a formal approach to the realization of verbal
agreement (and complementizer inection) which is based on the theoretical assump-
tions laid out above.
Notes
. If the relevant categorial features are weak, the inectional features of the verb and the argu-
ment are eliminated by covert movement of the relevant formal feature bundles (cf. Chomsky
1995).
. One motivation for this revision are instances of long-distance agreement in Icelandic where
the matrix verb agrees in number with a nominative DP that is contained in an embedded lower
clause, while the matrix subject position is occupied by a quirky dative subject; cf. the following
example taken from Sigursson (1996: 30):
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.28 (1626-1697)
(i) Mr
me.dat
virtust
seem-pl
[ r
they.nom
vinna
work
vel].
well
To me, they seem to work well.
. Similar to uninterpretable -features, structural Case is analyzed as an undifferentiated (or
unvalued) feature the manifestation of which depends on properties of the probe: nite T
assigns nominative, assigns accusative and the T head of control innitives assigns null case.
. Still, a DP with a deleted Case feature is visible to the computation and may block the estab-
lishment of an Agree or Move operation that targets a lower goal c-commanded by the DP in
question, giving rise to so-called defective intervention effects. Defective intervention effects are
only triggered by the head of an A-chain. In contrast, A-movement traces are invisible to the
probe-associate relation (Chomsky 2000: 131).
. Chomsky (2000, 2001a) further claims that unaccusative and passive P do not constitute a
phase either. However, see Legate (2003) for a set of arguments that unaccusative and passive
Ps are in fact phases as well.
. Thus, DM is an anti-lexicalist model, in the sense that no complete lexical elements (fully
specied for semantic, morphosyntactic and phonological features) are inserted into the syn-
tactic derivation (cf. Marantz 1997 for discussion). Instead, the information that is traditionally
assumed to be part of a single lexical entry is divided into separate lexical entries that are part
of distinct subparts of the lexicon: one part of the lexicon contains abstract morphemes, that is,
only bundles of semantic and morphosyntactic features which are subject to early, that is pre-
syntactic insertion (such as e.g. C
[+wh]
, T
[+past]
etc.). Another part contains the set of Vocabulary
items which are inserted post-syntactically and provide phonological realizations of the abstract
morphemes. Finally, there is the so-called Encyclopedia which contains the non-linguistic (id-
iomatic) information associated with lexical items (e.g. dog: four legs, canine, pet, sometimes
bites etc.).
. Some authors make use of the notion abstract morpheme to refer to syntactic terminals in
order to avoid confusion with the traditional usage of the term morpheme as part-of-speech.
Note that throughout this book, the notion morpheme is used in the technical sense introduced
in the main text, that is, as referring to syntactic heads/terminal nodes.
. Most current morphological theories accept the notion that the phonological form of a given
afx should be separated from its morphosyntactic function. Thus, they adopt (in some form
or other) the Separation Hypothesis (Beard 1988, 1995; cf. Stump 1998; Borer 1998 for some dis-
cussion), according to which grammatical operations (which manipulate the constituency and
feature content of lexical items) are discrete fromthe morphological or phonological operations
that control the spell-out of words. Accordingly, it is widely held that afxes or other forms of
morphological marking (such as Umlaut or the modication of tone or accent) are merely the
exponents (Matthews 1991) of abstract morphological features such as [future], [plural], [1st
person]. Note that this distinction follows automatically from the overall structure of DM.
. Note that the Subset Principle alone is not sufcient to determine the winner of the compe-
tition if two Vocabulary items realize the same number of inectional features. For such cases,
two different solutions are proposed in the literature. The winner is either simply stipulated by
an extrinsic rule ordering (Halle & Marantz 1993) or follows from a universal hierarchy of mor-
phosyntactic features (cf. e.g. Noyer 1997; see Harley 1994; Harley & Ritter 2002 for the use of
structured feature geometries instead of simple hierarchies).
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.29 (1697-1764)
. The systematic differences between lexical and functional morphemes can perhaps be em-
ployed to explain certain characteristics of grammaticalization processes which are usually
assumed to involve a transition from lexical to functional categories, that is, from lexical to
functional morphemes in the approach outlined here. See Section 2.5 below for the analysis of
grammaticalization phenomena in a generative setting.
. Note that this exposition is somewhat simplifying. Current work in DM (cf. e.g. Marantz
1997; Embick 1997; Harley & Noyer 1999) usually assumes that traditional category labels such
as N, V, A have no signicance in the grammar. Instead, the category of a given element is
determined by the syntactic conguration in which it appears. More specically, it is assumed
that instead of traditional lexical categories, the lexicon contains category-neutral roots. The
morphosyntactic category (if this is a valid notion at all) of a given root is then determined by
the closest c-commanding functional morpheme. For example, the traditional category noun
corresponds to a root that is locally c-commanded and licensed by a D-head, the category verb
to a root locally c-commanded by (or T) etc. On this view, functional morphemes assume a
category-dening role. Importantly, one and the same root may show up as a verb or a noun
depending on the syntactic context where it is merged. This is then reected by the shape of the
Vocabulary items for roots, where LP can be read as Lexical Phrase:
(i) a. [
DP
D [
LP
___ ]] /dog/
b. [
DP
D [
LP
___ ]] /kt/
etc.
However, as long as it does not affect the argument made in the text I will continue to use the
familiar notion of different lexical categories.
. Another set of phenomena which is frequently analyzed in terms of Morphological Merger
is the positioning of clitics, in particular second position clitics of the Wackernagel type. See for
example Schtze (1994) on second position clitics in Serbo-Croatian.
. According to Matthews (1991: 180) it is important to distinguish between genuine cumula-
tive exponence and two other phenomena that may give rise to similar surface forms. The latter
include fused marking (or so-called portmanteau forms), where regular phonological processes
(sandhi) lead to a fusion of adjacent (simple) exponents similar to French au < le. Another
phenomenon that may give rise to the impression of apparent cumulative exponence is so-called
overlapping exponence: for example, in Latin, voice and subject agreement are realized by a single
formative in 2nd person plural forms: -tis in laudatis you praise and -min in laudamin you
are praised. However, in 3rd person plural forms, separate afxes for voice and agreement can
be discerned: compare laudant they praise with laudantur they are praised (-nt 3pl, -ur pas-
sive) (cf. Stump 1998: 31f.). This contrasts with cases of genuine cumulative exponence, where
the exponents of two or more inectional features coincide completely.
. See Arregi (1999) for an analysis of the person and number inection of Basque and Halle
(1997) for an analysis of the Latin noun declension in terms of Fission.
. Note that impoverishment rules may perhaps also be used to account for a number of ap-
parently syntactic anti-agreement effects. For example, the absence of verbal agreement in the
context of wh-subject extraction, which can be obsrved in a number of Northern Italian dialects
(cf. e.g. Brandi &Cordin 1989), may result from an impoverishment rule that deletes agreement
features in the presence of a wh-feature, leading to the insertion of the default 3sg ending.
. Note that this does not amount to say that there are no changes affecting the syntax of a
given language apart from parametric change. For example, there might be changes that have
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.30 (1764-1828)
to do with the frequency of a given syntactic construction (such as, say, V1 structures). Impor-
tantly for our purposes, however, it is only parametric changes that are of theoretical interest
and signicance, in the sense that only they can reveal something about the structure and the
workings of the language faculty/UG, see Lightfoot (1991, 1999) and Hale (1996) for discussion.
. Note that an approach in terms of default settings is not always compatible with the Subset
Principle. For example, Hyams (1986) claims that the default setting for the pro-drop parameter
is [+pro-drop], which is clearly in conict with the Subset Principle, since [+pro-drop] permits
a set of grammatical sentences that includes the set generated by the setting [pro-drop] (cf.
OGrady 1997 for discussion).
. Concerning the question of when a given cue is attested robustly factors such as unambigu-
ous expression and frequency play a role. In the case of V2, for instance, only those utterances
that can only be analyzed as
SpecCP
[XP] are taken to express the cue in question. Furthermore, it
has been observed that in the Germanic V2 languages, only 30% of main clauses exhibit an
arbitrary XP in clause-initial position, while 70% are subject-initial. Lightfoot (1999) is led
to conclude that these 30% of nonsubjects occupying SpecCP represent the threshold that is
sufcient to count as a robust instantiation of
SpecCP
[XP].
. See Bobaljik (2003) and Haeberli (2004) on the relative signicance of syntactic and mor-
phological cues. Haeberli argues that the impact of morphological or syntactic cues may differ
from parameter to parameter (for similar considerations cf. Anderson 1980). Interestingly, he
claims that mismatches between morphology and syntax (e.g., a grammar with syntactic ev-
idence for verb movement but without rich verbal inection) triggers grammar competition,
that is, the existence of a second internalized grammar where the mismatch is resolved (cf. Kroch
1989). Over time, the latter more harmonic (or economic) variant may win out over the older
grammar.
. According to Hale (1996: 16) the difference between change and diffusion can be character-
ized as follows: In the case of change, there has been imperfect transmission of some feature of
the grammar. The acquirers input sources had features X, Y, and Z and the acquirer constructed
a grammar which had features X, Y and W. The difference (W instead of Z) represents a change
(Z>W). In the case of diffusion, the acquirer had input sources with features X, Y, Z and other
input sources (or another input source) with features A, Y, W and constructed a grammar with
features X, Y, and W. Note that there is no imperfect transmission of the relevant features: the
child had feature W in his or her input sources and constructed a grammar with feature W.
(Similarly for X and Y).
. Apart from the transition from OV to VO order, the rise of do-support and the loss of verb
movement are two of the best studied examples of this kind of long-term change in the history
of English (cf. Ellegrd 1953; Lightfoot 1979, 1991, 1999; Kemenade 1987; Kroch 1989; Roberts
1993a; Pintzuk 1999, among many others).
. Furthermore, it is generally assumed that grammaticalization is a unidirectional process
(cf. Hopper & Traugott 1993; Lehmann 2002 for discussion). Thus, if an element has entered a
grammaticalization cline it must continue along the individual stages of the cline but cannot go
back, from e.g. afx to clitic or pronoun. However, it has been pointed out that this strong claim
is confronted with counterexamples in which a former afx developed into a clitic. For example,
in Estonian, (clitic) question particles and emphatic particles appear to have developed from
former afxes (Campbell 1991). In a similar vein, the possessive clitic s can be shown to have
evolved out of a former genitive sufx in the history of English (Allen 1997).
JB[v.20020404] Prn:2/08/2005; 9:38 F: LA8102.tex / p.31 (1828-1892)
Chapter 2. Theoretical preliminaries ,
i. Note that if grammaticalization reects universal properties of functional heads in the way
sketched above, then the study of this type of diachronic change can reveal important insights
into the universal prole of functional categories, which is a major research interest in current
grammar models. See Roberts and Roussou (2003: Ch. 5.3) for discussion.
i|. Note, moreover, that it is generally assumed that many functional heads (such as T in En-
glish or D in languages without overt determiners such as Tagalog) are phonologically empty
which is the endpoint of phonological deciency.
i,. Note that on these assumptions, the category agreement seemingly does not to qualify as
a functional category in the same sense as D, T or C, since it is apparently not quanticational.
See Chapter 3 for discussion of the phrase-structural status of agreement.
io. Note that von Fintel (1995) assumes that the transition from lexical content morpheme to
functional morpheme usually involves an intermediate stage, that is, a content morpheme that
carries logical meanings and is of high semantic type (examples include adjectives such as mere,
former or propositional attitude verbs like believe or deny).
i. The account of grammaticalization proposed by Roberts and Roussou (2003) is based on
the idea that a functional category F which requires PF realization (labeled F*) can satisfy this
requirement either via Move (F*
Move
) or Merge (F*
Merge
). Grammaticalization processes then
usually involve a (initial) change from F*
Move
> F*
Merge
.
i8. Roberts and Roussou (2003) claim that this kind of change took place in previous stages of
the Northern Italian dialect Veneto, drawing on data from Poletto (1995, 2000). Since this di-
alect already exhibits a paradigm of verbal agreement sufxes, the reanalysis in (25) apparently
leads to multiple realization of a single set of agreement features, possibly an instance of (illicit)
multiple exponence. To avoid this unwelcome result (and for various other reasons), Roberts
and Roussou assume that AgrSP splits into separate person (PersP) and number (NumP) pro-
jections, basically following proposals by Poletto (2000) and Manzini and Savoia (2002). The
pronoun is then reanalyzed as the realization of Pers
0
while the verb occupies Num
0
. Under
these assumptions, the reanalysis is captured by the following structural change (p. 182):
(i) [
PersP
DP
i
[
Pers
V ] [
NumP
. . . [
VP
t
i
... [
PersP
[
Pers
D [
NumP
[
Num
V ] ...
i. Apart from their incompatibility with prenominal genitive possessors in some languages
such as German or English, cf. *the Peters book.
o. Of course, this raises the question of which categories are included in the set of functional
categories determined by UG. I adopt the view that properties of UG determine what kinds of
functional categories may exist in natural languages, that is, there is an innate denition of
what is a possible functional category. Howeves, in contrast to work by, for example, Cinque
(1999), it is assumed that the set of (theoretically possible) functional categories compatible
with this denition does not have an extension in every human language. Instead, only the core
functional categories C, T, , D are taken to be universally present, whereas the acquisition of a
wider array of functional categories is dependent on relevant evidence in the input (for related
views cf. Thrinsson 1996; Bobaljik & Thrinsson 1998; Haeberli 2002; Bobaljik 2003).
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.1 (45-125)
Chapter 3
The structural design of agreement
. Introduction
An investigation of the historical emergence of agreement morphology necessarily
builds on a theory of agreement in general. Therefore, the purpose of this chapter
is to develop a formal account of predicate-argument agreement.
Which are the key issues a theory of agreement must address? Even a very elemen-
tary inspection reveals that agreement differs fundamentally from other inectional
categories like Tense and Aspect. The latter require merely a single (morpho-) syntac-
tic operation which serves to license the relevant inectional features by establishing a
structural relation between a verbal head and a higher functional head. This structural
relation is usually modeled in terms of head movement. The resulting head complex
is then interpreted by the morphological component giving rise to an inected verb.
In contrast, the realization of agreement is a more complex process: in addition to
the combination of verb stem and inectional afx, the feature content of the agree-
ment marker has to be determined via feature matching with the -set of a nominal
argument. We therefore have to acknowledge the existence of two separate structural
relations: a relation between two heads (e.g., verb movement leading to afxation of
an agreement marker to the verb) and feature matching between an agreement head
(the agreement target) and an arguments -set (the agreement controller).
1
Accordingly, a formal model of syntactic agreement must provide an answer to
the following fundamental questions: (i) How does the verb combine with the agree-
ment afx (concerning both syntax and morphology)? (ii) How is feature matching
between a certain argument and the agreement afx established? A third issue which
is intimately related to these questions concerns the phrase-structural representation
of agreement, that is, the question of whether agreement features project a separate
functional agreement phrase (AgrP) in the syntax.
The central proposals of this chapter can be summarized as follows. First, it is ar-
gued that agreement morphemes do not head their own projection in the structure
of the clause. Rather, they attach parasitically to other functional heads the content
of which is interpretable at the interfaces. Second, the feature content of agreement
morphemes is not identied under spec-head agreement. Instead, the relevant struc-
tural relation is assumed to be closest c-command (i.e., the Agree relation, Chomsky
2000). Third, it is claimed that agreement morphemes are provided with phonolog-
ical content post-syntactically (Late Insertion) and that the particular properties of
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.2 (125-183)
the placement of agreement markers follow from the way complex head adjunction
structures are linearized in a autonomous morphological component, Morphological
Structure.
The chapter is organized as follows. Section 3.2 focuses on the phrase-structural
representation of agreement. More specically, it addresses the question of whether
agreement heads project a separate phrasal category in the structure of the clause.
Section 3.2.1 reviews a number of conceptual arguments that bear on this issue. In
Section 3.2.2, I discuss relevant evidence from word order facts. Section 3.2.3 deals
with the morphological realization of agreement, giving particular attention to the po-
sition of agreement markers relative to the verb stem and other inectional markers.
Section 3.3 discusses the structural relation under which predicate-argument agree-
ment is established, focusing on the question of whether an approach that makes use
of the Agree mechanism recently proposed by Chomsky (2000) is to be preferred over
the standard analysis in terms of spec-head agreement. It is argued that decisive ev-
idence in favor of an Agree-type approach comes from long-distance agreement in
the Nakh-Dagestanian language Tsez. Section 3.4 deals with previous approaches to
the linearization of inectional markers, arguing that the order of inectional afxes
cannot be determined by the syntax alone, contra recent proposals by Julien (2002).
Section 3.5 briey summarizes the empirical and conceptual ndings reached so far.
In Section 3.6, I develop the approach to predicate-argument agreement advocated in
this book. Section 3.6.1 argues that agreement heads/morphemes are base-generated
as adjuncts to other functional heads. Section 3.6.2 presents an analysis where feature
matching between a controller and an agreement target is established under closest c-
command (i.e., Agree). Under the assumption that Agr-morphemes may in principle
attach to a variety of functional heads, this analysis predicts that Case and agreement
can be dissociated. It is shown that this prediction is borne out by agreement facts
in Georgian and participle agreement in French. Section 3.6.3 develops a formal ac-
count of the spell-out of inectional markers, focusing on the creation of linear order.
Section 3.6.4 summarizes the basic traits of the analysis developed so far. Section 3.7
argues that we must acknowledge the existence of a further mechanism that may give
rise to agreement phenomena, drawing on evidence from complementizer agreement
in Germanic. Section 3.7.1 provides a critical review of previous analyses of the phe-
nomenon in question, arguing that a purely syntactic account is not tenable. In Section
3.7.2, I develop a new analysis of complementizer agreement that attributes the exis-
tence of inection in the C-domain to the post-syntactic insertion of Agr-morphemes.
Section 3.8 provides a summary of the issues discussed in this chapter.
. The phrase-structural representation of agreement
As already noted in Chapter 2 above, in the pre-minimalist stage of the Principles
and Parameters framework the standard approach to subject-verb agreement posited
a spec-head conguration between the subject and the functional In node which
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.3 (183-218)
Chapter 3. The structural design of agreement
contains the inectional features normally associated with the verb (i.e., tense, mood,
and agreement features) and mediates the subject-predicate relation (Chomsky 1981,
1986b). In later work, it was reasoned that the features/afxes which were previously
assumed to be part of In should head their own projections in the syntax (cf. e.g.
Pollock 1989; Kayne 1989; Belletti 1990; Chomsky 1991), leading to a more articulate
conception of clause structure with separate projections for T(ense) and subject/object
agreement (AgrSP and AgrOP respectively):
(1) CP
C
AgrSP
AgrS
C
AgrS TP
AgrOP T
T
AgrO
AgrO VP
In this approach, feature matching between an arguments -set and the -set of an
agreement morpheme is attributed to a spec-head relation between these two ele-
ments, that is, the argument DP moves to SpecAgrSP/SpecAgrOPto license the feature
content of AgrS/AgrO. In addition, the spec-head relation in question is assumed to re-
sult in Case assignment, in the sense that agreement projections provide the structural
conguration in which the Case features carried by V (accusative) and T (nominative)
are assigned to nominal arguments under spec-head agreement (which necessitates
movement of T and V to the relevant Agr-heads). In line with the morphosyntactic
analyses of the time, it is assumed that the verb stem moves to AgrS/AgrO to pick up
the relevant inectional morphemes (Ouhalla 1991; Chomsky 1991, among others).
Following Baker (1985, 1988), the successive head movement operations that combine
the verb with its inections are mirrored by the order of inectional afxes, with lower
functional morphemes being realized closer to the verb stem than higher ones (the
so-called Mirror Principle, see below). Furthermore, it is sometimes assumed that the
hierarchical order of functional projections may be subject to parametrization (Laka
1990; Ouhalla 1991), since in some languages like German or Turkish tense morphol-
ogy is closer to the verb stem than agreement, while it is the other way around in
languages such as Berber.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.4 (218-254)
However, this approach to agreement phenomena has not gone unchallenged.
The following section reviews a set of arguments that have been put forward against
the existence of separate syntactic agreement projections. These arguments can be
divided into three types: (i) evidence from word order facts; (ii) evidence from the
morphological realization of agreement; (iii) conceptual arguments.
.. Conceptual arguments
Already early on it was pointed out that the assumption of separate agreement pro-
jections raises conceptual problems due to the fact that the nature of Agr differs fun-
damentally from that of other inectional categories. For example, Speas (1991: 411)
notes that typical functional categories such as Tense, Aspect or Negation are not de-
pendent on some other constituent for their interpretation, whereas the content of Agr
is determined in relation to some other constituent with which Agr shares -features.
In other words, Agr lacks an independent meaning and is more accurately character-
ized as a relational instead of a (separate) functional category (Iatridou 1990; Speas
1991; Mitchell 1994; Chung 1998).
This line of reasoning is taken up in more formal terms in Chomsky (1995: 349ff.),
who argues that agreement projections should be eliminated from the clause structure
since they lack features that receive an interpretation at the interface levels PF and LF.
Chomsky points out that agreement projections are present only for theory-internal
reasons, namely to provide the structural conguration in which the feature content
of T, V (Case, non-interpretable -features) is checked against the feature content of
nominal arguments. As a substitute for an AgrP-based theory of Case and agreement
checking, he proposes that the relevant non-interpretable features are added to the
core functional categories T and , which contain sets of features interpretable at
the interfaces. Accordingly, Case and the predicates non-interpretable -features are
checked in an additional specier of these categories.
2
Thus, the removal of AgrPs leads
to a theory where heads may project multiple speciers (see Ura 2000, 2001 for discus-
sion). The following phrase marker illustrates the resulting clause structure, including
the A-/head movement operations that result in Case and agreement checking.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.5 (254-303)
(2) CP
C
TP
T subj.
C
[ [ V[ ]]T]
T
P
t
subj.
obj.
VP t
n
t
V
t
obj.
Apart from these conceptual considerations, there is also empirical evidence indicating
that the idea of separate agreement projections may be awed. The relevant data are
discussed in the following sections.
.. Evidence from word order facts
One of the original arguments for splitting the IP into separate tense and agreement
projections comes from word order facts in French innitives. Pollock (1989) argues
that the diverging placement properties of non-nite auxiliaries and main verbs rela-
tive to sentence negation and certain classes of adverbs should be taken to indicate that
there are two different landing sites for verb movement in the inectional domain. His
argument goes as follows. In innitives, the auxiliaries avoir have and tre be can
appear on either side of the negation pas.
(3) a. Ne pas
neg neg
tre
be
heureux
happy
est
is
une
a
condition
prerequisite
pour
for
crire
write
des romans.
novels
Not to be happy is a prerequisite for writing novels.
b. Ntre pas heureux est une condition pour crire des romans.
(Pollock 1989: 373)
Pollock therefore concludes that non-nite auxiliaries can optionally undergo move-
ment to In/T (which is assumed to be higher than NegP) in French. In contrast,
non-nite main verbs can appear to the left or to the right of VP-adverbs but must
follow the negation pas:
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.6 (303-361)
(4) a. Compltement
completely
perdre
lose
la
the
tte
head
pour
for
les
the
belles
pretty
tudiantes,
students
cest
that is
dangereux!
dangerous
To completely lose ones head over pretty students is dangerous.
b. Perdre compltement la tte pour les belles tudiantes, cest dangereux!
(Pollock 1989: 377f.)
(5) a. Ne pas
neg neg
sembler
seem
heureux
happy
est
is
une
a
condition
prerequisite
pour
for
crire
write
des romans.
novels
Not to seem happy is a prerequisite for writing novels.
b. *Ne sembler pas heureux est une condition pour crire des romans.
(Pollock 1989: 374)
On the assumption that VP-adverbs mark the left edge of the VP, these examples are
taken to indicate that non-nite main verbs can undergo optional movement to an
inectional head that is located between NegP and VP. Pollock identies this head as
Agr, heading a separate AgrP. Moreover, he claims that both non-nite main verbs and
auxiliaries can undergo short verb movement to this position, cf.
(6) tre
be
souvent
often
triste...
sad
To often be sad... (Iatridou 1990: 562)
Hence, according to Pollock the structure of the clause is as follows:
(7) [
CP C
[ C [
TP
[
T
T [
AgrP
[
Agr
Agr [
VP
[
V
V ]]]]]]]]
However, shortly after Pollocks paper was published, Iatridou (1990) presented a set
of French data that challenges the assumption that there is a separate AgrP between
NegP and VP. First, she discusses the pair of examples in (8).
(8) a. Pierre
Pierre
a
has
peine
hardly
vu
seen
Marie.
Marie
b. Pierre a vu peine Marie. (Iatridou 1990: 563)
(8a) follows straightforwardly from Pollocks analysis: the nite auxiliary moves to T,
while the non-nite main verb stays behind in situ (to the right of the VP-adverb).
However, Iatridou points out that (8b) creates a problem for Pollocks account. Note
that both the nite auxiliary and the non-nite main verb precede the VP-adverb. Ac-
cording to Pollocks proposals, this conguration implies that both verbs have moved
over the adverb, which occupies a xed position. However, if verb movement has to
obey the Head Movement Constraint (Travis 1984), then the nite auxiliary must
move through Agr before it reaches T. But then, there is no head position left where the
participle can move to. Pollock tries to handle the problematic example by assuming
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.7 (361-435)
that the structure of sentences like (8b) is actually as shown in (9), where the participle
moves to the head of a separate AgrP that is located between VP
aux
and VP
main
. To
ensure that participle movement lands in a position to the left of peine, he assumes
that the adverb is adjoined to VP
main
.
(9) ... [
AgrP
[
Agr
Agr [
VPaux
[
Vaux
V
aux
[
AgrP
[
Agr
Agr[
VPmain
peine
[
VPmain
[
Vmain
V
main
]]]]]]]]]
Iatridou notes that this proposal partially weakens Pollocks argument for the existence
of AgrP, since it is now possible to base-generate auxiliaries in the correct surface order
in examples such as (6) (which Pollock analyzes as resulting from short movement
of the auxiliary). Even more problematic is the following data set, which includes not
one, but two VP-adverbs:
(10) a. Souvent
frequently
mal
badly
faire
make
ses
ones
devoirs,
homework
cest
that is
stupide.
stupid
To frequently do ones homework badly is stupid.
b. Faire souvent mal ses devoirs...
c. Souvent faire mal ses devoirs... (Iatridou 1990: 567)
The crucial example here is (10c). Whereas (10a) and (10b) can be analyzed as the
result of non-movement and short verb movement, respectively, it is not possible to
account for (10c) under Pollocks assumptions: if both adverbs are adjoined to the VP,
there is simply no head position available where the non-nite main verb can move
to. Iatridou concludes that Pollocks arguments in favor of a separate AgrP do not go
through and that the French data in questionis better handled by a more exible place-
ment of adverbs. In addition, she suggests an alternative analysis of agreement which is
based on the idea that predicate-argument agreement does not involve a separate pro-
jecting Agr-node, but merely reects a certain structural (spec-head) relation between
two elements.
.. The morphological realization of agreement
Some of the original conceptual motivations for splitting the Innode into a number
of functional heads concerned the interface between syntax and morphology. First,
the fact that the In node contained different sets of morphosyntactic features (such
as Tense, Agr, etc.) was deemed conceptually unattractive. Second, in line with the
Mirror Principle proposed by Baker (1985), it was generally assumed that syntax and
morphology are basically isomorphic, that is, ideally, each separate inectional afx
should correspond to a separate functional head.
3
(11) The Mirror Principle
Morphological derivations must directly reect syntactic derivations (and
vice versa).
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.8 (435-499)
Thus, if a given functional head is apparently associated with more than a single set of
inectional features, this is attributed to head movement operations which yield head
adjunction structures (in line with the Head Movement Constraint, Travis 1984). One
important consequence of the Mirror Principle is that one should be able to ascertain
the order of embedding of functional categories simply by inspecting the order of in-
ectional markers on the verb stem in a given language. While this line of research has
proven to be very fruitful with respect to functional categories such as Tense, Aspect
and Negation (cf. e.g. Laka 1990; Ouhalla 1990, 1991; Baker 1996; Zanuttini 1997,
and most recently Julien 2002), the evidence coming from the placement of agree-
ment markers is less straightforward and often rather problematic for the hypothesis
that there is a separate functional agreement projection (cf. Laka 1990; Speas 1991;
Mitchell 1994; Spencer 1997; Julien 2002). This section discusses a set of morphosyn-
tactic properties of agreement which indicate that there is a fundamental difference
between agreement and other inectional categories such as tense, aspect, or negation,
for which it is generally assumed that they head their own projection in the syntax.
The rst difference concerns the distribution of inectional markers. Functional
categories like tense, aspect, negation, or complementizers usually have a unique ex-
pression in any given clause, that is, they generally occur (i) only once and (ii) in
a xed position.
4
The following facts demonstrate that agreement may behave com-
pletely differently. First, we nd (redundant) multiple realizations of one and the same
agreement marker on various verbal elements. Second, agreement markers may ap-
pear in different positions dependent on the (morpho-) syntactic context. In addition,
agreement may be signaled by a set of discontinuous markers scattered over the clause.
Finally, agreement may be realized by a combination of these options.
Multiple agreement on different (verbal) heads is found in languages such as
Swahili where subject agreement (=agreement in noun class) is repeated on every
element of the verbal complex:
Swahili
(12) wa-toto
cl2-children
wa-li-kuwa
cl2-past-be
wa-me-ki-soma
cl2-perf-cl7-read
ki-tabu
cl7-book
The children had read the book. (Krifka 1995: 1416)
Other languages with multiple subject agreement include Sudanic languages like Mbay
(Central Sudanic, AgrS on verb and tense marker/aux) or Lango (Nilotic, AgrS on verb
and negation):
Mbay
(13) m- a
1sg-fut
m-nd 6g o
1sg-buy
I will buy. (Keegan 1997: 70)
Lango
(14) n
I
-p
1sg-neg
-wt
1sg-go.perf
kmpl
Kampala
I didnt go to Kampala. (Noonan 1992: 142)
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.9 (499-586)
Another possibility is traveling agreement, where the agreement markers attach to
different verbal elements dependent on the syntactic context. For example, in the Mon-
Khmer language Sengoi, subject agreement is expressed on the tense marker/auxiliary
in the future tense, but on the verb in the past tense:
5
Sengoi
(15) a. Ke
3sg
ki-ha
3sg-fut
muit.
enter
She will enter.
b. Guru
teacher
ajeh
that
ya
past
ki-lei.
3sg-come
The teacher came. (Means et al. 1986)
A combination of multiple agreement and agreement scattering is found in languages
such as Serbo-Croat, where the auxiliary inects for person and number, while the
main verb shows agreement for gender and number of the subject (note that the
pronoun may be dropped):
(16) on-i
3-masc.pl
su
be.3.pl
dol-i
came-masc.pl
They came. (Corbett 2000: 244)
An evenmore complex systemis exhibited by the Finnic language Livonian, where each
element of the verbal complex agrees with a specic subset of the subjects -features
(negation: person; aspectual auxiliaries: person and number; participle: number):
Table 1. Multiple agreement in Livonian (Mitchell 1994: 114).
neg-agrs(person) be-agrs(person & number) give-part-agrs(number)
1sg -b o- and-n-
2sg -d o- and-n-
3sg -b o- and-n-
1pl -b o-m and-n-D
2pl -d o-tt and-n-D
3pl -b o-tt and-n-D
Under the assumption that the presence of agreement morphology should be taken
as evidence for a separate AgrP (in line with the Mirror Principle), these facts require
the existence of multiple agreement phrases to account for subject agreement alone.
As noted above, this contrasts signicantly with the behavior of functional heads such
as T or Asp, which typically occupy a single xed position; the latter neither spread to
other positions nor are they realized in different positions dependent on the syntactic
context. Alternatively, one might resort to the assumption that some of the different
realizations of agreement are to be attributed to agreement morphemes/features that
are generated outside of AgrP. Yet, the latter solution undermines the whole idea that
agreement morphology corresponds to functional agreement nodes that head their
own projection (cf. Mitchell 1994: 114).
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.10 (586-786)
Further support for the claim that agreement does not head a separate projec-
tion in the syntax comes from the fact that cross-linguistically, the ordering proper-
ties of agreement markers differ signicantly from those of other inectional mark-
ers such as tense or aspect. The following discussion is based on empirical ndings
of Julien (2002). Based on an extensive typological study covering 530 languages,
Julien discusses the relation between word order and the realization and position
of tense and aspect markers. She shows that these inectional categories show only
a very limited amount of cross-linguistic variation with respect to their relative or-
dering. More specically, she claims that the following generalizations hold across
languages (p. 235):
(17) Attested sequences of inectional markers
a. (S) T-Asp-V (O)
b. (S) T-V-Asp (O)
c. (S) (O) V-Asp-T (O)
d. S Asp-V O T
(18) Non-attested sequences
a. *(S) Asp-T-V (O)
b. *(S) Asp-V-T (O)
c. *(S) V-T-Asp (O)
Thus, it appears that aspect markers are always closer to the verb stemthan tense mark-
ers (if both followor precede the verb stem). Julien argues that this observation directly
follows from the Mirror Principle if it is assumed that there is a universal hierarchy
of functional projections (cf. e.g. Cinque 1999) in which T is obligatorily generated
above Asp. Interestingly, Julien shows that with respect to their placement properties,
agreement markers are much less well-behaved than tense and aspect markers. Table 2
presents Juliens ndings concerning the relative position of verb stem, tense (T) and
subject agreement marker (AgrS) in her language sample.
6
Even a cursory look at Table 2 reveals that there is a vast amount of variation
concerning the position and realization of agreement markers across languages. For
example, we can see that all six theoretically possible linearizations of V, T and AgrS
are attested.
7
The overall picture gets even more complex if object agreement marking
is taken into consideration as well, cf. Table 3.
8
A closer look at Table 3 reveals that of the 24 theoretically possible orderings (ab-
stracting away from the complications created by portmanteau forms), only four are
not attested in Juliens sample: *AgrO-T-AgrS-V, *T-AgrO-V-AgrS, *AgrO-T-V-AgrS,
and *AgrS-V-AgrO-T.
9
Thus, it appears that there is a signicant difference between
functional categories such as T and Asp on the one hand and Agr on the other. While
the relative order of tense and aspect markers is rather xed (reecting the fact that T
and Asp universally occur in a xed hierarchical relation), apparently almost anything
goes with respect to agreement markers. If this is again taken to reect the hierarchical
position of Agr, then it follows either (i) that the hierarchical position of agreement
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.11 (786-807)
Table 2. Relative order of verb stem, tense marker, and AgrS marker, by word order (see
Table 5.2 of Julien 2002: 249).
Morpheme Word order
order V-initial V-medial V-nal Uncertain Total
1. AgrS/T V 2 2
2. AgrS/T+V 1 4 3 1 7
3. V+AgrS/T 3 4 17 2 26
4. V+T+AgrS 4 17 42 8 64
5. V+AgrS+T 3 10 3 16
6. V+AgrS T 1 2 3
7. AgrS+T+V 4 7 5 14
8. AgrS T V 6 6
9. AgrS+T V 5 1 5
10. AgrS+V+T 4 29 28 7 57
11. AgrS+V T 1 5 5
12. T+V+AgrS 4 2 2 1 9
13. T V+AgrS 1 1
14. T+AgrS+V 5 5 3 1 9
15. T+AgrS V 1 2 1 3
16. T AgrS+V 3 3
projections is subject to parametrization across languages (in contrast to T or Asp),
basically following proposals by Ouhalla (1991), or (ii) that the phrase-structural
properties of Agr differ fundamentally from those of other functional categories such
as T or Asp. Taken together with the other arguments put forward in this section, it
appears that the second option in fact represents the more accurate choice. I am there-
fore led to conclude that Agr-nodes do not head their own projection in the syntax, in
contrast to substantial functional categories such as C, T or Asp, which are projecting
syntactic heads, characterized by a cross-linguistically uniform order of embedding
(this conclusion is also reached by Julien 2002: 6). Rather, Agr-morphemes may attach
parasitically to other substantial functional categories. This explains the multitude of
serialization patterns if it is further assumed that languages may differ with respect to
the functional head the Agr-morpheme attaches to. A radical implementation of this
insight is that agreement heads are actually not part of the syntax at all, as claimed
by Anderson (1992), Marantz (1992), and Halle and Marantz (1993), for example. In-
stead, these authors propose that agreement morphemes are added to the structure
post-syntactically at Morphological Structure. According to this view, only substan-
tial functional categories such as T, Neg or Asp are represented in the syntax, while
agreement is purely morphological in nature, merely reecting structural relations es-
tablished in the syntactic derivation (see Section 3.6 below for some discussion and a
more moderate proposal according to which at least canonical instances of agreement
are represented in the syntax, but parasitic on other functional heads).
10
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.12 (807-807)
Table 3. Relative order of verb root, tense marker, AgrS marker, and AgrOmarker, by word
order (see Table 5.3, Julien 2002: 253).
Morpheme order Word order
V-initial V-medial V-nal Uncertain Total
1. AgrS/O-V-T 8 11 2 19
2. T-V-AgrS/O 3 3
3. V-T-AgrS/O 1 4 3 7
4. V-AgrS/O-T 1 1 2
5. T-AgrS-AgrO-V 1 2 1 1 3
6. AgrS-T-AgrO-V 1 1 2 4
7. AgrS-AgrO-T-V 1 1 1
8. AgrS-AgrO-V-T 1 3 7 1 11
9. T-AgrO-AgrS-V 1 1 1 2
10. AgrO-AgrS-T-V 2 2
11. AgrO-AgrS-V-T 1 6 1 8
12. T-V-AgrS-AgrO 1 1 2
13. V-T-AgrS-AgrO 1 4 6 9
14. V-AgrS-T-AgrO 1 1 2
15. V-AgrS-AgrO-T 1 1
16. T-V-AgrO-AgrS 1 2 3
17. V-T-AgrO-AgrS 2 6 1 9
18. V-AgrO-T-AgrS 5 1 6
19. V-AgrO-AgrS-T 1 2 1 4
20. V-AgrO-T/AgrS 1 1
21. T-AgrS-V-AgrO 2 2 1 1 4
22. AgrS/T-V-AgrO 2 2 2
23. AgrS-T-V-AgrO 2 2 3
24. AgrS-V-T-AgrO 3 3 2 6
25. AgrO-V-T-AgrS 7 7
26. AgrO-V-AgrS-T 4 4
27. AgrO-V-T/AgrS 1 1 2
This view on agreement markers is further supported by the fact that agreement
markers are preferably bound elements (cf. Chapter 1). (19) lists the sequences that are
theoretically possible, but not attested in Juliens sample (focusing on AgrS only):
(19) a. *V AgrS/T
b. *V T AgrS
c. *V T+AgrS
d. *V+T AgrS
e. *V AgrS T
f. *V AgrS+T
g. *AgrS T+V
h. *AgrS V T
i. *AgrS V+T
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.13 (807-869)
j. *T V AgrS
k. *T+V AgrS
l. *T AgrS V
Note that the vast majority of the non-attested patterns (9 out of 12) involve a free
AgrS marker. In fact, there are only six genera in Juliens sample that exhibit free AgrS
markers. All the relevant languages are of the type SVO (mostly Creoles), with free
subject agreement and tense markers that both precede the verb (i.e., AgrS T V):
(20) Generalization: The distribution of free AgrS markers
A free AgrS marker is only possible in an SVOlanguage in which the sequence
of inectional markers is AgrS T V.
On the assumption that Agr heads its own projection, this nding is rather surpris-
ing. Rather, we would expect Agr to occur as frequently as a non-bound element as
other substantial functional categories like tense auxiliaries, aspect markers or com-
plementizers. In contrast, if agreement does not head its own projection but attaches
parasitically to other functional heads, it is predicted that Agr is preferably realized
as an afx that attaches either to its functional host or to a lexical head moved to the
functional head complex in question (a similar point is made with respect to T+Agr
portmanteau forms in Speas 1991).
11
. The structural relation involved in feature matching
The second ingredient of the agreement relation concerns the structural conguration
in which feature matching between an argument and a predicate can be established.
The following proposals are found in the literature: (i) the spec-head conguration
(Koopman 1992, 1996, 2003; Sportiche 1998; Kayne 1989, 1994; Chomsky 1991, 1993,
1995; Watanabe 1996; Belletti 2001; Julien 2002); (ii) government (Polinsky &Potsdam
2001); (iii) local c-command (i.e., the Agree relation, Chomsky 2000; Carstens 2003;
Bejar 2003); (iv) a combination of some of these proposals (cf. e.g. Chomsky 1986b;
Chung 1998; Bobaljik & Wurmbrand 2004). This section focuses on the approaches
that are prevalent in current syntactic theorizing, that is, the spec-head relation and
an analysis of agreement in terms of the Agree relation proposed in Chomsky (2000,
2001a, 2001b). It is shown that facts from so-called long-distance agreement (hence-
forth LDA) favor an Agree-type analysis over the traditional approach in terms of
spec-head agreement.
It is commonly assumed that the agreement relation is strictly local, in the sense
that the controller and the target must be clause-mates. In an analysis based on the
spec-head conguration, this condition follows automatically from the standard as-
sumptions (i) that agreement involves A-speciers only (similar to case assignment)
and (ii) that A-movement is typically clause-bounded, that is, A-movement may not
cross a CP boundary.
12
Note that this kind of locality holds for both subject and object
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.14 (869-904)
agreement. In both cases, feature matching between an argument and the predicate
is licensed by moving the argument to the closest A-specier (following Chomsky
1995, SpecTP for subject agreement and SpecP for object agreement), which cre-
ates a checking conguration between the interpretable -set of the argument and
the corresponding non-interpretable features of the predicate. In contrast, the locality
conditions imposed by the operation Agree are somewhat different, due to the fact that
the agreement domain extends to the left edge of a lower (strong) phase. Hence, ma-
terial at the left edge of an embedded CP is accessible for the establishment of (object)
agreement as well.
13
Thus, the two approaches to agreement give rise to a set of slightly
different agreement domains, which are illustrated in (21):
(21) Agreement domains
a. Spec-head
Subject agreement
[ C [ spec T [ spec [ V ... [ spec C [ ... ]]]]]]
Object agreement
CP

TP P VP CP TP
b. Agree
Subject agreement
[ C [ spec T [ spec [ V ... [ spec C [ ... ]]]]]]
Object agreement
CP

TP P VP CP TP
In standard cases of subject or object agreement, the empirical coverage of both ap-
proaches is identical (at least roughly, cf. Koopman 2003 for some discussion). To
decide which analysis is to be preferred over the other, we must therefore look for
cases where the subtle difference between (21a) and (21b) with respect to locality has
an effect on the realization of agreement. This section shows that the locality con-
ditions imposed by the spec-head relation are violated in instances of long-distance
agreement (LDA) where a matrix verb can apparently access a controller that is part
of an embedded clause.
14
Therefore, the existence of LDA constitutes a problem for
an analysis of agreement that makes use of the spec-head relation (cf. e.g. Polinsky &
Potsdam 2001; Bruening 2001; Bobaljik & Wurmbrand 2004). However, before we can
decide whether the extension of the agreement domain made available by the work-
ings of the operation Agree is sufcient to capture the problematic data, we must take a
closer look at the phenomenon in question (the following discussion is based on work
by Polinsky & Potsdam 2001; Polinsky 2003; Bobaljik & Wurmbrand 2004).
Genuine LDA involves a matrix verb (the target) that selects for a clausal com-
plement and agrees with an argument (the controller) contained in the embedded
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.15 (904-965)
clause.
15
The following denition of the phenomenon in question is given in Polinsky
(2003: 282):
(22) Long-distance agreement
Agreement between a controller and a target that are in distinct clauses.
Languages which reportedly exhibit this form of long-distance agreement are Tsez
(Polinsky & Potsdam 2001) and the Algonquian languages Pasamaquoddy (Bruen-
ing 2001) and Innu-aimn (Branigan & MacKenzie 2002). The following discussion
will focus on the relevant facts in the Nakh-Dagestanian language Tsez, which is spo-
ken in the Caucasus region and exhibits basic SOV order (cf. Polinsky & Comrie 1999;
Polinsky & Potsdam 2001; Comrie 2003; Polinsky 2003).
Verbal agreement in Tsez is organized in an ergative-absolutive fashion, that is,
the verb agrees in gender and number with its absolutive argument (i.e., either the
direct object of a transitive or the subject of an intransitive verb):
16
In the following
examples, Roman numerals indicate gender distinctions.
(23) a. kid
girl.abs.II
y-iki-x
II-go-pres
The girl goes.
b. u- a
boy-erg
bikori
snake.abs.III
b-exu-r-si.
III-die-caus-pstwit
The boy killed the snake. (Comrie 2003: 323)
However, in cases where the matrix verb selects a clausal argument, it may agree either
with the whole embedded clause or with the absolutive argument of the embedded
clause. This state of affairs is illustrated by the following pair of examples:
(24) a. eni-r
mother-dat
[u- a
boy-erg
magalu
bread.abs.III
b- ac-ru-i]
III-eat-pstprt-nmlz-abs
r-iy-xo.
IV-know-pres
The mother knows that the boy ate the bread.
b. eni-r
mother-dat
[u- a
boy-erg
magalu
bread.abs.III
b- ac-ru-i]
b-iy-xo.
III-know-pres
The mother knows that the boy ate the bread. (Polinsky & Potsdam
2001: 584)
In (24a), the matrix verb riyxo knows agrees in gender with its complement clause,
as can be seen from the fact that the verb is inected for default gender class IV agree-
ment, which is realized by the prex r-. However, in (24b), the matrix verb carries the
agreement prex b-, which serves to signal agreement with a gender class III absolutive
argument. In the example at hand, the only available controller of this particular agree-
ment marker is the absolutive-marked direct object of the embedded clause, magalu
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.16 (965-1044)
bread, which also controls gender class III agreement on the embedded verb. In other
words, (24b) exemplies a case of LDA where the controller is apparently separated
from its target by a clause-boundary.
Polinsky and Potsdam show that LDA is only possible when the controller is the
primary topic of the embedded clause (in other words, the presence of LDA signals
that the controller serves as the topic of the embedded clause). Thus, LDA is not pos-
sible if the embedded absolutive argument carries the clitic focus marker -kin, cf. (25),
or if there is another topic in the embedded clause. Example (26a) shows that the tem-
poral adverb Au yesterday does not block LDA when it occurs in situ. However, LDA
with the embedded absolutive DP is ruled out when the temporal adverb is topicalized
within the embedded clause, as in (26b):
17
(25) *eni-r
mother-dat
[tek-kin
book.II.abs-foc
y-igu
II-good
y a-ru-i]
be-pstprt-nmlz
y-iy-xo.
II-know-pres
The mother knows that the BOOK is good. (Polinsky & Potsdam 2001: 611)
(26) a. eni-r
mother-dat
[u- a
boy-erg
Au
yesterday
magalu
bread.III.abs
b- acrul-i]
III-eat-pstprt-nmlz
r/b-iy-xo.
IV/III-know-pres
The mother knows that the boy ate bread yesterday.
b. eni-r
mother-dat
[Au
yesterday
u- a
boy-erg
magalu
bread.III.abs
b- acrul-i]
III-eat-pstprt-nmlz
r/*b-iy-xo.
IV/III-know-pres
The mother knows that the boy ate bread yesterday.
Polinsky and Potsdam deduce from these facts that the embedded absolutive must
undergo overt or covert movement into the left periphery of the embedded clause to
be accessible for LDA.
Crucially, it can be shown that neither restructuring nor overt raising into the
matrix clause is involved. The following pair of examples demonstrates that the whole
embedded clause including the controller ka>at letter, which triggers long-distance
agreement on the matrix verb, can undergo fronting to clause-initial position:
(27) a. [kidb a
girl
ka>at
letter.II.abs
t atrui]
read
y-iyx
II-knows
uir.
boy
The boy knows that the girl has read the letter.
b. uir
boy
y-iyx
II-knows
[kidb a
girl
ka>at
letter.II.abs
t atrui].
read
The boy knows that the girl has read the letter. (Polinsky 2003: 302)
Therefore, it is quite clear that the controller has not undergone overt raising into
the matrix clause. Furthermore, in contrast to restructuring contexts, each clause
constitutes a separate domain for the scope of negation:
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.17 (1044-1105)
(28) a. [kidb a
girl
ka>at
letter.II.abs
tetr-nci-ru-i]
read-neg-pstprt-nmlz
y-iyx
II-knows
uir.
boy
The boy knows that the girl has not read the letter.
b. [kidb a
girl
ka>at
letter.II.abs
t atrui]
read
y-iy-x- anu
II-knows-pres-neg
uir.
boy
The boy does not know that the girl has read the letter.
(Polinsky 2003: 303)
These facts indicate that the construction in which long-distance agreement is ob-
served is a genuine biclausal structure. Moreover, the controller remains in the em-
bedded clause, at least in the overt syntax (see Polinsky & Potsdam 2001 for more
arguments involving scrambling, binding, and scope interactions). This seems to sug-
gest that the facts of LDA in Tsez cannot be accounted for if agreement is modeled in
terms of an overt spec-head conguration. Yet, we must rule out the possibility that
the controller undergoes covert raising into the matrix clause, thereby meeting the
normal locality condition on verbal agreement at LF. Polinsky and Potsdam (2001)
provide several arguments against this analysis and other conceivable alternatives that
situate the controller in the matrix clause. One such argument concerns the interpre-
tation of embedded quantied expressions, the scopal properties of which are taken to
indicate that the lower absolutive DP remains in the embedded clause throughout the
entire syntactic derivation (Polinsky & Potsdam 2001: 618f.). In Tsez, quantiers that
are clause-mates may give rise to scope ambiguities. In the following mono-clausal ex-
amples, either the subject or the object may take wide scope over the other (X > Y
reads as X has scope over Y):
18
(29) a. ibaw
every
>
w
ay- a
dog-erg
sis
one
ketu
cat-abs
han-si.
bite-pstwit
Every dog bit a cat. (dog > cat, cat > dog)
b. sida
one-obl
>
w
ay- a
dog-erg
ibaw
every
ketu
cat-abs
han-si.
bite-pstwit
A dog bit every cat. (dog > cat, cat > dog)
No such ambiguities arise when one of the two quantied DPs is contained in an em-
bedded clause. In other words, quantiers may not take scope out of their own clause
(Polinsky & Potsdam 2001: 618):
(30) a. sis
one
uciteler
teacher
[ibaw
every
ui
boy-I.abs
-ikixosi-i]
I-go-nmlz
r-iyxo.
IV-know
Some teacher knows that every boy is going.
b. ok: Some teacher is such that he knows that every boy is going. (teacher
> boy)
c. *Every boy is such that some teacher knows that he is going. (*boy >
teacher)
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.18 (1105-1157)
Now, if it is assumed that the embedded absolutive DP ends up in the matrix clause at
some point in the derivation (e.g., by covert movement) in order to be able to control
agreement on the matrix verb, then we should expect that this DP may take wide scope
over another quantied element of the matrix clause. Crucially, the expectation is not
borne out by the facts:
(31) a. sis
one
uciteler
teacher
[ibaw
every
ui
boy-I.abs
-ikixosi-i]
I-go-nmlz
-iyxo.
I-know
Some teacher knows that every boy is going.
b. ok: Some teacher is such that he knows that every boy is going. (teacher
> boy)
c. *Every boy is such that some teacher knows that he is going. (*boy >
teacher)
In (31a), the matrix verb shows gender class agreement with the embedded absolutive
DP ibaw ui every boy (class I). Nevertheless, the sentence has the same interpre-
tation as (30a), that is, ibaw ui every boy cannot take scope over the quantied
element sis uciteler one teacher in the matrix clause. Therefore, we can conclude
that ibaw ui every boy must be located in the embedded clause at the level of rep-
resentation where quantier scope is computed. More important for our purposes,
long-distance agreement cannot be licensed by raising of the embedded absolutive DP
in the covert syntax. These ndings can be summarized as follows:
(32) Long-distance agreement in Tsez
a. The structure is biclausal and the target is contained in the higher clause.
b. The controller, which carries absolutive case, remains in the embedded
clause at all levels of representation.
Polinsky and Potsdam (2001) conclude that the form of LDA found in Tsez cannot
be accounted for by an analysis in terms of spec-head agreement. Instead, they argue
for an analysis where the agreement relation closely resembles head government, al-
lowing the matrix verb to enter into an agreement relationship with a controller that
undergoes covert A-movement into the left periphery of the embedded clause.
Still, Koopman (2003) claims that these ndings can be reconciled with the idea
that agreement universally involves the spec-head conguration if an analysis in terms
of clausal pied-piping is adopted. Koopman assumes that an element that occupies
the left periphery of an embedded clause can pied-pipe the whole clause to establish
feature checking with a higher head under spec-head agreement (for related proposals
cf. Koopman & Szabolcsi 2000; Mller 2004; cf. Simpson & Bhattacharya 2003 for a
similar analysis of wh-in situ in Bangla). On this assumption, LDA between a matrix
verb and an embedded DP is analyzed as resulting from moving the whole embedded
clause containing the controller into the specier of an AgrP of the matrix clause:
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.19 (1157-1210)
(33) AgrP ...
CP
DP Agr ... C
Agr
C IP V t
CP
...
Koopman claims that in this conguration, feature checking between Agr and the -
set of the embedded absolutive DP is possible if it is assumed that spec-head agreement
between the DP and C leads to the presence of -features in C
0
/CP via feature perco-
lation. In turn, the resulting -feature set on C
0
/CP participates in feature checking
with Agr under spec-head agreement.
Note that this approach involves a number of non-standard assumptions. For
example, analyses that make use of clausal pied-piping are generally concerned with
A-dependencies,
19
for which it is commonly assumed that movement serves to satisfy
properties of both the moved element (scope) and the attracting head (clause type/wh-
feature). Thus, percolation of, for example, wh-features can be said to be motivated by
the need to satisfy scopal properties of an in-situ wh-phrase. This contrasts with the
percolation of A-related features involved in Koopmans analysis of LDA in Tsez, where
no property of the controller is involved: the -set of the absolutive DP is interpretable
and therefore need not be checked. Moreover, the absolutive DP has already partic-
ipated in Case and agreement checking in the embedded clause, so there is no need
for Case checking with the matrix verb either. Thus, there is no motivation for feature
percolation to occur. In addition, it appears that a spec-head relation that licenses the
presence of -features on a given head (which Koopman calls percolation) requires
that the relevant specier is an A-position (cf. e.g. Shlonsky 1994 on complementizer
agreement in Germanic). Again, this is not the case in the LDA examples at hand. To
conclude, it is very doubtful that the grammar makes available an operation like A-
pied-piping. Instead, it seems that the licensing of A-related features such as Case and
agreement may not lead to pied-piping of more material than the minimal phrase that
contains the attracted features.
Apart from these conceptual problems, a spec-head analysis of LDA can also be
shown to suffer from empirical shortcomings. Koopman explicitly assumes that agree-
ment always results from overt movement processes, that is, from a spec-head cong-
uration created in the overt syntax (which may be obscured by subsequent movement
operations). On this assumption, however, a spec-head analysis of LDA cases such as
(24b), repeated here as (34), faces a problem.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.20 (1210-1265)
(34) eni-r
mother-dat
[u- a
boy-erg
magalu
bread.abs.III
b- ac-ru-i]
b-iy-xo.
III-know-pres
In (34), the leftmost element of the embedded clause is not the absolutive DP mag-
alu, but rather the ergative DP u- a. Therefore, Koopmans analysis in terms of feature
percolation and spec-head agreement leads us to expect that the matrix verb can only
agree with the embedded ergative DP, since it occupies a position to the left of the ab-
solutive DP (which should lead to the presence of the ergative DPs -features in C if
the relevant position is SpecCP). Still, it is the absolutive DP which triggers agreement
on the matrix verb and not the ergative one. Since Koopman assumes that agreement
relations must be established in the overt syntax, this fact cannot be attributed to agree-
ment checking at LF involving covert movement of the absolutive DP into the left
periphery of the embedded clause. Moreover, it is not possible that the order ergative-
absolutive found in the embedded clause is created by further movement operations
after agreement checking has taken place: under standard assumptions about cyclicity,
no movement operations can take place in the embedded clause after this clause it-
self has undergone movement (in the case at hand, to SpecAgrP of the matrix clause).
Therefore, it seems that an analysis in terms of spec-head agreement is not capable of
explaining the form of LDA exhibited by Tsez. However, before we can turn to a con-
clusion, it remains to be shown that an account in terms of Agree can handle the data
set in question more adequately. Recall that Polinsky and Potsdam (2001) propose an
analysis where the agreement relation closely resembles head government, allowing the
matrix verb to enter into an agreement relationship with a controller located in the left
periphery of the embedded clause. Of course, this is already reminiscent of the Agree
mechanism and the concept of locality devised in Chomsky (2000, 2001a), where only
material located at the left edge of a phase (a stage in the syntactic derivation repre-
sented by e.g. a nite CP) is accessible to syntactic operations triggered by functional
heads higher up in the phrase marker (the Phase Impenetrability Condition). Thus, in
principle an analysis in terms of Agree seems to be well-equipped to handle LDA in
Tsez. Furthermore, as expected in the framework of Chomsky (2000), movement to
the left edge of a phase may feed operations triggered by higher functional heads (in
the case at hand, agreement), and LDA is subject to intervention effects created by
other elements topicalized in the embedded clause. Still, there are some open ques-
tions, the most obvious having to do with the fact that the operation that moves the
controller of LDA into the left periphery of the embedded clause apparently can take
place in the covert syntax (see above). This implies that agreement is established by
covert operations as well. This has an important consequence for the overall model of
grammar. On the assumption that the content of agreement morphemes must be val-
ued in the syntax before they can be subject to Vocabulary Insertion (Late Insertion),
the evidence from Tsez suggests that the morphological component must operate after
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.21 (1265-1347)
all syntactic movement operations have taken place, including covert movement (cf.
Bobaljik 2002; Bobaljik & Wurmbrand 2004). Thus, we arrive at a model of grammar
in which the point of Spell-out represents the interface to both Morphological Struc-
ture and LF, with apparent covert movement resulting from assigning a phonological
matrix to the base position of a chain (sometimes called Spell-out at LF, cf. Bobaljik
1995; Groat & ONeill 1996).
Summing up, this section has shown that the phenomenon of genuine LDA
which can be observed in languages like Tsez constitutes a serious problemfor the tra-
ditional approach to agreement in terms of a spec-head conguration. In addition, we
have seen that the slight extension of the agreement domain following from the work-
ings of the Agree mechanism enables a matrix verb to agree with an element in the left
periphery of its complement clause. This provides us with a rather elegant explana-
tion of the problematic data, lending further support to the claim that the structural
relation which agreement phenomena are based on is rather closest c-command (i.e.,
Agree) than the spec-head conguration.
20
. Syntactic approaches to word formation
This section deals with the relation between syntactic structure and morpheme order
(i.e., the order of phonological exponents). As already noted above, the most inuen-
tial approach to this topic (in the Principles and Parameters framework) is based on
the Mirror Principle (Baker 1985, 1988), repeated here for convenience:
(35) The Mirror Principle
Morphological derivations must directly reect syntactic derivations (and
vice versa).
On the assumption that the verb combines with its inectional afxes (which are
generated in functional heads) via successive head movement operations that must
respect the Head Movement Constraint, the Mirror Principle ensures that the order
of inectional markers on the verb stem reects the hierarchical properties of a head
adjunction structure derived in the syntax. For example, if the verb moves rst to Asp
and then to T, the morphological exponent of Asp must be closer to the verb stem than
the exponent of T (if both afxes are on the same side of the stem).
However, the assumption that the formation of complex words (here: the inected
verb) takes place in the syntactic component (via head movement) faces the problem
that verbs can also carry inection in languages that apparently lack the operation
of verb movement. Here, the principal example is English, for which it is standardly
assumed that the nite verb does not leave the VP. If adverbs like often are taken to
mark the left edge of the VP, thenthe following contrast indicates that the verb does not
move into the inectional domain in the overt syntax (cf. Emonds 1978; Pollock 1989):
(36) a. Floyd often eats bananas.
b. *Floyd eats often bananas.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.22 (1347-1386)
Still, the verb clearly carries inection, either the 3sg.pres.indicative /-z/ or the past
tense marker /-d/. To obviate this problem, Chomsky (1993) proposes a lexicalist alter-
native which does not any longer make use of the assumption that the verb undergoes
head movement to pick up inectional markers in higher functional heads. Instead,
inectional markers are added to the verb prior to lexical insertion. Subsequently, the
verb enters the syntactic derivation fully inected and undergoes syntactic movement
to functional heads to check whether it carries the appropriate inectional features.
The problem posed by examples such as (36) is now accounted for by the assumption
that the relevant checking operations may be postponed to the covert syntactic com-
ponent. Thus, it is assumed that in English, the verb undergoes invisible LF-movement
to check off tense and agreement features in higher functional heads. The overt/covert
distinction is implemented by the assumption that the V-feature that triggers verb
movement may be either strong or weak (recall that only strong features require overt
movement prior to Spell-out). This serves to account for the empirical facts, but gives
up the idea that there is a correlation between syntactic operations and the order of
inectional afxes on the verb: if checking of inectional features may take place af-
ter Spell-out, it is rather clear that the relevant movement operations cannot have
any effect on the shape of the inected verb. Thus, a checking model as envisaged by
Chomsky (1993) makes no predictions at all with respect to verbal morphology, since
word formation takes place prior to syntax and inectional features may be checked
by either overt or covert movement. There is no causal relation between the absence
of head movement and the absence of inectional morphology: a verb spelled-out in
situ may still carry inection.
21
Recently, Julien (2002) has proposed a radical alternative to previous accounts of
word formation, claiming that the order of inectional markers and the sufx/prex
distinction are completely determined by the syntactic component (which might be
called a Radical Mirror Principle; similar ideas are already present in Kayne 1994). Fol-
lowing Kayne (1994), she assumes that hierarchical relations created in the syntactic
component exhaustively determine word order. More specically, PF (or MS) maps
asymmetric c-command relations into linear order (the Linear Correspondence Ax-
iom, LCA). In addition, Julien claims that this approach carries over to word-internal
ordering properties of stems and inectional markers. Moreover, properties such as
being a prex or a sufx are taken to merely reect the hierarchical position of a given
terminal node in the syntactic component. For example, it is claimed that head move-
ment is always left-adjunction, leading uniformly to a realization of the target head
as a sufx (since by assumption, the target is asymmetrically c-commanded by the
adjoined head):
(37) T
V T
uniformly maps into V+T
Apart from head movement, Julien recognizes two other congurations that can give
rise to afxation of inectional material. Crucially, in these congurations, the el-
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.23 (1386-1450)
ements forming a complex word do not constitute a single X
0
-unit. Instead, Julien
assumes that separate syntactic heads may be interpreted by MS/PF as one grammat-
ical word if they are immediately adjacent to each other (adopting a Late Insertion
model). Accordingly, a conguration where the head of XP and the head of a phrase
YP moved to SpecXP are adjacent can lead to sufxation of the head of XP to the head
of YP (cf. Kayne 1994 for this proposal concerning word formation in strict SOV lan-
guages). This is illustrated in (38), where SpecTP is lled by a VP (remnant). On the
assumption that the complements of V have undergone further licensing movement
operations, V and T are adjacent. In this conguration, then, V+T can be realized as a
single word via Late Insertion.
(38) TP
VP
... T t
VP
V
T
Julien claims that structures generated by so-called roll-up movement, where VP
moves into SpecP, P into SpecAspP, AspP into SpecTP etc., are the the principal and
perhaps the only means of deriving complex words (p. 101) in SOV languages. Note
that by assumption, both head movement and the conguration in (38) give rise to suf-
xing. What about prexes? Here, Julien assumes that two heads which do not form a
X
0
-complex can still be realized as one grammatical word if they are linearly adjacent
due to evacuation movement of intervening material. Thus, in a conguration such as
(39), T may be realized as a prex on V (moved to ) if nothing intervenes between
these heads. As a consequence, prexation is always the outcome of non-movement.
22
(39) TP
T
T P
VP V +
...
Julien assumes that the structural congurations in (37)(39) may give rise to either
free or bound inectional markers. In other words, the syntax has nothing to say about
whether markers are free or bound.
23
One of the most interesting aspects of Juliens proposal is that it links the order of
inectional markers to general syntactic properties (word order) of a given language.
For example, it has often been noted that SOV languages show predominantly sufxal
inection (cf. Sapir 1921; Greenberg 1963; Bybee 1985; Bybee et al. 1990; Siewierska
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.24 (1450-1475)
& Bakker 1996, and many others). In Juliens framework, this correlation is accounted
for by assuming that strict SOVlanguages such as Japanese always have the structure in
(38) and lack the option of head movement (cf. Kayne 1994). In other words, it appears
that Juliens account makes clear predictions with respect to which afx sequences
A,B,C should be possible if a language has a certain set of syntactic properties X,Y,Z
(a language with verb movement must show sufxes, prexation of functional mate-
rial implies the absence of verb movement etc.). Combined with the assumption that
functional categories such as T or Asp occur in a universal order of embedding (e.g.,
T is always generated above Asp), this leads to a rigid, syntactically constrained the-
ory of how X
0
-complexes are linearized, which is at rst sight conceptually attractive.
However, similar to the related approach to word order (cf. Kayne 1994 and subse-
quent work), such an account requires quite an amount of heavy syntactic machinery
(e.g., massive roll-up and remnant movement). In particular, the shift of explanatory
burden from morphology to syntax often leads to highly complex syntactic structures
and sometimes contra-intuitive derivations simply to account for a certain sequence of
inectional markers. For example, as already mentioned in Note 22, English must be
analyzed as having overt verb movement into the inectional domain in order to ac-
count for its sufxal verbal inection, despite massive counterevidence (cf. Emonds
1978; Pollock 1989; Chomsky 1991, 1993; Halle & Marantz 1993; Roberts 1993a;
Rohrbacher 1994; Bobaljik 1995, among others). Another general problem arises in
connection with the assumption that the majority of SOV orders is generated by suc-
cessive roll-up movement of complements into the specier of their selecting heads.
Since the speciers of the inectional domain are required to host constituents subject
to roll-up movement (i.e., P moves to SpecAspP, AspP to SpecTP etc.), there are no
positions left for argument licensing (Julien follows Kayne 1994 in assuming that every
head projects only a single specier). Accordingly, Case and agreement checking must
take place VP-internally, contra standard assumptions.
24
Thus it appears that a purely syntactic approach to morpheme ordering gives rise
to a number of serious conceptual problems. Moreover, it can be shown that the spe-
cic analysis in question is also empirically inadequate, predicting the existence of
some sequences that are not existent and excluding others that are attested across lan-
guages. Concerning the former point, recall that Julien (p. 235) notes that the following
sequences of inectional markers are not attested across languages:
(40) Non-attested sequences
a. *(S) Asp-T-V (O)
b. *(S) Asp-V-T (O)
c. *(S) V-T-Asp (O)
However, note that under Juliens assumptions, even if T is universally generated above
Asp, at least the order Asp-T-V can be derived rather easily by rst moving the Asp-
head to T, giving rise to sufxation (i.e., Asp+T). Subsequently, this complex may
attach to the verb (in situ) if nothing intervenes, leading to the sequence Asp-T-V:
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.25 (1475-1535)
(41) TP
T
T AspP
V ...
VP T t
Asp
Asp
... V
Thus, Juliens theory is not as restrictive as it appears at rst sight, due to the diversity
of (i) possible syntactic derivations and (ii) congurations feeding word formation
her theory allows for. Below, it is shown that a much more conservative theory of
morpheme ordering is equally successful in restricting the amount of possible orders.
Another empirical problem has to do with the realization of agreement. As noted
above, Julien assumes that agreement features are added to other functional heads
(Julien specically mentions Fin, T and heads lower than T such as Asp or Voice).
However, she is not very specic about the exact representation of agreement features.
That is, it remains unclear whether agreement features are merely added to the feature
set of a given functional head or if they are base-generated as separate heads adjoined
to other functional heads (as proposed in this book, see Section 3.6.1 below). Im-
portantly, both options can be shown to be rather unwieldy within the theoretical
approach developed by Julien.
Assume that Agr-features/morphemes are added to a functional head which is as-
sociated with morphological content as well, such as T. If the Agr-features are merely
added to the feature content of T, the question arises of what determines the linear or-
der of the phonological exponents of Agr and T. It is fairly clear that the linear relation
in question cannot be determined by the syntax, since the set of features contained
in T (e.g., [+past] and [3sg]) does not correspond to a syntactic structure involving
c-command relations. Accordingly, the position of Agr (i.e., its realization as pre-
x/sufx) must be determined by the morphological component alone, presumably
via lexical properties of the phonological exponent of Agr. This, however, would un-
dermine the whole approach devised by Julien (cf. p. 255, for example: It is a main
thesis of the present work, however, that the position of an element in the surface or-
der is never directly encoded in its lexical entry.). However, if the order of T and Agr
added to T must follow from a structural conguration, then the only option left in
Juliens framework is left-adjunction of Agr to T, that is, the structure in (42), where
Agr must be realized to the left of T (or any other potential host).
(42) T
T Agr
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.26 (1535-1592)
Yet, this can be shown to give rise to serious complications for Juliens approach. More
specically, all instances where AgrS occurs to the right of T cannot be analyzed any
longer as resulting from the presence of Agr-features on T. This order is found in 79
genera in Juliens language sample, including the most frequent sequence V+T+AgrS,
which is found in the Germanic, Celtic and Romance families, for example. In the
case of Germanic and Romance, then, this requires that (i) AgrS is added to a func-
tional head above T (under Juliens assumptions presumably Fin) and (ii) that the
nite verb undergoes overt movement to this head (roll-up movement is conned to
genuine OV languages). However, at least in embedded clauses of the Germanic VO-
languages (apart from Yiddish and Icelandic), it is rather obvious that the verb does
not move that far up in the clause, but rather stays behind to the right of negation, cf.
the following examples from Danish (which is representative for the other Mainland
Scandinavian languages including Faroese):
25
(43) a. *Jeg
I
spurgte
asked
hvorfor
why
Peter
Peter
havde
had
ikke
not
lst
read
den.
it
b. Jeg
I
spurgte
asked
hvorfor
why
Peter
Peter
ikke
not
havde
had
lst
read
den.
it
I asked why Peter didnt read it. (Vikner 1995: 145)
Thus, even with the proviso that AgrS features may be located in a number of different
functional heads, Juliens system is not capable of deriving the ordering patterns found
in Germanic in a satisfactory way. This state of affairs results from the fact that Juliens
idea that afx order is completely determined by the syntax is too strict. The failure
to account for the linearization of Agr-morphemes casts serious doubts on the overall
enterprise of providing a purely syntactic analysis of the linear order of inectional
markers. Note, however, that at least some of the problems of Juliens system might be
remedied if Agr-features added to, say, T might be spelled out to the left or to the right
of T. Then, however, the claim that afx order is exhaustively determined by the syntax
must be dropped.
In Section 3.6.3 below, I will outline an alternative theory of the linear order of
inectional markers which is based on similar assumptions as Juliens approach, but
is more exible, adopting the more conservative assumption that linear order is only
partially determined by the syntactic component, which leaves open certain points of
variation which are resolved in the post-syntactic morphological component.
. Interim summary: The design of agreement
So far, the discussion in this chapter has provided us with the following insights into
the structural design of agreement. First, it has beenshown that the properties of agree-
ment differ signicantly fromthose of functional categories such as T, Neg or Asp. This
has been taken to suggest that agreement features/morphemes do not head their own
projection in the syntax. Rather, they are parasitic on other functional heads (cf. Ia-
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.27 (1592-1660)
tridou 1990; Speas 1991; Marantz 1992; Halle & Marantz 1993; van Gelderen 1993;
Mitchell 1994; Chomsky 1995, 2000; Halle 1997; Julien 2002). Second, Section 3.3
has provided further arguments (apart from those originally put forward in Chom-
sky 2000; cf. Chapter 2 above) that the structural conguration in which the feature
content of Agr-morphemes is licensed/valued is closest c-command (i.e., the Agree re-
lation) rather than spec-head agreement. Third, we have seen that it is presumably not
possible to establish a one-to-one relation between the output of the syntactic compo-
nent and the ordering of inectional markers on the verb stem. Rather, it appears that
the syntax constrains the set of possible afx sequences, but that lexical information
(such as the prex/sufx distinction) matters as well when the output of the syntactic
component is interpreted at Morphological Structure.
. Toward a realizational theory of agreement
In this section, I will give an outline of a theory of agreement which is based on the
insights reached above. The approach presented here is framed in a realizational model
of grammar, where morphology operates after syntax, interpreting the output of the
syntactic component (Distributed Morphology; cf. Chapter 2 above). The section is
organized as follows. First, Section 3.6.1 addresses the question of how agreement
should be represented in the grammar if we abandon the idea that agreement heads
project phrasal categories. Section 3.6.2 deals with the structural relation under which
feature matching between a controller and an agreement morpheme is achieved. Fi-
nally, Section 3.6.3 discusses aspects of word formation in the model of agreement
envisaged in this book, focusing on how syntactic structures are mapped into linear
sequences of phonological exponents.
.. The structural representation of agreement morphemes
If agreement features/morphemes do not head their own projection in the syntax,
the question arises of how they are to be represented in the structure instead. As
noted above, two major lines of reasoning can be discerned in the literature: agree-
ment features/morphemes are added to other functional heads either before or after
the syntactic component. According to the latter line of thinking, agreement is a
purely morphological phenomenon and agreement heads are completely absent from
the syntactic component. Instead, they are added post-syntactically at Morphologi-
cal Structure as adjuncts to substantial functional categories such as T, Neg or Asp
which are represented in the syntax, while agreement merely reects structural rela-
tions established in the syntactic derivation (Marantz 1992; Halle & Marantz 1993;
Halle 1997).
26
This mechanism is sometimes referred to as the insertion of dissociated
morphemes (Embick 1997). In Distributed Morphology, this morphological operation
is often used to account for case and agreement phenomena. For example, Marantz
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.28 (1660-1688)
(1992), Halle and Marantz (1993), and Halle (1997) analyze subject-verb agreement
in terms of the post-syntactic adjunction of an Agr-morpheme to T. However, I hes-
itate to adopt this mechanism for canonical instances of agreement for the following
reasons. First, an analysis where agreement is generally attributed to the insertion of
dissociated morphemes requires that MS has powerful syntax-like mechanisms at its
disposal, which are necessary for detecting the correct agreement controller, to value
Agr-morphemes via copy operations etc. That is, this analysis seems to establish a syn-
tax after the real syntax, which is conceptually unattractive. Second, as shown above
in Section 3.3, long-distance agreement phenomena are sensitive to intricate syntac-
tic restrictions such as intervention effects created by topicalized constituents. It is
rather doubtful that these restrictions can be handled by morphological mechanisms
alone in a satisfactory way. Still, it seems that at least for some instances of multiple
agreement, where agreement is realized in more than one place in a given clause, an
approach in terms of a post-syntactic insertion of agreement morphemes might be on
the right track (see Section 3.7.2 below for an analysis of complementizer agreement
in Germanic which makes use of this mechanism). I therefore adopt a somewhat less
radical alternative according to which (syntactic) agreement features/morphemes are
present in the syntax, but parasitic on other functional heads (Iatridou 1990; Speas
1991; van Gelderen 1993; Mitchell 1994; Chomsky 1995, 2000; Julien 2002). More pre-
cisely, I assume that agreement morphemes are part of the numeration, but do not
head their own projection in the syntax. Instead, agreement morphemes are merged
with other substantial functional heads before the latter are combined with a phrasal
complement.
27
Canonical subject-verb agreement is then the result of an agreement
morpheme attached to T (henceforth Agr-on-T). That is, the following X
0
-complex
consisting of an Agr-morpheme adjoined to T is merged with P in the syntax:
(44) T
Agr T
As far as the syntactic derivation is concerned, the assumption that discrete, albeit
non-projecting Agr-morphemes are adjoined to other functional heads does not dif-
fer substantially from the idea that agreement features are merely added to the feature
content of other functional heads (Chomsky 1995, 2000). At rst sight, the latter ap-
proach seems to be preferable on conceptual grounds, since it completely removes
Agr-heads (and adjunction structures such as (44)) from the syntactic derivation. Still,
I choose to represent agreement via separate syntactic heads/morphemes, since this fa-
cilitates a more straightforward morphological processing of agreement morphemes,
in particular with respect to the creation of linear order (see Section 3.6.3).
28
By assumption, any of the core functional categories C, T, or D (which may
represent a wider array of functional categories) is in principle capable of hosting
agreement morphemes (cf. Julien 2002 for some discussion and examples where sub-
ject agreement is analyzed as resulting from an Agr-morpheme attached to Fin, Asp or
Voice). The actual presence of an agreement morpheme is then dependent on whether
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.29 (1688-1724)
the learner can detect the presence of an Agr-morpheme on a certain functional head
on the basis of the evidence available to him/her.
.. Feature matching under closest c-command
In Section 3.3 above, it has already been pointed out that the structural relation
underlying predicate-argument agreement should be modeled in terms of closest c-
command between an agreement morpheme and an appropriate goal, that is, the
operation Agree proposed by Chomsky (2000). In this section, the workings of Agree
in establishing feature matching in canonical instances of subject and object agreement
are briey illustrated. Recall that by assumption, subject-verb agreement involves the
presence of an agreement morpheme adjoined to T, while object agreement results
from an agreement morpheme added to . If syntactic trees are built up in a bottom-
up fashion, object agreement is checked after has been merged with its complement
VP which contains the object (note that the following phrase markers abstract away
from head movement).
(45)
P
DP
subj.
VP
DP
obj.
V Agr
Agree
In the structure (45), the head complex [
Agr []] may enter into an Agree relation

with the feature set of the object DP (under closest c-command), which leads to fea-
ture matching between the interpretable -set of the object and the content of the
Agr-morpheme, giving rise to object agreement (and Case licensing). Eventually, the
object DP moves to a specier of if the latter carries an EPP feature which triggers that
operation. If Merge preempts Move, it is expected that the subject is introduced before
object movement takes place (Chomsky 2000).
29
Subsequently, the resulting struc-
ture is merged with T, which may host an agreement morpheme as well. Again, the
head complex [
T
Agr [T]] initiates an Agree operation which targets the closest active
DP with an appropriate feature content, leading to subject agreement and nominative
licensing:
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.30 (1724-1762)
(46)
P
DP
obj.
DP
subj.
VP
t
obj.
V Agr
T
TP
spec
T
T Agr
Agree
Note that the intervening object DP does not block the Agree relation between the head
complex in T and the subject DP due to the fact that elements of the same minimal do-
main are equidistant to a higher probe (cf. e.g. Chomsky 2000: 122, 130). This analysis
reects the intuition that agreement should not be implemented via a projecting func-
tional head (see Section 3.2 above). Rather, agreement is more accurately characterized
as a relational category (Iatridou 1990; Speas 1991; Mitchell 1994; Chung 1998) the
content of which is licensed by entering into an Agree relation with the interpretable
-set of a DP argument.
The theory of agreement outlined so far has certain consequences for the rela-
tion between Case and agreement. Recall that Chomsky (2000, 2001a) follows the
standard assumption that Case and agreement licensing is correlated. In other words,
object agreement is associated with the Agree relation initiated by , which also serves
to license accusative Case, while subject agreement is associated with T in combina-
tion with nominative assignment/licensing. However, on the assumption that Agr-
morphemes can in principle attach to a variety of functional heads apart from T or
(see above), it should be possible for Case and agreement to be dissociated from
each other (cf. Pesetsky & Torrego 2004 for related conclusions). In other words, it is
expected that there are languages where Case is assigned by T and (or, rather by T
and the head complex +V if it is assumed that the capability to assign accusative is
associated with the lexical category V), while the agreement relation is mediated by
other functional heads that host Agr-morphemes. In such languages, Case assignment
should be independent of the realization of agreement. A rst indication that some-
thing along these lines is on the right track comes from long-distance agreement in
Tsez, repeated here for convenience:
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.31 (1762-1837)
Tsez
(47) a. eni-r
mother-dat
[u- a
boy-erg
magalu
bread.abs.III
b- ac-ru-i]
r-iy-xo.
IV-know-pres
b. eni-r
mother-dat
[u- a
boy-erg
magalu
bread.abs.III
b- ac-ru-i]
b-iy-xo.
III-know-pres
The mother knows that the boy ate the bread. (Polinsky & Potsdam
2001: 584)
As noted above, the matrix verb know may choose to agree either with the whole
complement clause, as in (47a), or with an absolutive DP that receives an interpre-
tation as the topic of the embedded clause (and is therefore associated with the left
edge of the embedded CP phase), cf. (47b). Note that in the latter case, the agreement
controller magalu bread has already undergone Case (absolutive) and agreement
checking in the embedded clause. Thus, it appears that a DP with a checked Case
feature may still be accessible for a further Agree operation that serves to value the
uninterpretable -features of a predicate located in a higher phase, giving rise to a
form of multiple agreement where a single argument repeatedly triggers agreement
on different predicates. This state of affairs requires a modication of the assumption
that an uninterpretable -set can only be valued by a DP with an activating unchecked
structural Case feature. For example, one might pursue the idea that agreement check-
ing is completely independent of Case checking (contra Chomsky 2000). Somewhat
less radical, we may assume that the nal disappearance of an activating Case feature
that is marked for deletion may be postponed to a later stage of the derivation (cf. Pe-
setsky & Torrego 2001; Carstens 2003). For the time being, I adopt the latter proposal.
For the sake of concreteness, I assume that in Tsez, an uninterpretable Case feature that
is marked for deletion is nally deleted at the point where the relevant DP is sent to
Spell-out. On the assumption that the left edge of a phase is spelled-out together with
the complement of the next higher phase head, a Case-marked DP that occupies the
left edge of an embedded CP phase remains accessible to further syntactic operations
triggered by a probe in the next higher phase (i.e., the matrix P).
30
Further support for the claim that there is no direct relation between Case check-
ing and agreement relations comes from Georgian (for details cf. Harris 1981; Hewitt
1996; see King 1994 for a similar conclusion with respect to Georgian). Georgian ex-
hibits a split ergative system: in the present tense, future and other so-called series I
tenses, Georgian exhibits nominative case on the subject and dative on the object.
31
In
contrast, in the so-called series II tenses such as the aorist (also referred to as sim-
ple past) or the subjunctive, the subject carries ergative marking, while the (direct)
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.32 (1837-1891)
object is marked with nominative (or absolutive) case (examples taken from Comrie
1978: 351f.):
32
Georgian Present
(48) a. Student-i
student-nom
midi-s.
go.pres.3sg
The student goes.
b. Student-i
student-nom
ceril-s
letter-dat
cer-s.
write.pres.3sg
The student writes the letter.
Georgian Aorist
(49) a. Student-i
student-nom
mivid-a.
go.aor.3sg
The student went.
b. Student-ma
student-erg
ceril-i
letter-nom
da-cer-a.
prev-write.aor.3sg
The student wrote the letter.
Interestingly, the agreement pattern exhibited by the verb is not affected by the case
marking changes in question. Agreement always works in a nom-acc fashion, that is,
the ergative subject of a verb in the aorist (series II) is marked on the verb in the
same way as the nominative subject of a verb in the present tense (series I). This is
illustrated by the following examples where the verb agrees with the 1st person subject,
independent of whether the latter is marked by ergative or nominative case (takenfrom
Bejar 2003: 180).
33
(50) me
I.nom
tseril-s
letter-dat
v-cer.
1-write-pres
I am writing the/a letter.
(51) me
I.erg
cign-eb-i
book-pl-nom
da-v-cer-e.
prev-1-write-indic.aor
I wrote the books.
These facts suggest that there is no correlation between the realization of Case and
the realization of agreement in Georgian (cf. Marantz 1992 for a similar conclusion,
but a different analysis in which both case and agreement morphemes are added post-
syntactically). Note that this is expected within the analysis of agreement proposed
above where Case and agreement licensing may be associated with different functional
heads (and where a deletion-marked Case feature may remain accessible to further
operations under certain conditions). I will not attempt to sketch a possible anal-
ysis of the intricate agreement facts of Georgian, a topic the complexity of which
justies a monograph in itself (see Bejar 2003 for a recent analysis in terms of the
probe/goal system).
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.33 (1891-1954)
A similar point can be made with respect to participle agreement in French. Kayne
(1989) shows that participle agreement in French is dependent on overt syntactic
movement. The following examples illustrate that the participle is inected for the
-features of the object only if the latter undergoes overt movement to the left of the
participle:
(52) a. Marie
Marie
a
has
fait/*-e
made/-fem
la
the
robe.
dress
(no movement)
Marie has made the dress.
b. La
the
robe
dress
a
has
t
been
fait-e.
made-fem
(NP-movement)
The dress has been made.
c. Marie
Marie
la
it-has
fait-e.
made-fem.
(clitic movement)
Marie has made it.
d. Combien de tables
how-many of-the tables
Paul
Paul
a
has
repeint-es?
repainted-fem.pl
(wh-movement)
How many tables has Paul repainted?
In Kaynes (1989) analysis, participle agreement results from a spec-head relation be-
tween the object and the participle in the overt syntax. This relation is established as a
by-product of independently triggered movement operations such as wh-movement,
NP-movement or clitic-movement. In other words, the object passes through the rel-
evant specier on its way further up in the tree, triggering participle agreement. In
contrast, objects that remain to the right of the participle fail to trigger agreement,
since they do not enter into a spec-head relation with the participle.
As pointed out by Koopman (2003), it seems that the gist of this elegant analysis
cannot be captured by an approach that makes use of the Agree mechanism. Due to
the fact that the object is always locally c-commanded by the participle, one would
rather expect that participle agreement is always possible, independent of movement:
A post participle object in French certainly occurs in the same phase as the participle,
and is locally c-commanded by the participle. It is therefore mysterious why the Probe
of the participle must fail in this context. (Koopman 2003: 6).
In the following, however, it is shown that an analysis in terms of Agree is capable
of capturing the correlation between overt object movement and participle agreement
if it is assumed that accusative Case and object agreement licensing are associated with
different functional heads in French, in line with the approach to agreement advocated
in this book. Note that in the case of French, this appears to be a rather natural move,
since it is fairly clear that the realization of object agreement is independent of Case
assignment: the object receives Case independent of whether it triggers agreement on
the participle. In other words, agreement is dependent on movement, whereas Case
is not. This suggests that the relevant Case-assigning head is lower in the structure
than Agr, as for example in the following structure in which accusative Case is licensed
under Agree by , while participle agreement results from a separate Agree operation
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.34 (1954-2003)
initiated by an agreement morpheme added to a higher functional head (abstracting
away from head movement):
34
(53)
P
DP
obj.
DP
subj.
VP
t
obj.
V
participle
Agree
Part
Part Aux
Part
Agr Part
Agree
Following Pollock (1989) (among others), I assume that the participle moves into
a position below the base position of the nite auxiliary to combine with its agree-
ment morphology (note that this structure is necessary in all approaches that assume
movement of the participle to a higher Agr head; otherwise the trace left by the auxil-
iary should always block participle movement due to the HMC). This head might be
identied as Asp or some other functional head which occupies a low position in the
inectional eld. For ease of exposition, I have labeled the relevant functional head
simply Part(iciple).
Importantly, the structure (53) allows us to account for the correlation between
overt object movement and participle agreement if the concept of locality developed
in Chomsky (2000, 2001a) is adopted (see Chapter 2 above for details). Recall that the
syntactic derivation proceeds in phases, which are identied as CP and P. Locality
of movement is guaranteed by the following condition, which ensures that operations
triggered by higher functional heads can target only material at the left edge of the next
lower phase, that is, they cannot look into the complement of the phase head.
(54) Phase Impenetrability Condition (Chomsky 2000: 108)
In phase with head H, the domain of H is not accessible to operations
outside , only H and its edge are accessible to such operations.
In other words, the Phase Impenetrability Condition requires that elements must move
to the left edge of their phase to be accessible for movement operations triggered by
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.35 (2003-2053)
higher functional heads. For example, wh-movement of the object to SpecCP requires
that the object has previously moved to the left edge of P. Under this concept of lo-
cality, it is now possible to account for the correlation between overt object movement
and participle agreement in French in the following way. After the object has been
Case-licensed by an Agree relation with , it remains in its base position inside VP.
In this position, it cannot be targeted by an Agree relation started by Agr adjoined to
Part
0
, since Agr cannot look down into the complement of the phase head . Hence,
no participle agreement is possible if the object stays in situ. This analysis implies that
Case checking does not depend on the existence of an unvalued -set in the probe, in
contrast to Chomsky (2000). For the sake of concreteness, I adopt the traditional idea
that Case checking involves an uninterpretable Case feature on T (nominative) and V
(accusative) which is checked against a matching uninterpretable feature on D (cf. e.g.
Chomsky 1993; see Pesetsky & Torrego 2004 for the related claim that agreement i.e.,
-feature checking and case are independent phenomena).
However, if the object undergoes overt movement, triggered by a [+wh] C-head,
for example, it must pass through the left edge of P for reasons of locality. In this
intermediate position (the outer SpecP in (53)), it is accessible for an Agree relation
initiated by the head complex Agr+Part, giving rise to participle agreement. Again, if
we continue to assume that Case renders a goal active in the sense of Chomsky (2000),
this analysis requires that the Case feature marked for deletion is still accessible after
movement to the left edge of P.
35
We can therefore conclude that an analysis in terms
of Agree is capable of deriving the distribution of participle agreement in French, in
contrast to the critique by Koopman (2003). Note that the explanation presented here
becomes only available if we assume that Case and agreement licensing are associated
with different functional heads, lending further support to the concept of agreement
developed above.
.. Agreement and word formation
In the framework adopted in this book, inected verbs are not assembled in the lex-
icon, but rather built in the syntactic and/or morphological component and later
realized by the insertion of phonological exponents (see Chapter 2 for details). Accord-
ingly, an inected verb can only be spelled-out if it is combined with its inectional
afxes (e.g., T, Agr, Asp, Mood etc.) prior to Vocabulary Insertion.
36
In other words,
the requirement that an inectional afx must attach to a lexical host is treated as a
morphophonological requirement that must be satised prior to PF (similar to the
Stray Afx Filter, Baker 1985, 1988). This can be accomplished either in the syntac-
tic component by overt head movement to a higher functional head or at MS by
Morphological Merger which combines the verb root with its inectional morphemes
post-syntactically under (structural) adjacency, giving rise to the impression of appar-
ent syntactic lowering (afx hopping as in the case of nite main verbs in English, cf.
Halle & Marantz 1993; Bobaljik 1994, 1995; Embick & Noyer 2001; Baker 2002).
37
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.36 (2053-2098)
This section deals with the mechanisms that determine the spell-out of inectional
morphemes, focusing on the creation of linear order. Similar to Julien (2002), I assume
that the linearization of morphemes is a function of the way syntactic structures are
mapped into sequences of phonological exponents. More precisely, the linearization
of syntactic structures is assumed to be part of the Spell-out procedure, that is, the
mapping to PF. Following Halle and Marantz (1993), Bobaljik (1995), Embick and
Noyer (2001), I assume that linear order is created by operations connected to Vocab-
ulary Insertion that are part of a separate morphological component, Morphological
Structure (MS).
However, in contrast to Julien, I do not assume that the information provided by
the syntactic output sufces to completely determine the order of inectional markers
on the verb. Instead, I will adopt the more conservative view that afx order is only
partially determined by the syntactic structure. The idea that the insertion of phono-
logical material takes place after syntax at a separate level of Morphological Structure
(MS) suggests that the linearization of terminal nodes, that is, the phonological realiza-
tion of sequences of words and afxes, is also a function of MS: by assumption, linear
order is not a property which is directly encoded in syntactic representations. Instead,
the output of the syntactic component contains only hierarchical information (i.e.,
c-command and dominance relations) which is interpreted by MS and subsequently
mapped into certain linear orders, (cf. Bobaljik 2002; Chomsky 2005). For each pair of
(sister) nodes [X,Y] in a binary branching structure, the syntactic component does not
determine whether X precedes Y or vice versa. Instead, the assignment of precedence
relations is a function of MS. More specically, I assume that the mapping of syntactic
structures into linear orderings is part of the operation of Vocabulary Insertion (cf.
Embick & Noyer 2001: 562).
(55) The Late Linearization Hypothesis
The elements of a phrase-marker are linearized at Vocabulary Insertion.
Still, the order of inectional markers is taken to reect the hierarchical relations
in head adjunction structures created by syntactic movement. On the standard as-
sumption that the morphological derivation must mirror the syntactic derivation (the
Mirror Principle, Baker 1985, 1988), the phonological exponent of a lower functional
head must be closer to the verb stem than the phonological exponents of higher func-
tional heads.
38
This follows if Vocabulary Insertion proceeds in a bottom-up fashion,
as normally assumed in Distributed Morphology. That is, Vocabulary Insertion affects
the verbal or nominal root before it affects functional heads the root adjoins to (so-
called root-out insertion). For example, in the following head adjunction structure,
Vocabulary Insertion targets rst the verbal root. Subsequently, the phonological ex-
ponent of the closest functional head is inserted (here: Asp) and combines with the
verb stem. Finally, the exponent of T is attached to the existing sequence of phono-
logical exponents, ensuring that the exponent of T is the outermost element in the
resulting word.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.37 (2098-2165)
(56) T
Asp T
Asp V
In this fashion, the effects described by the Mirror Principle follow from the way lin-
ear order is created at the point of the derivation where phonological exponents are
inserted into the terminal elements of the syntactic structure. The ordering restric-
tions imposed by the hierarchical structure assembled in the syntactic component are
supplemented by lexical properties of individual Vocabulary items to create the nal
ordering instructions which are sent to the articulatory system.
39
The lexical proper-
ties in question determine the orientation of afxes (i.e., the prex/sufx distinction).
In other words, the linearization of a head adjunction structure depends on selectional
properties of bound forms: prexes select a host to their right, whereas sufxes require
a host to their left. On these assumptions, a structure such as (56) may give rise to the
sequences in (57), but excludes those in (58):
40
(57) a. V+Asp+T (uniformly sufxing)
b. T+Asp+V (uniformly prexing)
c. T+V+Asp (T prex, Asp sufx)
d. Asp+V+T (T sufx, Asp prex)
(58) a. *V+T+Asp
b. *Asp+T+V
The sequences in (57cd) represent possible but typologically marked options, which
are not very frequent across the worlds languages. This can be attributed to the as-
sumption that there is normally a default orientation for inectional afxes in a given
grammar, that is, languages prefer to be uniformly prexing or sufxing (cf. e.g. Bybee
et al. 1990; Hale 1996).
41
In contrast to the purely syntactic approach to afx order proposed in Julien
(2002), the model outlined in this section requires no non-standard assumptions.
Moreover, under the assumption that the order of embedding of functional categories
is not subject to parametrization, the above model appears to be at least as restrictive
as Juliens much more complicated model. Recall, for example, that the (non-existing)
sequence Asp-T-V can be derived in Juliens approach as the result of Asp-to-T move-
ment plus word formation with the V-head which stays in situ (see Section 3.4 above).
In contrast, this order is ruled out in the model presented here without further stip-
ulations: a structure like (56) can only be mapped into a linear sequence where the
exponent of Asp is closer to the verb stem than the exponent of T.
Let us now explore how agreement morphemes play out in this model. Above, it
has been shown that cross-linguistically, the ordering properties of agreement markers
differ considerably from those of other inectional formatives. This was taken to sug-
gest that agreement heads do not project a separate phrase, but are adjoined to other
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.38 (2165-2212)
functional categories. The multitude of ordering possibilities observed with agreement
markers were then attributed to the assumption that agreement morphemes can at-
tach to a variety of functional heads. In what follows, it is argued that the special
ordering possibilities of agreement markers may also result directly from the partic-
ular phrase-structural character of agreement morphemes as elements which adjoin
to other functional heads. The basic idea is that the resulting head adjunction struc-
tures are compatible with different linear orderings and that a given grammar may
choose between the possible outcomes. Compare the following two head adjunction
structures, where the verb root combines with functional heads X and Y:
(59) a. b.
Y
X Y
X V
X
X Y
X V
(59a) represents the structure created by successive head movement to the functional
heads X and Y, while (59b) represents the structure resulting from verb movement to
a functional head X that hosts a base-adjoined agreement morpheme Y. Above, we
noted that a structure like (59a) rules out the linear orderings *V-Y-X and *X-Y-V,
where the higher functional head is realized closer to the verb stem than the lower
functional head. This was attributed to the idea that linearization (i.e., Vocabulary
Insertion) proceeds in a bottom-up fashion, rst affecting the lexical root (i.e., the
most deeply embedded terminal node), then the next higher terminal element etc.
In the case of a normal head adjunction structure such as (59a), the choice of the
next higher element is straightforward. Each terminal element corresponds to a unique
head domain, which excludes the next higher head domain. This is indicated in the
following phrase marker:
(60)
Y
X Y
X V
Following (Chomsky 1986: 9), exclusion is dened as follows:
(61) Exclusion
excludes if no segment of dominates .
In the course of mapping syntactic structures into linear order, the insertion procedure
moves upward in the phrase marker, processing one head domain after the other. The
lexical root constitutes a separate linearization/head domain by default, since it is the
starting point of Vocabulary Insertion.
42
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.39 (2212-2273)
Let us now address the question of whether (59b) imposes similar restrictions on
the mapping into linear order. Again, Vocabulary Insertion starts at the lexical root.
Subsequently, the insertion procedure has to determine the next terminal node that is
targeted by Vocabulary Insertion. However, (59b) crucially differs from (59a) in that
both X and Y are part of the same head domain, with Y adjoined to X.
(62)
X
X Y
X V
In other words, in contrast to (59a), the head domain of X does not exclude Y and
the insertion procedure has in principle access to both terminal elements in the head
domain. I assume that in this structure, insertion may target either X or Y as the next
terminal node where a phonological exponent is inserted. Under the assumption that
the choice between X and Y (during Vocabulary Insertion) is open to parametrization
in structures like (59b), we derive the result that the kind of head adjunction structures
created by the insertion of agreement heads may give rise to a greater variety of pos-
sible linear orderings across languages even in a subset of languages that is uniform
concerning the adjunction site of Agr (e.g., in which AgrS is always adjoined to T).
.. Section summary
In this section, I have developed a realizational model of predicate-argument agree-
ment. To account for the special distributional properties of agreement markers, it
was proposed that agreement heads are base-generated as adjuncts to substantial
functional heads. Here, the usual choices are T for subject agreement and for ob-
ject agreement. By assumption, however, agreement morphemes may attach to any
other functional head, which is one reason for the multitude of ordering possibili-
ties found across languages. Subsequently, I addressed the question of how feature
matching between the controllers -set and the agreement morpheme is established
during the syntactic derivation. Following the basic insight reached in Section 3.3
above that an analysis in terms of the operation Agree is to be preferred over the
traditional mechanism of spec-head agreement, I claimed that the feature content
of agreement morphemes is identied under closest c-command with an appropri-
ate goal (the interpretable -set of the DP that controls agreement). Under the as-
sumption that agreement morphemes may attach to other functional heads apart
from T and , this analysis predicts that Case and agreement licensing can be dis-
sociated. This prediction is borne out in Georgian where the verb agrees with the
external argument, regardless of whether the latter carries nominative or ergative
case. Moreover, it was shown that participle agreement in French is also amenable
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.40 (2273-2331)
to an analysis where AgrO is dissociated from the functional head that assigns ac-
cusative Case. The proposed analysis in terms of Agree also accounts for the fact
that participle agreement requires overt object movement. The latter property was
attributed to the Phase Impenetrability Condition, which forces movement of the ob-
ject to the left edge of P, raising the object into the agreement domain of the higher
AgrO-morpheme. Finally, I addressed the question of how agreement markers are at-
tached to the verb stem. I argued that the linearization of inectional morphemes
is only partially determined by hierarchical relations established in the syntax and
that the ordering instructions sent to the articulatory system are also inuenced by
lexical properties of individual Vocabulary items (i.e., the prex/sufx distinction).
On the assumption that the mapping of syntactic structure into linear orderings is
a function of Vocabulary Insertion, I derived the effects of the Mirror Principle by
assuming that head adjunction structures are processed in a bottom-up fashion. Fur-
thermore, it became clear that the workings of the mapping procedure introduce
another possible source of cross-linguistic variation with respect to the linear place-
ment of agreement markers apart from the possibility that Agr may attach to different
functional heads.
. Multiple agreement: Inected complementizers in Germanic
Drawing on evidence from the phenomenon of complementizer agreement, this sec-
tion argues that we must acknowledge the existence of another agreement mechanism
in addition to the analysis of canonical syntactic agreement (in terms of Agree) out-
lined in the previous section. More specically, it is claimed that complementizer
agreement (and presumably other instances of multiple agreement) results from the
post-syntactic insertion of a so-called dissociated agreement morpheme at the level of
Morphological Structure.
A number of non-standard varieties of Germanic is characterized by the pres-
ence of inection in the C-domain, usually referred to as complementizer agreement
(cf. Bayer 1984; Altmann 1984; Wei 1998, 2002 on Bavarian; Bennis & Haegeman
1984; Haegeman 1990, 1992; Shlonsky 1994; de Vogelaer et al. 2002 on (West) Flem-
ish; de Haan & Weerman 1986; Hoekstra & Marcz 1989 on Frisian; Zwart 1993a,
1993b, 1997a on dialects of the eastern and southern Netherlands; Hoekstra & Smits
1999 for an overview). Thus, we can observe a peculiar form of multiple agreement
where there is not only an agreement relation between the subject and the verb, but
also between the subject and elements located in the CP domain (primarily comple-
mentizers). This kind of agreement is pretty uncommon across the worlds languages,
but frequently exhibited by German and Dutch dialects, which share another typo-
logically marked property, V2. Therefore, it is often assumed that the availability of
inected complementizers is linked to the V2 property. The phenomenon is illus-
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.41 (2331-2369)
trated in (63) with examples fromWest Flemish (Bennis &Haegeman 1984; Haegeman
1990, 1992):
West Flemish
(63) a. da-n *(=k)
that-1sg=clit.1sg
ik
I
werk-en
work-1sg
That I work
b. da-t *(=j)
that-2sg=clit.2sg
gie
you
werk-t
work-2sg
c. da-t *(=j)
that-3sg=clit.3sg.masc
ij
he
werk-t
work-3sg
d. da-t (=ze)
that-3sg=clit.3sg.fem
zie
she
werk-t
work-3sg
e. da-t (=t)
that-3sg=clit.3sg.neut
tet
it
werk-t
work-3sg
f. da-n (=me)
that-1pl=clit.1pl
wunder
we
werk-en
work-1pl
g. da-t *(=j)
that-2pl=clit.2pl
gunder
you.pl
werk-t
work-2pl
h. da-n (=ze)
that-3pl=clit.3pl
zunder
they
werk-en
work-3pl
In the above examples, we can see that the complementizer carries an inectional
ending the shape of which varies in relation to the -features of the subject. While
1sg, 1pl, 3pl subjects trigger the ending -n, -t is found with 2sg, 3sg, 2pl subjects. In
West Flemish, the form of complementizer agreement is identical to the verbal agree-
ment endings. Note that the inected complementizers are followed by subject clitics
which are obligatory in some person/number combinations and optional in others
(obligatory: 1sg, 2sg, 3sg.masc, 2pl; optional: 3sg.fem, 3sg.neut, 1pl, 3pl).
43
West Flemish differs in two important ways from the other Germanic languages
that exhibit inected complementizers. First, West Flemish has a complete paradigmof
complementizer agreement for all persons and numbers. Second, only in West-Flemish
a clitic and a full subject pronoun can co-occur in addition to complementizer agree-
ment, as shown by the examples in (63). The examples in (64) demonstrate that clitic
doubling is impossible with full referential subject DPs and that the clitic is obligatory
if no other subject is present (examples taken from Shlonsky 1994: 354).
(64) a. *da-t=ze
Marie
Marie
werk-t
work-3sg
That Marie works
b. da-t *(=ze)
werk-t
work-3sg
That she works
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.42 (2369-2449)
In other German and Dutch dialects, complementizer agreement is generally restricted
to certain person/number combinations. In most varieties of Bavarian, complemen-
tizer agreement is found only in 2nd person contexts, in the eastern Netherlands, it is
restricted to 1pl, in the south to 1pl and 3pl, and in Frisian to 2sg (plus 2pl in some
varieties, similar to the Brabants dialect of Dutch).
44
Another kind of typological variation concerns the shape of the agreement afxes
found on C. In the majority of varieties these are identical with the verbal agreement
endings (cf. the West Flemish data in (63)). However, in a set of dialects spoken in
the eastern Netherlands and Brabants, the shape of complementizer agreement differs
from verbal agreement (Zwart 1993b; similar facts can be observed with 1pl in some
Lower Bavarian dialects, see below). This is illustrated by the following example from
an eastern dialect, where the 1pl ending on the complementizer is -e, while it is -t on
the verb (in clause-nal position):
45
(65) datt-e
that-1pl
wij
we
speul-t
play-1pl
That we play (Zwart 1993b: 253)
Finally, in those varieties where complementizer agreement differs from verbal agree-
ment, the former replaces the latter in inversion contexts (cf. e.g. Zwart 1993b: 254).
Again, relevant examples come from the eastern Netherlands (similar examples are
found in Brabants):
46
(66) a. Wij
we
speul-t/*-e.
play-1pl
We play.
b. Waar
where
speul-e/*-t
play-1pl
wij?
we
Where do we play? (Zwart 1993b: 254)
The examples in (66) show that the nite verb must carry regular verbal agreement
(found e.g. on verbs in clause-nal position) when it appears in a subject-initial main
clause, as in (66a), whereas in inversion contexts, it displays the agreement ending
normally found on complementizers, cf. (66b). Thus, the ultimate shape of the nite
verb is determined by the syntactic position in which it occurs.
47
In some varieties of Germanic, the presence of agreement in C licenses referential
pro-drop. This is true of Frisian (2sg) and Bavarian (2sg, 2pl), for example:
Frisian
(67) a. Kom-st
come-2sg
(do)
you
jn?
tonight
Do you come tonight?
b. dat-st
that-2sg
(do)
you
jn
tonight
kom-st
come-2sg
That you come tonight (Zwart 1993b: 256)
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.43 (2449-2561)
Bavarian
(68) a. Kumm-st
come-2sg
pro noch
to
Minga,
Munich
dann
then
muas-st
must-2sg
pro me
me
bsuacha.
visit
If you come to Munich you must visit me.
b. Kumm-ts
come-2pl
pro noch
to
Minga,
Munich
dann
then
mass-ts
must-2pl
pro me
me
bsuacha.
visit
If you come to Munich you must visit me. (Bayer 1984: 211)
(69) a. ob-st
whether-2sg
pro noch
to
Minga
Munich
kumm-st
come-2sg
...whether you come to Munich
b. ob-ts
whether-2pl
pro noch
to
Minga
Munich
kumm-ts
come-2pl
...whether you (pl) come to Munich (Bayer 1984: 240)
Hence, Frisian and Bavarian differ from West Flemish, which requires the presence of
a subject clitic in the absence of a full pronoun/DP subject, as shown in (64b) above.
Dialects spoken in the east and south of the Netherlands behave similar to West Flem-
ish the presence of complementizer agreement does not license pro-drop (examples
taken from Zwart 1993b: 257):
Southern varieties
(70) a. Komm-e
come-pl
*(ze)?
they
Do they come?
b. ovv-e
whether-pl
*(ze)
they
komme
come-pl
whether they come
Eastern varieties
(71) a. Speul-e
play-1pl
*(we)?
we
Do we play?
b. datt-e
that-1pl
*(wij)
we
speul-t
play-1pl
That we play
The descriptive ndings reached so far are summarized in Table 4.
48
The following section briey reviews a selection of the most prominent theoretical
approaches to the phenomenon of complementizer agreement in Germanic, introduc-
ing more data from other languages and highlighting a number of general issues that
an adequate analysis has to address.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.44 (2561-2586)
Table 4. Properties of complementizer agreement in Germanic.
Paradigm Identical AgrC replaces Additional clitic AgrC
with AgrS AgrS (inversion doubling: licenses
contexts) AgrC+clit.+pron. pro-drop
West Flemish complete yes no yes no
Frisian defective (2sg;
in some
varieties also
2pl)
yes no no yes
NL: Eastern defective (1pl) no yes no no
dialects
(Hellendoorn)
NL: Southern defective (1pl,
3pl)
yes no no no
dialects
Bavarian defective (2sg,
2pl; plus 1pl in
some varieties)
yes (apart
from 1pl)
no (apart from
1pl)
no yes
.. Previous accounts of complementizer agreement
The phenomenon of complementizer agreement is commonly attributed to a struc-
tural relation that holds between C and a lower inectional head. This relation is often
modeled in terms of syntactic head movement which provides C with inectional fea-
tures (e.g. Hoekstra & Marcz 1989: I-to-C; Zwart 1993b: AgrS-to-C). Alternatively, it
has been proposed that the inection found on the complementizer heads its own pro-
jection in the C-domain and is licensed in a specier-head relation (Shlonsky 1994). In
the following, it is shown that both analyses are confronted with a number of concep-
tual and empirical problems. In addition, I will examine whether the data in question
can be more successfully accounted for by an analysis that makes use of the Agree
mechanism, as proposed by Carstens (2003). Note that the following discussion is
rather selective, focusing on the major lines of thought on this topic, while alternative
approaches are only briey commented on.
... In-to-C movement
The traditional approach to complementizer agreement attributes the inection on
C to a structural relation that holds between C and a lower inectional head, which is
usually modeled in terms of syntactic head movement.
49
For example, Hoekstra and
Marcz (1989) attribute the presence of agreement morphology on C to In-to-C
movement. The problematic fact that agreement is realized on the nite verb as well is
analyzed in terms of feature percolation from the C+In complex down to the verb.
This general idea is adopted in slightly modied form in various publications of Jan-
Wouter Zwart (1993a, 1993b, 1997a, 2001), who argues that AgrS-to-C movement is
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.45 (2586-2651)
triggered as a means to (case-) license the subject in SpecAgrSP in embedded clauses.
In order to account for the fact that we nd multiple agreement on both the verb and
the complementizer, Zwart assumes (1993b: 267, fn. 21) that it is not the complete
AgrS head that moves to C, but only an abstract agreement morpheme, which is in-
volved in the formal licensing of the subject. In addition, he proposes that AgrS carries
an additional (agreement) feature that engages in the licensing of the inected verb
(at LF), which stays in situ in the overt syntax.
50
The latter assumption is supposed
to ensure that complementizer agreement does not block the presence of agreement
morphology on the verb.
51
According to Zwart, further support for an analysis in terms of (general) AgrS-
to-C movement comes from the eastern dialects where the morphological realization
of complementizer agreement differs from that of verbal agreement, repeated here for
convenience:
(72) a. Wij
we
speul-t/*-e.
play-1pl
We play.
b. Waar
where
speul-e/*-t
play-1pl
wij?
we
Zwart argues that this asymmetry follows from the fact that the verb occupies different
head positions in (72a) (AgrS) and (72b) (C). The fact that the nite verb apparently
inects for complementizer agreement in (72b) is analyzed as another overt morpho-
logical reex of AgrS-to-C movement (recall that Zwart assumes that AgrS-to-C takes
place in all embedded clauses and in main clauses with subject-verb inversion).
At closer inspection, however, the phenomenon in question constitutes a serious
problem for an analysis in terms of head movement. Relevant data come from Hellen-
doorn, an East Netherlandic dialect (Carstens 2003; Ackema &Neeleman 2004), where
complementizer agreement can be observed with 1pl subjects, as shown in (73).
Hellendoorn
(73) datt-e
that-1pl
wiej
we
noart
to-the
park
park
loop-t
walk-1pl
That we are walking to the park (Carstens 2003: 397)
Similar to other dialects spoken in the eastern Netherlands, the realization of 1pl com-
plementizer agreement (-e) differs from the relevant verbal agreement ending (-t).
Furthermore, nite verbs in subject-verb inversion contexts exhibit the agreement
morphology typical of complementizers (1pl only), whereas normal verbal agree-
ment is found on verbs in subject-initial main clauses:
Hellendoorn
(74) a. Volgens
according-to
miej
me
lop-e
walk-1pl
wiej
we
noart
to-the
park.
park
According to me we are walking to the park.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.46 (2651-2731)
b. Wiej
we
loop-t
walk-1pl
noart
to-the
park.
park
We are walking to the park. (Carstens 2003: 398)
In other words, Hellendoorn seems to exhibit exactly the same behavior as the other
eastern dialects discussed by Zwart. Interestingly, however, the presence of an adjunct
that intervenes between C and the subject blocks the availability of complemen-
tizer agreement in Hellendoorn. This restriction holds for both main and embedded
clauses, as can be seen from the examples in (75).
Hellendoorn
(75) a. dat/*datt-e
that/that-1pl
[op
on
den
the
wrmsten
warmest
dag
day
vant
of-the
joar]
year
wiej
we
tegen
against
oonze
our
wil
will
ewrkt
worked
hebt.
have
That on the warmest day of the year we have worked against our will
b. Volgens
according-to
miej
me
loop-t/*lop-e
walk-1pl/walk-1pl
[op
on
den
the
wrmsten
warmest
dag
day
vant
of-the
joar]
year
ook
also
wiej
we
noart
to-the
park.
park
According to me we are also walking to the park on the warmest day of
the year. (Carstens 2003: 398)
On the assumption that complementizer agreement is the result of AgrS-to-C move-
ment, its non-availability in V2-clauses such as (75b) where exactly this movement
operation has taken place comes as a surprise. More generally, under standard as-
sumptions, it is clear that the intervening PP should not block head movement (in
both main and embedded contexts).
52
Similar adjacency effects can be observed in the other Germanic varieties which
exhibit complementizer agreement, cf. the following example from Bavarian, where
complementizer agreement is impossible if an adverb intervenes between C and a full
subject pronoun:
53
Bavarian
(76) *obwoi-st
although-2sg
woartscheints
probably
du
you
ins Kino
to-the movies
ganga
gone
bist
are
although you probably went to the movies (Gnther Grewendorf, p.c.)
We can therefore conclude that an analysis in terms of head movement cannot ade-
quately account for the phenomenon of complementizer agreement (see e.g. Carstens
2003: 401ff. for further arguments against Zwarts analysis of complementizer agree-
ment). In the next section, we will consider an alternative approach that is based on
the assumption that agreement is universally licensed in a specier-head conguration,
attributing complementizer agreement to the presence of an AgrP in the CP domain.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.47 (2731-2767)
... AgrP within CP
Shlonsky (1994) develops an account of complementizer agreement which is based
on a split CP structure and makes use of the idea that agreement phenomena always
involve a specier-head relation between the controller and a syntactic agreement head
(Belletti 1990, 2001; Chomsky 1993; Koopman 2003; cf. Roberts 1994 for a related
approach). The analysis focuses on West Flemish, but is claimed to extend to the other
Germanic languages that exhibit inected complementizers. Shlonsky proposes the
following structure for the left periphery of the clause, where AgrC is the host of the
agreement features that give rise to inected complementizers:
(77) [
CP
[
C
C [
AgrCP
[
AgrC
AgrC [
IP
... ]]]]]
It is claimed that the content of AgrC is licensed either via insertion of a subject
clitic in its specier or by movement of a full nominal subject to SpecAgrCP, which
by assumption is an A-position. Thus, the additional inection in the C system is li-
censed in a separate agreement relation apart from the specier-head conguration
within IP (which serves to establish verbal agreement). Shlonsky further assumes that
the -features located in AgrC attach to the complementizer via syntactic AgrC-to-
C movement. The analysis of an example such as (78) is illustrated by the phrase
marker in (79). The full subject pronoun is located in its Case position SpecIP, where
it licenses the set of -features located in I
0
, giving rise to verbal agreement. The -
features of AgrC enter into a separate licensing relation with the subject clitic, which is
base-generated in SpecAgrCP.
(78) da-t (=ze)
zie
she
werk-t
work-3sg
(79)
AgrCP
IP
AgrC Spec
AgrC
Spec I
I VP zie
t
V
werk-t
C
CP
C
da
-t
ze
In the absence of a subject clitic, the subject (either a full DP or pronoun) is assumed to
undergo A-movement to SpecAgrCPto license the -features in AgrC.
54
The adjacency
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.48 (2767-2828)
effects observed above are taken to follow from the assumption that nothing can inter-
vene between C (hosting the inected complementizer) and the clitic/full pronoun in
SpecAgrCP (Shlonsky rules out adjunction to AgrCP).
However, this analysis faces the problem that the relevant spec-head relation be-
tween the subject and AgrC is never overtly realized, either because AgrC-to-C move-
ment obscures the existence of a previous spec-head relation and the subject appears to
the right of the inected complementizer or because there is simply no subject overtly
present, as for example in Bavarian, which exhibits pro-drop in 2nd person contexts
(cf. (69) above, repeated here for convenience):
(80) a. ob-st
whether-2sg
pro noch
to
Minga
Munich
kumm-st
come-2sg
...whether you come to Munich
b. ob-ts
whether-2pl
pro noch
to
Minga
Munich
kumm-ts
come-2pl
...whether you (pl) come to Munich (Bayer 1984: 240)
Another shortcoming of Shlonskys analysis concerns the hierarchical relation between
the functional head that carries the agreement features and the lexical element where
agreement is realized morphologically. It is standardly assumed that the lexical host of
agreement morphology has to move up to a higher functional head to check agreement
features. In other words, Agr is higher than its lexical host. However, in Shlonskys
account, its the other way around (note that an analysis in terms of I-to-C movement
is subject to the same criticism).
55
In addition to these conceptual issues, Shlonskys analysis can be shown to be in-
compatible with data fromother languages with inected complementizers. Recall that
Shlonsky claims that his account carries over to languages such as Frisian and Bavar-
ian, which in contrast to West Flemish generally exclude the presence of a subject clitic
in contexts with complementizer agreement. Thus, it is exactly these contexts that li-
cense pro-drop in Bavarian (Frisian behaves similarly, as noted above). This raises the
question of how the content of AgrC can be licensed when the specier of AgrCP is
apparently not lled. Shlonsky is led to assume that in Bavarian (and Frisian), SpecA-
grCP is occupied by a non-overt counterpart of the subject clitics of West Flemish,
which serves to identify the -features present in AgrC in the contexts that license
complementizer agreement (i.e., 2sg and 2pl). However, note that Bavarian has an
elaborate system of subject clitics, exhibiting gaps only for the 2nd person nomina-
tive forms (in contrast to claims by Shlonsky (p. 362); cf. Altmann 1984; Bayer 1984;
Wei 1998; see Chapter 5 for details). Now, under Shlonskys analysis, it is totally un-
expected that complementizer agreement is only found in those cases where no (overt)
subject clitic exists. Instead, we should expect to nd inected complementizers rather
in those instances where overt subject clitics exist, that is, 1sg/pl, 3sg/pl, contrary to
the facts.
Finally, the analysis proposed by Shlonsky makes wrong predictions for languages
which show a reex of Agr-on-C in main clauses as well. Shlonsky adopts an asymmet-
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.49 (2828-2890)
rical analysis of V2 (basically following Travis 1984; Zwart 1993a), according to which
the nite verb moves to AgrC in subject-initial clauses and further to C in clauses with
topicalized non-subjects. However, verb movement to AgrC (and further up) does not
result in double afxation of AgrS and AgrC to the verb:
(81) *Werk-t-t
work-Agrs.3sg-Agrc.3sg
Marie?
Marie
Does Marie work? (Shlonsky 1994: 364)
To rule out examples such as (81), Shlonsky stipulates following Rizzi and Roberts
(1989), who analyze head movement as a substitution operation that the contents
of AgrC
0
must be deleted to create a position into which the inected verb may sub-
stitute (p. 367). However, we have seen above in example (74) (repeated here for
convenience) that in languages such as Hellendoorn (and other eastern dialects), there
is a reex of Agr-on-C in V2 clauses in the sense that the verb carries complementizer
agreement in inversion contexts, whereas it retains its normal agreement pattern in
subject-initial clauses:
Hellendoorn
(82) a. Volgens
according-to
miej
me
lop-e
walk-1pl
wiej
we
noart
to-the
park.
park
According to me we are walking to the park.
b. Wiej
we
loop-t
walk-1pl
noart
to-the
park.
park
We are walking to the park.
These data create two problems for Shlonskys analysis. First, it is unclear why the con-
tent of AgrC is not deleted, but rather overwrites the content of AgrS in dialects such
as Hellendoorn. Second, within the asymmetrical analysis of V2 adopted by Shlonsky,
the contrast between (82a) and (82b) suggests that the -features giving rise to com-
plementizer agreement are associated rather with C instead of AgrC, since the verb
carries complementizer agreement only when it occupies the higher head (in inversion
contexts).
Summing up, we have seen that an analysis of complementizer agreement in terms
of a separate specier-head relation in the (split) CP domain can avoid some of the
conceptual problems that arise in the traditional head movement approach, for exam-
ple by providing separate licensing congurations for distinct agreement phenomena.
However, exactly the strong assumption that agreement always involves a specier-
head conguration necessitates a number of additional assumptions to account for the
West Flemish data. Furthermore, it has been shown that this strong assumption leads
to serious complications and wrong empirical predictions if data from other Germanic
languages are taken into consideration. We can therefore conclude that complemen-
tizer agreement in Germanic should not be modeled in terms of a specier-head
relation, which corroborates our ndings in Section 3.3 above concerning the struc-
tural relation involved in predicate-argument agreement. In the next section, we will
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.50 (2890-2973)
see whether the problems encountered so far can be resolved by an analysis that makes
use of the Agree mechanism proposed in Chomsky (2000, 2001a).
... Multiple Agree
Recently, Carstens (2003) has proposed an analysis of complementizer agreement in
Germanic that makes use of the probe/goal mechanism developed in Chomsky (2000,
2001a, 2001b). Carstens assumes that the agreement morpheme on C (in her anal-
ysis: Fin) is valued under closest c-command (i.e., via the operation Agree) by the
interpretable -features of the subject which occupies SpecTP.
56
Before this Agree op-
eration takes place, the subjects -set has already been accessed by the uninterpretable
-features of T (in its base position SpecP, prior to movement to SpecTP), which
serves to Case-license the subject and ensures that the nite verb carries agreement
morphology. Note that this instance of multiple Agree represents a deviation from
the assumption that a category with an inactive (i.e., deletion-marked) Case feature
should no longer be accessible to syntactic operations (Chomsky 2000, 2001a). That
is, it should not be possible for the subject to enter into an additional probe-goal
relation with the non-interpretable -features located in C after its Case feature has
been deletion-marked by T. Accordingly, Carstens slightly relaxes Chomskys original
proposal and claims that a Case feature that has been marked for deletion remains syn-
tactically active until the next strong phase is reached (cf. Pesetsky & Torrego 2001 and
Section 3.6.2 above for similar assumptions). This enables the subject to participate in
multiple Agree relations, in the case at hand rst with T and then with C.
Again, it can be shown that this analysis does not account for a number of restric-
tions on complementizer agreement in a satisfactory way. The rst set of problematic
data has to do with the fact that the availability of complementizer agreement is subject
to an adjacency requirement, as noted above. Recall that in Hellendoorn, for example,
the presence of an adjunct which intervenes between Cand the subject blocks the avail-
ability of complementizer agreement. This restriction holds for embedded clauses as
well as inversion contexts (cf. (75) above, repeated here as (83)).
Hellendoorn
(83) a. dat/*datt-e
that/that-1pl
[
PP
op
on
den
the
wrmsten
warmest
dag
day
vant
of-the
joar]
year
wiej
we
tegen
against
oonze
our
wil
will
ewrkt
worked
hebt.
have
b. Volgens
according-to
miej
me
loop-t/*lop-e
walk-1pl/walk-1pl
[
PP
op
on
den
the
wrmsten
warmest
dag
day
vant
of-the
joar]
year
ook
also
wiej
we
noart
to-the
park.
park
According to me we are also walking to the park on the warmest day of
the year.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.51 (2973-3016)
Carstens attributes the absence of complementizer agreement in (83) to an interven-
tion effect in the sense of Chomsky (2000, 2001a). More precisely, she claims that the
adjoined adverbials in (83) bear an abstract Case feature that identies the adverbial as
a possible goal for the -set in C. As a consequence, the adverbial disrupts closest c-
command of the subject by C
0
(p. 398), thereby blocking the valuing and realization
of complementizer agreement. However, apart from the fact that it is quite unusual
to assume that PP adverbials like op den wrmsten dag vant joar carry a Case feature,
this analysis leads to a serious problem with respect to the Agree relation that serves
to value the -set of T. Recall that Carstens assumes that the -set of T initiates an
Agree relation with the subject in SpecP. Now, under her analysis, we should expect
that adverbials that intervene between T and the base position of the subject should
give rise to the same kind of intervention effects that are taken to block Agree between
C and the subject in SpecTP. Of course, this is not the case. Therefore, we can con-
clude that the analysis of Carstens (2003) does not provide an adequate account of the
Hellendoorn data either. In the next section, I will present an alternative account of
complementizer agreement which is based on the idea that the relevant phenomena
should receive an explanation in terms of morphological operations carried out in the
post-syntactic level of MS.
.. A Late Insertion account of complementizer agreement
The fact that complementizer agreement is sensitive to an adjacency requirement sug-
gests that this form of agreement is not a purely syntactic phenomenon, but at least
partially determined by properties of PF (or MS). A similar conclusion is reached
by Ackema and Neeleman (2004: Ch. 7), who develop an account of complemen-
tizer agreement which is based on the assumption that feature checking may also be
accomplished in the phonological component of grammar (i.e., PF). More precisely,
they propose the following PF feature checking rule, which applies if C and the subject
occur in the same prosodic phrase (marked by braces):
57
(84) Germanic complementizer agreement
{[C (Prt) (Add) (Plr)] [D (Prt) (Add) (Plr)]}
{[C (Prt
i
) (Add
j
) (Plr
k
)] [D (Prt
i
) (Add
j
) (Plr
k
)]}
(Ackema & Neeleman 2004: 241)
The rule in (84) serves to identify the set of -features associated with C (Prt = Par-
ticipant, Add = Addressee, Plr = Plural) with the relevant (interpretable) -features of
the subject. The adjacency effect illustrated in (83) is then explained as follows: due to
the presence of an intervening XP between C and the subject, rule (84) cannot apply,
since the complementizer and the subject are in two different prosodic domains.
Thus, an account in terms of PF feature checking seems to be able to capture the
major empirical facts of the phenomenon under consideration and is therefore supe-
rior to the syntactic analyses discussed above. However, this analysis requires that PF
is endowed with powerful syntax-like properties like identifying, checking and ma-
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.52 (3016-3087)
nipulating -features. This becomes particularly clear when we take a closer look at
Ackema and Neelemans explanation of the fact that complementizer agreement is al-
ways with the subject and never with the object. Even in contexts where C forms a
prosodic phrase with a scrambled object (excluding the subject), the complementizer
cannot agree with the object and must appear in its non-agreeing form:
(85) a. dat
that
[zulke
such
boeken]
i
books
zelfs
even
Jan
Jan
t
i
niet
not
leest
reads
That even John does not read such books
b. {dat zulke boeken} {zelfs Jan} {niet leest} (Ackema & Neeleman 2004: 242)
To prevent that the feature content of C is identied with the objects -features, Ack-
ema and Neeleman (p. 242) assume that PF feature checking can only access elements
in A-positions. By assumption, then, the scrambled object in (85) is not visible for
the relevant PF rule since it occupies an A-position. However, this requires that PF is
not only capable of identifying, checking and manipulating -features. In addition,
PF must be sensitive to the A/A-distinction. In other words, Ackema and Neele-
man claim that information which is only legible in the syntactic module (types of
syntactic positions) must also play a crucial role in PF-operations, contra standard as-
sumptions. This establishes a syntax after the syntax proper, which is a rather dubious
theoretical move.
Next, it is shown that Ackema and Neelemans account also faces a serious empir-
ical problem. More specically, it appears that the evaluation of agreement features in
the C-system does not involve a dependency between C and the subjects -features.
58
Rather, complementizer agreement is shown to be parasitic on the overt presence of
an inected verb at PF.
The assumption that C carries its own set of non-interpretable -features, which
initiates a checking relation with the subject (either in the syntax or at PF), predicts
that the establishment of complementizer agreement is independent of the realization
of verbal agreement. At least in Bavarian, however, this expectation is not borne out
by the facts. Consider the comparative clauses in (86) (Bayer 1984: 269):
59
(86) a. DResl
the-Resl
is
is
gresser
taller
[als
than
wia-st
as-2sg
du
you
bist].
are
Resl is taller than you are.
b. *DResl
the-Resl
is
is
gresser
taller
[als
than
wia-st
as-2sg
du].
you
c. DResl
the-Resl
is
is
gresser
taller
[als
than
wia
as
du].
you
In comparatives, overt agreement on C leads to ungrammaticality if the nite verb
is absent from the structure, cf. (86b). The sentence becomes acceptable when the
complementizer bears no inection, cf. (86c), an apparent violation of the general-
ization that complementizer agreement is obligatory in all contexts (for 2nd person),
see above. This suggests that the possibility of complementizer agreement depends
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.53 (3087-3144)
on the presence of a nite verb within the same clause. In other words, it appears that
Agr-on-C is parasitic on the presence of another already valued agreement morpheme,
Agr-on-T. Crucially, these examples showthat agreement between the complementizer
and the subject cannot be implemented in terms of a checking relation between the set
of -features in C and the lower subject. Otherwise one would expect examples such
as (86b) to be grammatical (the -set of C should be able to enter into a checking rela-
tion with the subjects -set). This conclusion holds for a syntactic analysis in terms of
Agree (Carstens 2003) as well as for an account involving PF checking rules as proposed
by Ackema and Neeleman (2004).
Still, comparatives such as (86b, c) provide further evidence for the suggestion that
complementizer agreement must operate post-syntactically. It is standardly assumed
that comparatives are to be analyzed as the result of post-syntactic PF-operations that
delete the inected verb in the second clause, as shown in (87) (cf. Bresnan 1973;
Lechner 2001).
60
(87) DResl
the-Resl
is
is
gresser
taller
[als
than
wia (*-st)
as-2sg
du
you
bist].
(are)
Resl is taller than you are.
Under this scenario, we would not expect any interaction with complementizer agree-
ment if insertion and valuation of agreement morphemes were to take place in the
syntax, since the nite verb would be present throughout the whole syntactic deriva-
tion, being subject to deletion only after the structure has been transmitted to Mor-
phological Structure (and, nally, PF). In other words, it would remain a mystery why
post-syntactic deletion of the nite verb affects the realization of Agr-on-C in these
contexts in the way it does. In contrast, this interaction comes out much more natu-
rally in a Late Insertion model where the insertion and valuation of Agr-morphemes
on C take place after syntax as well, in the sense that the post-Spell-out operations
that bring about agreement may be sensitive to other post-Spell-out operations such
as deletion of the nite verb in examples like (86b, c). A similar argument can be made
with respect to sluicing, that is, instances where an IP within a wh-CP is elided (cf. e.g.
Ross 1969; Merchant 2001):
(88) I
I
woas
know
dass-ts
that-2pl
ihr
you
a
a
Madl
girl
gseng
seen
hoabts,
have-2pl
owa
but
I
I
woas
know
net
not
wo (*-ts)
where-2pl
ihr
(you
a
a
Madl
girl
gseng
seen
hoabts.
have-2pl)
I know that you have seen a girl, but I dont know where. (Gnther Grewen-
dorf, p.c.)
In examples such as (88), we can observe that complementizer agreement is not avail-
able after the lowest IP has been deleted at PF. Again, this sensitivity to post-syntactic
processes is totally unexpected if it is assumed that complementizer agreement is es-
tablished by syntactic operations. We can therefore conclude that certain properties of
complementizer agreement in Germanic cannot be accounted for by a purely syntactic
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.54 (3144-3179)
analysis in terms of Agree. Instead, the sensitivity to properties such as linear adjacency
or PF-deletion seems to indicate that the phenomenon of complementizer agreement
is to be located in the post-syntactic components of grammar, that is, at MS or, more
generally speaking, the mapping to PF.
To account for these observations, I suggest a hybrid model of agreement in which
agreement phenomena may result from either the pre-syntactic or the post-syntactic
insertion of agreement morphemes. As proposed in Section 3.6 above, canonical sub-
ject agreement is then analyzed as the result of an agreement morpheme which enters
the syntactic derivation as part of T
0
and is valued by the operation Agree:
(89) [ ... [ T + Agr ... [ subject ... ]]]
CP TP P
Agree
In contrast, complementizer agreement is analyzed as resulting from a morphological
operation, the post-syntactic insertion of a dissociated Agr-morpheme at the level of
MS. Hence, I assume that the constituent structure of morphemes derived in the syn-
tax can be modied by the post-syntactic insertion of (functional) morphemes which
adjoin to syntactic terminal nodes (cf. Marantz 1992; Halle & Marantz 1993; Noyer
1997; Embick 1997). Following Embick (1997), these morphemes are called dissoci-
ated, since they are not present in the syntactic derivation and merely reect properties
expressed by structural congurations in the syntax proper.
61
Accordingly, dissociated
morphemes are not part of the numeration. The insertion process is illustrated by the
following pair of phrase markers, where (90a) represents the output of the syntactic
component and (90b) shows the structure resulting from the insertion of a dissociated
Agr-morpheme to the C head, which is subsequently subject to Vocabulary Insertion
(note that the Agr-morpheme on T is already present and valued in the syntax).
(90) a.
C
T
TP C
subj.
VP T
Agr T
V t
V
T
CP
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.55 (3179-3230)
b.
C
T Agr
(inserted at MS)
TP C
subj. C
VP T
Agr T
V t
V
T
CP
How can we ensure that the feature content of the post-syntactically inserted agree-
ment morpheme on C matches the -features of the subject? By assumption, feature
matching between Agr-on-C and the subject does not take place directly, but is me-
diated by another agreement morpheme that has been valued by a syntactic Agree
relation. Recall that the Bavarian comparatives in (86)(87) suggest that the licens-
ing of Agr-on-C depends on the presence of (an already valued) Agr-on-T. To capture
this observation, I assume that the feature content of the dissociated Agr-morpheme
on C is determined by the feature content of an Agr-morpheme that has already
been valued in the syntax, that is, Agr-on-T. More precisely, I propose that a disso-
ciated morpheme which expresses agreement with an argument (here: the subject) is
a copy of a relevant Agr-morpheme that has been valued by a syntactic Agree relation
(here: Agr-on-T). This mechanism ensures feature identity between these different
types of Agr-morphemes (which both reect the -feature content of the same argu-
ment).
62
Thus, we can formulate the following (informal) condition on the insertion
of dissociated Agr-morphemes.
63
(91) The insertion of a dissociated Agr-morpheme is parasitic on the presence of
an Agr-morpheme that has been valued in the syntax.
Accordingly, complementizer agreement is a primarily morphological process which is
nevertheless sensitive to the syntactic environment. This account explains the restric-
tion on complementizer agreement observed in Bavarian comparatives if we assume
that the insertion of dissociated Agr-morphemes is a morphological operation that
applies after the deletion of the inected verb in comparatives (i.e., the X
0
-structure
formed by adjoining the verb to the T+Agr complex in (90)).
64
In Distributed Morphology, rules operating at MS are usually subject to locality
constraints. More specically, operations such as Impoverishment or Morphological
Merger may involve only structurally adjacent morphemes (cf. Halle & Marantz 1993
and Note 37 above). With respect to the insertion of dissociated Agr-morphemes, the
relevant condition must capture the intuition that the relation between the syntacti-
cally valued Agr-morpheme and its late-inserted copy is sufciently local. This can be
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.56 (3230-3294)
achieved by the following condition on the insertion of dissociated Agr-morphemes
(and the denition of structural adjacency in (93)).
65
(92) Insertion of dissociated Agr-morphemes
A dissociated Agr-morpheme can attach to a functional head X only if X is
structurally adjacent to another functional head Y hosting an Agr-morpheme
that has been valued in the syntax.
(93) Structural adjacency
A head X is structurally adjacent to a head Y iff
(i) X c-commands Y
(ii) There is no projecting syntactic head Z that
(a) is c-commanded by X and
(b) c-commands Y.
66
In other words, (93) guarantees that a head X is structurally adjacent to the head Y
of its complement. Hence, dissociated Agr-on-C can only be inserted if C is struc-
turally adjacent to T, the latter hosting a valued Agr-morpheme. It is now possible to
attribute the adjacency effects observed in Hellendoorn to the locality condition (92)
if we assume that scrambled XPs intervening between C and TP are not adjoined to
TP, but occupy the specier of a separate projection (Haeberli 2002: AgrSP, Grewen-
dorf 2004: TopP/FocP). Following proposals by Rizzi (1997) concerning the presence
of TopP/FocP in the left periphery, I assume that this projection is present only if it
serves to implement certain information-structural distinctions. Otherwise it is absent
from the structure. In the structure (94), then, the (optionally projected) TopP host-
ing the PP op den wrmsten dag vant joar disrupts structural adjacency between C and
T. This blocks the insertion of a dissociated Agr-morpheme on C. As a consequence,
complementizer agreement leads to ungrammaticality in contexts such as (94).
(94) *[
CP
datt-e
that-1pl
[
TopP
[op
on
den
the
wrmsten
warmest
dag
day
vant
of-the
joar]
year
[
TP
wiej
we
tegen
against
oonze
our
wil
will
ewrkt
worked
hebt]]]
have
How does this analysis account for the fact that in a certain set of dialects, we can
observe a similar adjacency effect in matrix clauses?
67
Recall that the presence of a
scrambled XP prevents the exponent of Agr-on-C from replacing the verbal agreement
ending in examples like (75b), repeated here as (95).
Hellendoorn
(95) Volgens
according-to
miej
me
loop-t/*lop-e
walk-1pl/walk-1pl
[
TopP
[op
on
den
the
wrmsten
warmest
dag
day
vant
of-the
joar]
year
[
TP
ook
also
wiej
we
noart
to-the
park]].
park
According to me we are also walking to the park on the warmest day of the
year.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.57 (3294-3372)
In examples like (95), the syntactically evaluated Agr-morpheme adjoined to T is
part of the complex C-head, due to V-to-T-to-C movement. So the question arises
of whether the T+Agr complex is structurally adjacent to C in the relevant adjunc-
tion structure (96). This would generally license the insertion of a dissociated Agr-
morpheme in V2 contexts, predicting that adjacency effects are absent in matrix
clauses, contrary to the facts.
(96)
T C
1
C
2
T Agr
T V
However, note that on standard assumptions, a category does not c-command ma-
terial adjoined to it (Kayne 1994: 16; Chomsky 1995: 339f.).
68
As a consequence, C
1
does not c-command T in (96). The adjunction structure in (96) thus fails to meet
the conditions for the insertion of a dissociated Agr-morpheme, since T is not struc-
turally adjacent to C. It follows that in V2 contexts, the realization of complementizer
agreement can only be licensed by structural adjacency (i.e., c-command) between the
two-segment category [C
2
, C
1
] and the (non-pronounced) lower copy of the T+Agr
complex, which is disrupted by the TopP hosting the scrambled XP in (95).
Interestingly, not all elements that intervene between C and an additional subject
(or rather, the TP) block the realization of complementizer agreement (in contrast to
claims by Carstens 2003: 401, Fn. 20: My understanding is that agreement of C with
a subject across any intervening material is impossible). In Bavarian, which exhibits
a similar adjacency requirement as Hellendoorn (as shown in (97)), modal particles
such as aber, halt, ja and clitic object pronouns may intervene between Agr-on-C and
TP/the subject (cf. e.g. Altmann 1984; Nbling 1992):
69
(97) *obwoi-st
although-2sg
woartscheints
probably
du
you
ins Kino
to-the movies
ganga
gone
bist
are
Although you probably went to the movies (Gnther Grewendorf, p.c.)
(98) dass-st
that-2sg
oaba
modprt
du
you
ibaroi
everywhere
dabei
with-it
bist
are
That you really are involved everywhere (Altmann 1984: 205)
(99) wia-sd-n
when-2sg-clit.3sg
du
you
gseng
seen
hoast
have-2sg
When you saw him (Pfalz 1918: 231)
Note that these examples do not represent counterexamples to the analysis proposed
above if we assume that the structural position of clitics and modal particles differs
from that of scrambled XPs, which move into a specier position of a TopP or FocP
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.58 (3372-3420)
intervening between C and TP. With respect to modal particles of Germanic, it is often
assumed that they are base-generated as adjuncts (here: TP-adjuncts) (cf. e.g. Abra-
ham 1995).
70
Accordingly, they do not require the projection of a separate TopP or
FocPand do not disrupt structural adjacency between Cand TP. Concerning the place-
ment of object clitics, I assume that their ultimate surface position is determined by
late MS-processes such as Prosodic Inversion, which apply at the mapping to PF (cf.
Bonet 1991; Halpern 1992; Schtze 1994). Therefore, they reach their surface position
after the insertion and valuation of dissociated Agr-morphemes has been completed.
Again, no interaction between these two processes is expected. In other words, the
correct generalization seems to be that only scrambled XPs which undergo syntac-
tic movement to a topic or focus position between C and TP block the realization of
complementizer agreement, while base-generated adjuncts (i.e., modal particles) and
object clitics that have undergone PF-movement do not disrupt the structural adja-
cency between C and T. This difference is expected on the analysis of complementizer
agreement presented here, but difcult to account for in a purely syntactic approach
to the phenomenon in question.
Similar adjacency effects can be observed in connection with so-called agreement
weakening in Standard Dutch (cf. e.g. Ackema & Neeleman 2003, 2004). As is well-
known, the 2sg agreement ending /-t/ is dropped (i.e., replaced by /-/) in inversion
contexts, resulting in a form homophonous to the 1st person singular:
(100) a. dat
that
jij
you
dagelijks
daily
met
with
een
a
hondje
doggy
over
over
straat
street
loop-t
walk-2sg
That you walk with a doggy in the street every day
b. Jij
you
loop-t
walk-2sg
dagelijks
daily
met
with
een
a
hondje
doggy
over
over
straat.
street
c. Dagelijks
daily
loop-
walk
jij
you
met
with
een
a
hondje
doggy
over
over
straat.
street
Similar to complementizer agreement, agreement weakening is blocked when a scram-
bled XP intervenes between the verb in C and the subject pronoun:
(101) *Volgens
according-to
mij
me
ga-
go
[op
on
de
the
heetste
hottest
dag
day
vant
of-the
jaar]
year
zelfs
even
jij
you
naar
to
het
the
park.
park
Ackema and Neeleman (2003, 2004) develop an account of Dutch agreement weak-
ening which is based on similar assumptions as their analysis of complementizer
agreement (see above). More specically, they propose an impoverishment rule which
operates at PF and deletes the verbs [Addressee] feature if the verb and the 2sg subject
pronoun are part of the same prosodic phrase (marked by braces):
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.59 (3420-3486)
(102) Dutch Agreement Weakening
{[V Prt Add] [D Prt Add]} {[V Add] [D Prt Add]}
After impoverishment, a less specied exponent may be inserted into the relevant ter-
minal node (/-/, which is specied for [Participant] only), resulting in the weakening
effect. The absence of agreement weakening in (101) is then attributed to the fact that
the nite verb and the subject are no longer in the same prosodic domain if a scram-
bled XP intervenes between these two elements. As a result, the impoverishment rule
in (102) cannot apply. Thus, according to Ackema and Neeleman, complementizer
agreement and agreement weakening are effected by two different PF rules which are
both sensitive to prosodic domains: complementizer agreement is attributed to a rule
(see (84) above) which accomplishes PF feature checking, while agreement weakening
is taken to be the outcome of a PF impoverishment rule.
However, note that the mechanism proposed in this section, that is, the insertion
of dissociated Agr-morphemes, makes available a single analysis of these supercially
unrelated phenomena. More specically, I assume that in Standard Dutch, a dissoci-
ated Agr-morpheme is added to the C-head in 2sg contexts. In other words, I claim
that Standard Dutch exhibits a form of complementizer agreement as well, despite ap-
pearances. By assumption, the exponent of the relevant Agr-morpheme is phonetically
empty (i.e., /-/), in contrast to varieties with overtly inected complementizers. As a
result, the presence of Agr-on-C cannot be detected in embedded clauses of Standard
Dutch. Still, its insertion does give rise to visible effects in the context of agreement
weakening, where the exponent of Agr-on-C, /-/, replaces the regular agreement
ending 2sg /-t/ (recall that in a given X
0
-complex, only the highest of two identical
Agr-morphemes is spelled out, see Note 62). Similar to other instances of comple-
mentizer agreement, the adjacency effect illustrated in (101) can then be attributed to
the fact that the scrambled XP disrupts structural adjacency between C and T, which
prevents Agr-on-C from being licensed.
To sum up, this section has argued that complementizer agreement in Germanic
should be analyzed as resulting from the post-syntactic insertion of a dissociated Agr-
morpheme, the feature content of which is identied under structural adjacency with
another Agr-morpheme that has been valued in the syntactic derivation. We thus have
to recognize the existence of a morphological mechanism giving rise to agreement phe-
nomena, in addition to the syntactic licensing of Agr-morphemes discussed in Section
3.6. It seems likely that this analysis of complementizer agreement in Germanic car-
ries over to other cases of multiple agreement, as for example in the following Swahili
sentence (cf. (12) above, repeated here for convenience):
(103) wa-toto
cl2-children
wa-li-kuwa
cl2-past-be
wa-me-ki-soma
cl2-perf-cl7-read
ki-tabu
cl7-book
The children had read the book. (Krifka 1995: 1416)
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.60 (3486-3553)
Similar to complementizer agreement in Germanic, the multiple occurrences of the
agreement marker wa- (signaling class 2 noun class agreement) can be analyzed as
resulting fromthe post-syntactic insertion of a dissociated Agr-morpheme, the content
of which is identied in relation to a single Agr-morpheme that has been valued in the
syntactic derivation under closest c-command with the DP wa-toto the children. At
this point, however, I will leave the exact specics of such an analysis open for future
research (see Chapters 46 for the role of dissociated Agr-morphemes and multiple
agreement in the grammaticalization of new agreement morphology).
. Conclusion
In this chapter, I have outlined key features of a formal model of predicate-argument
agreement. Needless to say, the exposition scratched only on the surface of the vast
amount of diverse and intricate agreement phenomena found across the worlds lan-
guages, leaving aside topics such as anti-agreement, A-agreement, context-sensitive
agreement, or the role of morphological processes such as Impoverishment and Mor-
phological Merger/Fusion in the realization of agreement, each of which would justify
a monograph in itself (cf. e.g. Bobaljik 1995 on the realization of verbal inection
in Germanic; Noyer 1997 for an in-depth morphological analysis of complex agree-
ment patterns found in a set of Afro-Asiatic and native American languages; Chung
1998 for A-agreement in Chamorro; Bruening 2001 for long-distance agreement and
A-agreement in Algonquian; Bejar 2003; Rezac 2004 for an Agree-based analysis of
a set of intricate agreement phenomena such as context-sensitive agreement or so-
called agreement displacement in languages like Georgian, Algonquian, or Basque).
Instead, I focused onthe basic ingredients of the agreement relation, that is, the phrase-
structural representation of agreement morphemes, the structural mechanisms that
serve to identify their feature content and the way agreement markers attach to the
verb stem.
In Section 3.2, I presented a variety of arguments which suggest that agreement
differs signicantly from substantial functional categories such as T, Neg or Asp,
which are commonly assumed to receive an interpretation at the LF interface. These
differences were taken to suggest that Agr does not head a separate projection in the
syntax, but rather adjoins to other, substantial functional heads. Section 3.3 dealt
with the structural relation under which the feature content of these parasitic Agr-
morphemes is identied. Here I pointed out that evidence from long-distance agree-
ment in Tsez suggests that the relevant structural conguration is closest c-command
(i.e., the Agree relation) and not spec-head agreement. In Section 3.4, I focused on the
question of how afx order is linked to syntactic structure, arguing that it is not pos-
sible to completely derive the order of inectional markers from hierarchical relations
created in the syntax, contra Julien (2002).
In Section 3.6, the theoretical insights reached in the previous sections were
worked out in some more detail. I proposed that agreement heads are inserted into
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.61 (3553-3607)
the syntactic derivation as adjuncts to substantial functional heads, with the normal
choices of T for subject agreement and for object agreement. In addition, agree-
ment morphemes may attach to other functional heads, which is one reason for the
multitude of ordering possibilities found across languages. Following the basic insight
reached in Section 3.3, I claimed that feature matching between the controllers -
set and the agreement morpheme is established during the syntactic derivation via an
Agree operation initiated by an Agr-head (or, rather, its non-valued -set). Under the
assumption that Agr-morphemes may attach to other functional heads apart from T
and , this analysis predicts that Case and agreement licensing can be dissociated, that
is, implemented by different functional heads. This prediction was shown to be borne
out by long-distance agreement in Tsez and intricate agreement patterns exhibited by
Georgian where the verb agrees with the external argument, regardless of whether the
latter carries nominative or ergative case. In addition, I argued that participle agree-
ment in French is also amenable to an analysis in which AgrO is dissociated from the
functional head that assigns accusative case. The observation that participle agreement
requires overt object movement was then attributed to the Phase Impenetrability Con-
dition, which forces movement of the object to the left edge of P, raising the object
into the Agree domain of the higher AgrO-morpheme.
Next, I addressed the question of how agreement markers are attached to the verb
stem, arguing that linear order is only partially determined by the hierarchical relations
established during the syntactic derivation. More specically, I adopted a more con-
servative approach, assuming that the ordering instructions sent to the articulatory
system are inuenced by lexical properties of individual Vocabulary items (i.e., the
prex/sufx distinction), in addition to structural restrictions commonly attributed
to the Mirror Principle. Concerning the latter, I argued that its effects can be derived
if head adjunction structures are processed in a bottom-up fashion during Vocabulary
Insertion. As I pointed out, the workings of the mapping procedure introduce another
source of variation with respect to the placement of agreement markers: due to the fact
that an agreement morpheme is part of the head domain of the functional head X it
adjoins to, the insertion procedure may target Agr either before or after X, giving rise
to different linear orderings of X and Agr relative to the verb stem.
In Section 3.7, I took a closer look at complementizer agreement in Germanic.
It was shown that the realization of complementizer agreement is sensitive to factors
such as adjacency and PF-deletion of the inected verb, which creates a problem for
purely syntactic approaches involving mechanisms such as head movement, spec-head
agreement or Agree. Instead, these facts were taken to suggest an alternative account
according to which complementizer agreement is instantiated post-syntactically (i.e.,
at MS) via the insertion of dissociated Agr-morphemes which adjoin to C. By as-
sumption, the identication of the feature content of these post-syntactically inserted
Agr-morphemes does not involve a direct structural relation with an interpretable
-set. Rather, feature matching with the subject is mediated under structural adja-
cency with another agreement morpheme that has been valued by a syntactic Agree
relation before.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.62 (3607-3650)
We thus arrive at a theory of syntactic agreement where feature matching be-
tween a controller and a target may result from two different underlying mecha-
nisms: rst, the surface realization of predicate-argument agreement may result from
agreement morphemes that are added to other functional heads before the resulting
complex head enters the syntactic derivation. The feature content of syntactic agree-
ment morphemes is then identied/valued via an Agree operation, which leads to
feature matching with an appropriate set of interpretable -features under closest c-
command. In addition, agreement morphemes may be added after the syntax as copies
of syntactically evaluated Agr-morphemes. The insertion of these so-called dissociated
Agr-morphemes typically leads to instances of multiple agreement where agreement
with a certain argument is realized in several places in a sentence. In the remainder
of this book, it is argued that this mechanism plays an important role in the gram-
maticalization of new agreement markers, which often involves an intermediate stage
of multiple agreement where the new agreement marker co-occurs with the old one
before the latter is eventually replaced by the former.
Note that apparently, the analysis developed in this chapter has not much to say
about a number of the typological generalizations about agreement noted in Chapter
1 such as the sufxing preference or the observation that many languages lack verbal
agreement markers for 3rd person arguments. Some of these questions will be taken up
again in the next chapters, where it will be argued that a set of synchronic properties
of predicate-argument agreement can be explained as a result of the way agreement
markers arise historically.
Notes
. Note that the discussion in this chapter focuses on approaches to verbal agreement that have
been proposed within the range of Principles and Parameters theory. As a consequence, alter-
native approaches which make use, for example, of feature unication and which are prevalent
in frameworks such as Generalized Phrase Structure Grammar (GPSG) or Head-driven Phrase
Structure Grammar (HPSG) will not be commented on.
. An additional SpecTP is only necessary for multiple subject constructions as for example
transitive expletive constructions (TECs) in Icelandic (see e.g. Chomsky 1995: Ch. 4; Bobaljik &
Jonas 1996 for discussion of TECs).
. Note that the Mirror Principle was originally devised to account for the ordering of deriva-
tional morphemes. However, its implications for inectional morphology were immediately
recognized after the split IP structure had been introduced.
. Of course, there are instances of different aspect and tense markers within a clause, as in
examples with a future perfect like Joe will have left the party by two oclock where each of the
different tense markers/heads contributes some clearly identiable piece of information to the
interpretation of the clause. Thus, in contrast to what is shown below for agreement, it is likely
that each tense marker corresponds to a discrete functional head (cf. e.g. Zeller 1994). However,
note that there are very few languages which show the phenomenon of tense agreement, where
one and the same tense formative appears in more than one place in a given sentence (Joachim
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.63 (3650-3716)
Sabel informed me that in Malagasy, for example, tense morphology shows up on the verb and
on certain adverbials).
. A related phenomenon can be observed in Finnish, where subject agreement is normally
realized as a verbal sufx which alternatively attaches to the negation.
(i) a. puhu-i-n.
speak-past-1sg
I spoke.
b. e-n
neg-1sg
puhu-nut.
speak-past
I did not speak. (Mitchell 1994: 124)
. Table 2 is a slightly modied representation of Juliens gures (her Table 5.2). Note that in-
tervening markers of other categories are ignored. Bound markers are indicated by the notation
X+Y. In addition, fused or portmanteau marking, where it is not possible to segment a given
inectional marker into separate tense and agreement markers, is indicated by a slash between
the relevant categories (e.g., AgrS/T). Importantly, note that the numbers given in Table 2 re-
fer to genera/families, not to individual languages. If the members of a genus/language family
uniformly show a certain combination of word order and afx order, this is counted as a single
instance of this combination. If a given language family is not uniform with respect to word
and afx order, every single combination found in one genus is listed in the table. Therefore,
the gures in the body of the table do not necessarily add up to the gures in the rightmost
column. For example, the second type of afx combination, AgrS/T+V is found in seven gen-
era. However, the gures given in the body of the table do not add up to seven, but to nine.
Accordingly, some of the seven genera that exhibit the afx order AgrS/T+V must contain more
than one type of word order. An empty cell indicates that no language in Juliens sample exhibits
the relevant combination of word and afx order. Note that Table 2 does not make reference to
split agreement marking. For example, in languages such as Turkana or Cree, person marking
employs prexes whereas number is marked by sufxes; in Turkana (Dimmendaal 1983), the
sequence of inectional markers is apparently AgrSPerson+T+V+AgrSNumber.
. Differences concerning the possibility of a certain order apparently only have to do with
the bound/non-bound character of the inectional markers. For example, there are 64 genera
that exhibit V+T+AgrS, which is the most frequent pattern in Juliens sample, but there is no
language that shows V+T AgrS, that is, the same sequence of markers with a free instead of a
bound postverbal subject agreement marker.
. Intervening markers of other categories are ignored. Again, the gures in Table 3 refer to
genera. Note that Table 3 indicates only the linear order of the inectional afxes. No distinction
is made between free and bound markers. The notationAgrS/O refers to portmanteau markers
that realize both subject and object agreement.
. Interestingly, Julien (p. 253) notes that the afx sequences expected on the basis of the clause
structure proposed by Chomsky (1991), (1993) (AgrSP-TP-AgrOP-VP) are not very frequent
cross-linguistically: AgrS-T-AgrO-V is found in only 4 genera and its sufxal mirror image V-
AgrO-T-AgrS is found in only 6 genera. In contrast, it appears that the tense marker preferably
intervenes between the verb and sufxed AgrS and AgrO markers (cf. V-T-AgrS-AgrO, 9 genera;
V-T-AgrO-AgrS, 9 genera, both mostly SOV). Furthermore, the often stated claim(cf. e.g. Speas
1991) that object agreement is closer to the verb stem than subject agreement appears to be only
a (mild) tendency across languages: in Juliens sample, AgrO is closer to the verb stem than AgrS
in 42 genera, while its the other way around in 26 genera (note that these numbers refer only to
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.64 (3716-3800)
languages without fused AgrS/AgrO markers and where AgrS and AgrO are located on the same
side relative to the verb stem).
. Haeberli (2000) develops a mixed approach that relates the presence of separate agreement
projections to the richness of verbal agreement morphology. More precisely, he suggests that
agreement features are contained in T but project a separate AgrSP (above TP) if (i) the verb
carries separate tense and agreement sufxes (cf. Thrinsson 1996; Bobaljik & Thrinsson 1998;
Bobaljik 2003), or (ii) if the agreement morphology signals more than a default/non-default
two-way distinction. If these morphological conditions are not met (e.g., as in English, which ex-
hibits only a two-way distinction where the default agreement ending is - and the non-default
is 3.sg.pres.indic /-z/), no separate AgrSP is projected and the agreement features contained in
T are checked by the subject, which raises to SpecTP for EPP reasons.
. Another piece of evidence which supports this view on agreement comes from research on
agrammatic speech production. Friedmann and Grodzinsky (2000) study the linguistic abilities
of a Hebrew speaking agrammatic aphasic. They show that in agrammatic speech production,
tense is impaired, but agreement is intact. F&G conclude that these functional categories must
be dissociated in the syntax. From the fact that properties associated with the C-system (e.g.,
the formation of wh-questions) are equally impaired, they further conclude that impairment
of T implies impairment of other functional nodes above T. To be able to capture the different
behavior of tense and agreement in agrammatic speech, F&G assume that TP dominates AgrSP
in Hebrew. However, this set of assumptions fails to account for the fact that in languages like
French, Italian or German agrammatic aphasics show impairment patterns that are similar to
those found in Hebrew. In other words, F&G would predict that agreement is impaired in a
language where AgrSP dominates TP. But this is obviously not the case. F&G speculate that these
data can be handled if agreement is checked in a mechanismdifferent fromthat for tense. It may
not have a node of itself, but it checks in one of the other checking points below T and is thus
preserved. (p. 99). These ndings point into the same direction as the other empirical evidence
considered so far: there is a fundamental phrase-structural difference between agreement and
core functional categories such as Tense.
. Note that it is not necessary that the agreement controller originates in the same clause
as the target. If no CP boundary intervenes between the matrix and the embedded clause, A-
movement may feed agreement. Relevant examples include Raising in English, illustrated in
(i) and so-called long passive in German (Hhle 1978; Haider 1986, 1993; Wurmbrand 2001),
shown in (ii).
(i) [The books]
i
seem [ t
i
to be on sale].
(ii) a. weil
since
er
he
[den
the
Stuhl
chair.acc
zu
to
reparieren]
repair
versuchte
tried
Since he tried to repair the chair
b. weil
since
der
the
Stuhl
chair.nom
zu
to
reparieren
repair
versucht
tried
wurde
was
Since they tried to repair the chair
c. weil
since
[der
the
Stuhl
chair
und
and
der
the
Tisch]
table.nom
zu
to
reparieren
repair
versucht
tried
wurden
were
Since they tried to repair the chair and the table
In (i), the thematic subject of the embedded non-nite clause has undergone A-movement into
the matrix clause, from where it controls (number) agreement on the matrix verb. This possibil-
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.65 (3800-3862)
ity is commonly attributed to the assumption that raising verbs either select a non-nite IP/TP
(cf. e.g. Chomsky 2000) or delete the CP layer of the non-nite complement clause (Chomsky
1981). In a similar vein, long passive is only possible with innitival complements of so-called
restructuring verbs such as versuchen try, vergessen forget or beginnen begin, which give rise
to clause union effects (see e.g. Bech 1955; Grewendorf 1987; Sabel 1996; Wurmbrand 2001). In
restructuring innitives, passivization of the matrix verb affects the case-assignment properties
of the lower verb. Accordingly, the embedded object undergoes A-movement into the matrix
clause for case assignment, triggering agreement on the matrix auxiliary.
. In both approaches, the domain of subject agreement extends to material contained in the
VP if there is no external argument (i.e., no P) present, that is, in passives and unaccusatives.
In a spec-head analysis, the object can undergo movement to SpecTP for case and agreement
if no external argument intervenes (in SpecP). In an Agree-type analysis, the absence of an
external argument implies the absence of a P phase boundary. As a consequence, an Agree
operation initiated by T can access elements contained in the VP. Note that this raises the inter-
esting question of whether material in an embedded SpecCP is accessible for subject agreement
with passive/unaccusative matrix verbs.
. Recall that Chomskys original motivation for the introduction of the Agree mechanism
partially comes from instances of long-distance agreement in Icelandic, where a matrix verb
may agree in number with a nominative DP that is contained in a lower clause (cf. Chapter 2
above).
. Polinsky (2003) distinguishes between three types of (apparent) long-distance agreement.
First, the impression of long-distance agreement might arise in a conguration where two dis-
tinct but coreferent arguments occur in the matrix and embedded clause, respectively. However,
if the (local) agreement controller in the matrix clause happens to be phonetically empty, this
may give the impression that the controller is in fact located in the embedded clause:
(i) He wants of (you)
i
[to throw the ball at you
i
]. (Polinsky 2003: 283)
According to Polinsky, this form of mediated locality is found in a number of Algonquian
languages such as Fox, Cree, Blackfoot, or Pasamaquoddy. Another instance of only appar-
ent long-distance agreement may result from raising or restructuring operations which bring
the controller into the agreement domain of the matrix clause, similar to long passive in Ger-
man. Again, this gives rise to an agreement congurationwhich respects the clause-boundedness
constraint (examples taken from Polinsky 2003: 283):
(ii) He wants you
i
[to be thrown the ball at t
i
]. (subject-to-object raising)
(iii) He wants-throw you the ball. (clause union/restructuring)
Apart from the German long passive examples discussed above, similar phenomena can be
observed in languages where subject-to-object raising/ECM feeds object agreement on the
matrix verb. Polinsky mentions Imbabura Quechua, Kipsigis (Nilo-Saharan), Tsakhur (Nakh-
Dagestanian), and Hindi. Finally, there is the case of genuine long-distance agreement as exem-
plied by Tsez, in which the controller moves into the left periphery of the embedded clause
from where it can trigger agreement on a higher target. See Polinsky (2003) for discussion and
a set of diagnostics that help to distinguish between apparent and genuine long-distance agree-
ment. It seems that all instances of long-distance agreement are linked to a topic interpretation
of the embedded controller. Furthermore, some languages such as Pasamaquoddy (Bruening
2001) exhibit mediated long-distance agreement alongside genuine long-distance agreement
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.66 (3862-3924)
:io The Rise of Agreement
(with an element presumably in the left periphery of the embedded clause). This might be taken
to indicate that historically, cases of genuine long-distance agreement developed from cases of
mediated locality via reanalysis.
:o. Note that only vowel-initial verbs carry (prexal) agreement markers. If the verb begins with
a consonant, no agreement markers are observed.
:. Similar intervention effects are created by phrases marked by the topic particles -n(o) or
-gon (Polinsky & Potsdam 2001: 636). Furthermore, the presence of a complementizer sufx on
the embedded verb (instead of a nominalizing sufx) also prevents long-distance agreement, see
Polinsky and Potsdam (2001: 635).
:8. Ede Zimmermann pointed out to me that the particular choice of quantied expressions in
(29) ( vs. ) actually does not justify any rm conclusions with respect to scope ambiguities,
since one constellation () implies the other (). Accordingly, does not necessarily con-
stitute a separate reading (in the sense of a real scope ambiguity), but may be understood as a
special case (of the constellation ), arising if the restriction of the existential quantier hap-
pens to contain only a single element in a given discourse environment. In other words, more
tests would actually be necessary to ascertain the presence of scope ambiguities in examples such
as (29).
:. For example, Simpsonand Bhattacharya (2003) assume that pied-piping of non-wh material
in wh-questions is licensed by wh-feature percolation, which identies constituents larger than
wh-phrases as eligible for wh-movement.
io. Bobaljik and Wurmbrand (2004) argue for a hybrid approach where agreement may result
either from Agree or from a spec-head relation created by A-movement. Data from German
involving fronted VPs that contain a nominative are taken to indicate that the Agree mecha-
nism is needed to account for the agreement and case facts found in this construction. On the
other hand, anti-reconstruction effects found in the German long passive (in particular, the
phenomenon of scope freezing) seem to require an analysis of case/agreement checking that
necessarily involves A-movement into a specier position of the case-assigning functional head,
feeding agreement with the higher predicate.
i:. To capture the gist of the Mirror Principle, it is sometimes assumed that the inectional
features on the verb come in a hierarchical order which reects the order of checking relations
(and therefore the hierarchy of functional projections). See Grewendorf (2002: 42) for some
discussion.
ii. As a result of this analysis, English must be analyzed as having overt verb movement, contra
standard assumptions. An alternative analysis in terms of Morphological Merger under adja-
cency (Halle & Marantz 1993; Bobaljik 1995) is not available, since this is predicted to lead to
prexal inection, contrary to the facts.
i. For this reason, Juliens account does not provide a principled explanation for the fact that
postposed inectional markers showa strong tendency for being bound, in contrast to preposed
markers (cf. Bybee et al. 1990 and Chapter 1 above).
i|. In addition, similar problems arise due to Juliens assumption that adverbs are merged as
speciers of functional heads (cf. Cinque 1999). Again, this conicts with roll-up movement
in SOV languages. Julien therefore assumes that merging of adverbs blocks further roll-up
movement. This assumption leads to a variety of completely different derivations dependent
on whether a given sentence contains certain adverbs or not. Moreover, it should not be possible
that adverbs are merged in speciers of functional heads that contribute to the verbal morphol-
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.67 (3924-3998)
Chapter 3. The structural design of agreement :i:
ogy; otherwise the presence of an adverb in the relevant specier would block sufxation of the
relevant functional head.
i,. Note that Julien herself assumes that the nite verb in English targets a position very low in
the IP domain, somewhere between a low Asp-head and Voice
0
(in the system of Cinque 1999).
io. In a similar vein, Bobaljik (2005) argues that the realization of agreement is established
by post-syntactic operations. More specically, he claims that morphological agreement does
not involve syntactic -feature checking/valuation, but is rather dependent on morphological
case, a kind of information which is not available in the narrow syntax. See Anderson (1992)
for related ideas, expressed in an alternative Late Insertion framework where inection is not
analyzed as the concatenation of discrete morphemes. Instead, complex words are the result of
Word Formation Rules that interpret feature complexes generated in the syntactic component.
i. This gist of this analysis is reminiscent of the interarboreal head movement proposed in
Bobaljik (1995), Bobaljik and Brown (1997), the only difference being that in the case of agree-
ment morphemes, the complex head structures which enter the ongoing syntactic derivation
are created by the operation Merge instead of Move. Note that in both instances, the creation
of complex heads proceeds in line with the Extension Condition (Chomsky 1995). However,
as pointed out to me by Ian Roberts, the idea that X
0
-complexes may result from the opera-
tion Merge raises some non-trivial issues within the theory of Bare Phrase Structure (Chomsky
1995).
i8. However, note that there is a slightly more complex alternative conceivable which perhaps
combines the advantages of both approaches. More specically, we may assume that agree-
ment features added to functional heads (e.g., via Agree) are subject to Fission at the level
of MS, creating a separate Agr-morpheme for the purposes of Vocabulary Insertion and lin-
earization. This analysis removes Agr-heads from the syntactic computation, but maintains the
possibility of separate Agr-morphemes at the level of MS. Furthermore, recall that Fission is a
language specic process, which does not apply generally. This captures the observation that
cross-linguistically, agreement marking often fuses with other inectional afxes. However, for
expository reasons, I will continue to represent Agr-morphemes as adjuncts to other functional
heads.
i. For the present purposes, it does not matter whether the object moves to an outer specier
of P after the subject has been merged or whether object movement takes place prior to the
insertion of the subject. Alternatively, the object may be tucked in in a lower specier after the
subject has been merged. See Chomsky (1995, 2000) and Richards (2001) for discussion.
o. This raises certain questions with respect to parametrization. It is doubtful that the deletion
of uninterpretable features may wait until Spell-out in general. Otherwise we would expect that
long-distance (object) agreement is a much more widespread phenomenon (contrary to the
facts). Therefore, one might speculate that individual languages may differ with respect to the
point at which features marked for deletion are ultimately deleted. I will leave that question open
for future research.
:. In the history of Georgian, formerly distinct accusative and dative case endings have fallen
together into a single morphological case ending -s which is commonly referred to as dative in
the literature on Georgian.
i. In Georgian, tense is marked by a combination of so-called preverbs (abbreviated as prev)
and fused Tense/Agr sufxes on the verb.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.68 (3998-4057)
:ii The Rise of Agreement
. Note that this illustration is somewhat simplifying. Actually, the agreement patterns exhib-
ited by Georgian are much more complex. In Georgian, the controller of a certain agreement slot
on the verb may shift between subject and object dependent on the -features of the arguments
in question, which Bejar (2003) calls context sensitive agreement. Cf. the following statement
taken from Bejar (2003: 4): a verb cross-references the person feature of its object, unless the
object is 3rd person, in which case the person feature of the subject is cross-referenced. Number
is preferentially that of the subject, unless the subject is singular, in which case the number of
the object is cross-referenced, unless the object is 3rd person.
|. I am indebted to Shin-Sook Kim for discussion and valuable suggestions concerning the
analysis presented here.
,. Recall that by assumption, a deletion-marked Case feature remains is accessible to the
syntactic derivation until the relevant DP is sent to the interfaces (in the case at hand, until
completion of the next higher phase, that is, CP).
o. Thus, in contrast to proposals by Chomsky (1993), the option that the inectional features
of verbs are checked via covert raising is not available.
. Marantz (1988: 261) denes the operation Morphological Merger as follows:
(i) At any level of syntactic analysis (d-structure, s-structure, phonological structure), a rela-
tion between X and Y may be replaced by (expressed by) the afxation of the lexical head
of X to the lexical head of Y.
It is generally assumed that (structural) adjacency at MS is a necessary condition for Merger
to take place (cf. Bobaljik 1994). Thus, in present-day English, T is afxed to VP-internal V at
MS if nothing intervenes between these head positions; the presence of the negation not triggers
do-support to pick up the inectional morphology associated with T (cf. Chomsky 1957; Halle
& Marantz 1993). Note that this analysis requires that VP-adjoined material such as adverbs
does not count for the evaluation of adjacency at MS (cf. Bobaljik 1994, 1995, 2002 for concrete
proposals).
8. Of course, these effects can only be detected if the relevant inectional markers are located
on the same side of the verb stem and if their order is not affected by other processes that take
place at MS.
. In addition, morphological operations such as Local Dislocation may reorder elements/heads
which are adjacent and part of the same constituent at MS. Local Dislocation differs from Mor-
phological Merger in that the latter takes place prior to linearization whereas the former applies
to structures that have already been linearized (see Embick & Noyer 2001 for discussion).
|o. One might speculate that the realization of word order (i.e., of exponents that constitute
phonological words) proceeds along similar lines, resulting from the same mechanism. Note
that this is actually expected in an approach such as DM which does not recognize a theoret-
ically signicant distinction between morphosyntactic features which are realized as bound or
free markers, cf. Harley and Noyer (1999: 7): Features which will eventually be realized as a
subpart of a phonological word are treated no differently from features which will eventually be
realized as an autonomous word. That is, similar to the realization of afx sequences and the
prex/sufx distinction, word order (e.g., the VO/OV distinction) are presumably determined
by a combination of lexical properties of certain Vocabulary items and hierarchical structures
generated by the syntactic component, resulting in ordering instructions for PF. Note that the
existence of an additional projection level that contains the specier represents a complication
which does not exist for the computation of afx order in head adjunction structures. Since this
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.69 (4057-4129)
work is primarily concerned with the linear ordering of agreement markers, I will not further
address the realization of word order here (cf. Bobaljik 2002: 213ff. for a concrete proposal).
. Alternatively, it is conceivable that afxes select not only the direction where their host must
appear, but also its category. Then, the marked sequences (57cd) might follow from the fact
that cross-linguistically, T preferably attaches to Asp.
. A question arises with respect to lexical roots that are adjoined to other lexical roots, as for
example in the case of morphological causatives (cf. e.g. Baker 1988). The model developed here
predicts that in principle, insertion may either target the lower or the higher root rst. As far as
I can see, however, this creates no problems in the present approach, since it only affects the
order of the two lexical roots, while higher functional heads are excluded by the V
0
+V
0
adjunc-
tion structure. Accordingly, the latter must be realized outside of the V
0
+V
0
complex, which
captures the generalization that inectional morphology must appear outside of derivational
morphology changing grammatical functions (Baker 1988; Speas 1991).
. Liliane Haegeman (personal communication) informed me that there is a strong preference
to have clitic doubling in all person/number combinations, that is, the clitic is almost obliga-
tory (if the full subject is a pronoun; otherwise no clitics are found). However, the clitic can
be dropped somewhat more easily in 3sg.fem/neut, 1pl and 3pl contexts. According to Liliane
Haegeman, the deletion of the clitic is presumably governed by phonological factors, in the
sense that the clitic is always obligatorily present in the syntax, but can be deleted at PF if certain
phonological conditions are met. Asimilar proposal is made by Shlonsky (p. 352f., fn. 4). In light
of related phenomena discussed in Chapters 5 and 6, it seems to me that the (obligatory) subject
clitics should be analyzed as additional agreement morphemes on C, the grammaticalization of
which led to repair the previously highly syncretic agreement paradigm.
. To the best of my knowledge, these restrictions to certain person/number combinations are
still left unexplained. Hoekstra and Smits (1999) claim that the distribution of complementizer
agreement is governed by the following generalization:
(i) The Identity Generalization
Complementizer agreement only occurs when the agreement ending of the inverted aux-
iliary in the present tense is identical to the agreement ending of the inverted auxiliary in
the preterite.
Note, however, that even if (i) turns out to be correct, it is only a description of the distributional
facts, and not an explanation. In Chapter 6 it is argued that at least in the case of Bavarian,
the person/number restrictions ultimately receive a diachronic explanation. More precisely, it is
shown that Agr-on-C is a residue of the grammaticalizationof new verbal agreement endings in
the history of Bavarian, which affected only 2nd person forms (plus 1pl in some dialects).
. Zwart (1993b) claims that a similar difference can be observed for West Flemish. However,
Liliane Haegeman pointed out to me that the endings are in fact morphologically identical and
that the differences in question are most likely the result of phonological rules.
. Similar facts can be observed in a subset of Lower Bavarian dialects, where complementizer
agreement is also found with 1pl subjects. See Chapter 5 for details.
. Note that this phenomenon provides further support for the hypothesis that syntactic ter-
minals are supplied with phonological features after syntax (cf. Zwart 1997b). If the verb enters
the derivation in fully inected form (i.e., with phonological features), then the contextually
determined change from speult to speule constitutes a problem that cannot be circumvented
by base generating the verb as speule in (66b): due to the fact that the verb is rst accessed by
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.70 (4129-4199)
the functional head(s) T/Agr, it is expected that this checking relation requires the form speult,
which shows up when the verb is not moved to C. However, the change from speult to speule is
easily captured in a Late Insertion model in which the morphological component interprets the
representation generated by the syntactic component, assigning the head cluster V +T/Agr+C a
spell-out that differs from the spell-out of V +T/Agr.
. Frisian shows another restriction on complementizer agreement when the subordinate
clause is embedded under a verb that optionally selects a V2 complement (commonly referred
to as a bridge verb). Note that similar to the Scandinavian languages (but unlike German or
Dutch), the sentential complements including V2 clauses of this set of verbs are always
headed by a complementizer in Frisian. In these contexts, complementizer agreement is oblig-
atory when the verb stays behind in nal position, but excluded when the V2 option is chosen
(de Haan & Weerman 1986; Zwart 1993b):
(i) Heit
dad
sei
said
datst-o/*dat
that-2sg=you/that
do
you
soks
that
net
not
leauwe
believe
moast.
must-2sg
Dad said that you should not believe such things.
(ii) Heit
dad
sei
said
dat
that
do/*datst-o
you/that-2sg=you
moast
must-2sg
soks
that
net
not
leauwe.
believe
. See den Besten (1982) for an early account of complementizer agreement in terms of a rule
Move Tense and Bennis and Haegeman (1984) on West Flemish data. In an early analysis of
complementizer agreement in Bavarian, Bayer (1984) develops an account that is based on the
idea that in V2 languages, there is an abstract agreement relation between Comp, V/In and
the subject leading to co-indexation of all three elements. In the case of 2nd person subjects
in Bavarian, this form of agreement is overtly realized due to a linking rule that copies the -
features located in In to Comp. This overt manifestation of agreement serves to identify the
referential content of the subject DP, thereby licensing an empty pronominal pro in the subject
position.
. Slightly modifying Zwarts approach, Watanabe (2000) assumes that both T and V carry a
separate set of -features to guarantee the presence of agreement morphology on the verb.
. Note that Zwarts account of multiple agreement on C and V raises of number of conceptual
questions. First, the claimthat a single functional head can carry multiple sets of agreement fea-
tures is by no means a standard assumption and raises a number of questions. For example, it
is well-known that the paradigms of complementizer agreement are often defective (see above).
In contrast, verbal agreement paradigms tend to be much less defective (apart from a certain
amount of syncretism). This asymmetry between the paradigms of verbal agreement and com-
plementizer agreement is rather unexpected if both types of agreement features are represented
by the same functional head. Furthermore, in the theoretical model adopted by Zwart (the form
of Minimalism envisaged in Chomsky 1993, 1995), it is generally assumed that feature move-
ment must affect the whole set of formal features contained in a given functional head. In other
words, it should not be possible to move only a subset of the formal features contained in a given
functional head, in contrast to what is assumed by Zwart.
. This adjacency effect is also unexpected under the analysis proposed by Watanabe (2000),
who assumes that complementizer agreement results from a two-step operation in which the
subjects -features are rst copied onto T(as a result of Agree) and then accompany head move-
ment of the T
0
complex to C. Again, this analysis should lead us to expect that the verb always
carries complementizer agreement in V2 clauses, contrary to the facts.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.71 (4199-4287)
Chapter 3. The structural design of agreement :i,
,. A related adjacency effect can be observed in West Flemish:
(i) *da-t
that-3sg
morgen
tomorrow
zie
she
werk-t
work-3sg
That she is working tomorrow (Shlonsky 1994: 360)
However, note that West Flemish generally requires strict adjacency between C and the subject
(i.e., (i) is ruled out anyway, cf. Haeberli 2002, 2004). Thus, examples like (i) actually do not
allowany rmconclusions concerning the sensitivity of complementizer agreement to adjacency
effects in West Flemish.
,|. For cases where there is only a subject clitic present (cf. (64b) above), Shlonsky assumes that
SpecIP is lled by pro which is governed and Case-licensed by AgrC.
,,. In addition, the assumption that SpecAgrCP is an A-position raises a number of further
conceptual issues. First, there is no independent motivation for moving the full subject (in the
absence of a clitic) to this position apart from the licensing needs of AgrC (i.e., the realization
of agreement on the complementizer). Second, this operation conicts with the generalization
that a Case-marked NP cannot undergo further A-movement (see e.g. Chomsky 1981, 2000).
,o. Note that this mechanism is somewhat reminiscent of the idea of Bayer (1984) that there
exists an abstract agreement relation between Comp, V/In and the subject.
,. Ackema and Neeleman (2004: 235) stress that PF feature checking requires post-head adja-
cency: a phrase BP whose features are to enter into a checking relation must immediately follow
a head A that contains identical features [...] If another maximal projection intervenes ([
AP
A
XP BP]), or if the phrase precedes the head ([
AP
BP A]), the two will not be part of the same
[prosodic phrase], so that prosodic checking is impossible.
,8. Furthermore, the fact that complementizer agreement licenses pro-drop in varieties such
as Bavarian and Frisian creates another serious problem for Ackema and Neelemans approach.
More specically, Ackema and Neeleman would have to assume that an empty subject, which is
otherwise invisible at PF, is visible for the purposes of PF feature checking.
,. Note that similar facts hold for 1pl in a set of Lower Bavarian dialects (see Chapter 5 for
some discussion), cf.
(i) De
they
san
are
gscheider
more-intelligent
[(als)
than
wia-ma
as-1pl
mir
we
san].
are
They are more intelligent than we are.
(ii) *De
they
san
are
gscheider
more-intelligent
[(als)
than
wia-ma
as-1pl
mir].
we
(iii) De
they
san
are
gscheider
more-intelligent
[(als)
than
wia
as
mir].
we
(Bayer 1984: 271)
oo. Bresnan (1973) argues convincingly that comparatives are to be analyzed as the result of a
(PF) deletion rule. Compare the following pairs of sentences:
(i) a. Ive never seen a man taller than my father.
b. Ive never seen a taller man than my father.
(ii) a. Ive never seen a man taller than my mother.
b. Ive never seen a taller man than my mother.
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.72 (4287-4355)
While (ia) and (ib) can receive a more or less identical interpretation, (iia) and (iib) cannot.
More specically, the reading of (iib) is deviant, implying that the NP my mother refers to a
man. Bresnan shows that this puzzling fact receives an elegant explanation if it is assumed that
part of the than clause is deleted under identity with the head of the comparative. The difference
in interpretation between (iia) and (iib) can then be attributed to the different shape of the
antecedent (i.e., the head of the comparative) of the elided constituent. In (ia) and (iia), the
antecedent of the deleted constituent consists of a predicative AP, as illustrated in (iii) (note
that the special interpretation of comparative forms is commonly attributed to the presence
of a degree phrase containing the -er sufx, labeled QP here. The than clause is then assumed
to contain an identical degree phrase which encodes the degree commonly labeled x of a
property ascribed to the subject):
(iii) ... [
AP
[
QP
-er] tall][
CP
than [
IP
my father/mother [
I
is [
AP
[
QP
x ] tall]]]]]
Accordingly, what is compared is how tall the subject of the than clause is. In contrast, the
antecedent of the elided constituent in (ib), (iib) consists of a predicative NP:
(iv) ... [
NP
[
AP
[
QP
-er] tall] a man][
CP
than[
IP
my father/mother[
I
is
[
NP
[
AP
[
QP
x] tall] a man]]]]]
In this case, we compare how tall a man the subject of the than clause is. As a result, we get a
deviant reading with my mother as the subject of the than clause.
. From the fact that dissociated morphemes are inserted after Spell-out, it follows that they
are not interpreted at LF.
. Above we have already noted that C-related agreement marking cannot be realized in ad-
dition to canonical verbal agreement in contexts where the verb has undergone movement to
C. In other words, Agr-on-C cannot be realized in addition to Agr-on-T. Why is this form of
double agreement (i.e., *V+AgrT+AgrC) ruled out? Note that this question relates to all Ger-
manic varieties with complementizer agreement. Following Carstens (2003), I assume that the
impossibility of doubly inected nite verbs results from a morphological condition ensuring
that only the hierarchically highest Agr-morpheme is spelled out in a given head complex. Based
on proposals of Kinyalolo (1991), Carstens (2003: 407) phrases the relevant condition as follows
(where inert means that the relevant Agr-morpheme is not pronounced):
(i) Morphological Economy
In an adjoined structure, Agr on a lower head is inert iff its features are predictable from
Agr on a higher head.
If it is assumed that complex heads are processed in a bottom-up fashion during Vocabulary
Insertion (see Section 3.6.3 above), (i) can be implemented by a late MS-process which deletes
the lower Agr-morpheme in a head complex.
. The idea that complementizer agreement is parasitic on verbal agreement is further sup-
ported by the observation that across Germanic, there appear to exist no languages with comple-
mentizer agreement but without verbal agreement, while there are many languages that exhibit
verbal agreement in the absence of complementizer agreement (Hoekstra & Smits 1999). Thus,
it seems that cross-linguistically, the availability of complementizer agreement is dependent on
the overt realization of verbal agreement morphology.
. At rst sight, it seems that this analysis is contradicted by those varieties (like Hellendoorn,
or 1pl in Lower Bavarian) where the shape of Agr-on-C differs from that of Agr-on-T. However,
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.73 (4355-4414)
this phonological difference can be explained as the result of conditioned allomorphy which is
sensitive to the syntactic environment (i.e., the adjunction site of the Agr-morpheme, C vs. T)
or as pointed out to me by Ian Roberts the way the relevant agreement morpheme is valued
(via Agree/head movement or post-syntactic insertion as a dissociated morpheme). This form
of allomorphy is presumably introduced by the reanalysis of a subject clitic as a dissociated Agr-
morpheme. As a result, the phonological exponent of the latter will in most cases differ from the
existing verbal morphology due to its lexical source (the subject clitic), see Chapters 5 and 6 for
details.
. A similar theoretical notion of structural adjacency is proposed in Zeller (2001: 36). How-
ever, in contrast to the analysis proposed here, Zeller assumes that structural adjacency consti-
tutes a local domain between two heads that is only syntactically relevant.
. If clitics (and, possibly, proper names) are analyzed as D-heads, the restriction to projecting
syntactic heads is necessary to warrant that a clitic in SpecTP (in the conguration [
CP
C [
TP
D [
T
T ]]]) does not interrupt the structural adjacency between C and T (Ian Roberts, personal
communication).
. Recall that in some of these dialects, the realization of complementizer agreement is con-
ned to inversion contexts. This is illustrated by the following East Netherlandic examples ((65),
repeated here as (i)), in which the verb carries the exponent of Agr-on-C (/-6/) only in inversion
contexts (i.e., (ib)), while subject-initial clauses exhibit the regular agreement ending /-t/, cf.
(ia):
(i) a. Wij
we
speul-t/*-e.
play-1pl
We play.
b. Waar
where
speul-e/*-t
play-1pl
wij?
we
To account for this difference, I tentatively follow Travis (1984) and Zwart (1997a) and as-
sume that the dialects in question exhibit a structural difference between inversion contexts
and subject-initial clauses, in the sense that only the former involve movement to the C-system.
In contrast, subject-initial clauses result from a conguration where subject and verb occupy
specier and head of TP (sometimes called asymmetric V2), giving rise to regular agreement
morphology on the verb. Furthermore, note that not all dialects with different exponents for
Agr-on-C and Agr-on-T exhibit this contrast in matrix clauses. In Lower Bavarian, for example,
the exponent of Agr-on-C (1pl /-ma/) replaces the regular agreement ending in subject-initial
clauses as well (Wei 2002: 9). This can then be taken to indicate that both subject-initial clauses
and inversion structures involve movement to the C domain:
(ii) a. Mia
we
laff-ma/*laff-a
ran-1pl/ran-1pl
hoam.
home
We are running home.
b. Mia
we
gem-ma/*geng-an
go-1pl/go-1pl
hoam.
home
We are going home.
. Cf. the following denition of c-command given in Chomsky (1995: 339). X and Y are taken
to be disconnected if no segment of X contains Y:
JB[v.20020404] Prn:2/08/2005; 9:49 F: LA8103.tex / p.74 (4414-4448)
(i) C-command
X c-commands Y if (a) every Z that dominates X dominates Y and (b) X and Y are
disconnected.
. Similarly, object clitics may intervene between the subject and the inected complementizer
in West Flemish, which otherwise requires strict adjacency between C
0
and the subject (Liliane
Haegeman, personal communication):
(i) da-n
that-3pl
ze
her
Valre
Valre
en
and
Marie
Marie
nie
niet
gezien
seen
een
have-3pl
That Valerie and Marie have not seen her
This parallel between West Flemish and Bavarian suggests that the strict adjacency effect ex-
hibited by West Flemish (see Note 53) might possibly be attributed to the existence of a full
paradigm of complementizer agreement. As a result, C and T must be structurally adjacent in
all contexts.
. At rst sight, the shape of many modal particles seems to suggest an analysis as X
0
elements.
However, the fact these elements do not block V-to-C movement in the Germanic V2 languages
shows that they cannot be analyzed as syntactic heads. In addition, similar locality problems
arise for XP-movement into the C-domain if they are analyzed as speciers (cf. van Gelderen
2001 for discussion and further diachronic arguments against an analysis as speciers). These
facts can be taken to further support the idea that modal particles are base-generated in adjoined
positions.
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.1 (45-130)
Chapter 4
The transition from pronoun
to inectional marker
. Introduction
As already mentioned, it is commonly assumed that pronouns are the primary histori-
cal source of subject-verb agreement morphology. However, the discussion in Chapter
1 has shown that our knowledge of syntactic aspects of this historical process is still
rather limited. This chapter sets out to develop a theory of the syntactic environments
where new verbal agreement markers can evolve via a reanalysis of subject clitics. In
contrast to beliefs widespread in the literature, it is argued that this structural change is
not conned to a unique syntactic environment. Rather, there is quite a variety of syn-
tactic contexts that may feed the grammaticalization process in question. It is shown
that the set of historical scenarios which may license a reanalysis of subject pronouns
as agreement markers is dened by a small number of (necessary) syntactic conditions
that the change in question must satisfy. These conditions derive from the theory of
agreement developed in the previous chapter.
Of course, a formal analysis of the transition from clitic pronouns to agreement
markers must incorporate some diagnostics for telling apart clitics and agreement
markers. Accordingly, the present chapter discusses a set of syntactic and morpho-
logical diagnostics to decide whether a given element has already developed into some
form of agreement or should rather be analyzed as a clitic pronoun. Importantly, it is
argued that these criteria not only provide a tool for linguistic analysis, but also can be
conceived as cues for the learner during language acquisition.
The chapter is organized as follows. Section 4.2 introduces a set of criteria for
differentiating pronominal clitics and agreement markers. Section 4.2.1 is concerned
with syntactic criteria, while Section 4.2.2 deals with morphological criteria (basically
following Zwicky & Pullum 1983). In Section 4.3, I discuss a set of syntactic precon-
ditions on the categorial reanalysis of pronominal clitics. Based on these conditions, I
then introduce a set of historical scenarios where subject clitics can be reanalyzed as
markers of verbal agreement. The presentation of these scenarios is organized along
the common distinction between In-oriented (Section 4.4) and C-oriented clitics
(Section 4.5). Section 4.6 summarizes the conclusions reached in this chapter.
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.2 (130-182)
. Telling apart clitics and agreement markers
As noted in Chapter 1, the historical development of (bound) agreement markers nec-
essarily proceeds via a stage where clitics are reanalyzed as inectional afxes. For this
reanalysis to take place, the properties of clitics and agreement markers must be suf-
ciently similar so that the learner can misinterpret the former as the latter. However,
due to these similarities it is an equally difcult task for the linguist to decide whether a
given clitic has already developed into an agreement marker, which is a necessary pre-
condition for an adequate analysis of the diachronic development in question. Hence,
both the learner and the linguist must look for cues that decide on the categorial status
of the clitic in the data at issue. This section sets out to develop a set of relevant di-
agnostics to assess the categorial status of clitics/agreement markers. In Section 4.2.1,
I discuss a set of syntactic diagnostics while Section 4.2.2 focuses on morphological
criteria that separate clitics from agreement markers.
.. Syntactic criteria
Perhaps the most common syntactic criterion for distinguishing between clitics (and
incorporated pronouns) and agreement markers is complementary distribution with
a DP argument (cf. e.g. Corbett 2003 for some discussion): as a rule of thumb, genuine
agreement markers may co-occur with a DP argument, while clitics and incorporated
pronouns are arguments by themselves and may therefore not co-occur with an argu-
ment that receives the same -role. The way this diagnostic works can be illustrated
with the following examples from Italian and the Carib language Macushi (examples
taken from Corbett 2003: 186). In both Italian and Macushi, it is not necessary that the
arguments of the verb (in Italian, this holds only for subjects) are realized overtly. The
inected verb alone is sufcient to create a well-formed sentence:
1
Italian
(1) Mangi-a.
eat-3sg
He/she eats.
Macushi
(2) i-koneka-p-i-ya.
3sg-make-past-3sg-erg
He made it.
In Italian, the -features of the subject are expressed by the morpheme -a. In Ma-
cushi, the -features of the absolutive argument are expressed by the prex i-, while
the sufx -i marks the feature content of the ergative argument. Thus, at rst sight, the
person/number markers that appear on the nite verb in Italian and Macushi seem to
behave similarly. However, the languages behave differently when overt arguments are
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.3 (182-239)
Chapter 4. The transition from pronoun to inectional marker
added. In Italian, the subject can be optionally realized as an overt DP (e.g., an overt
subject pronoun), as in (3).
Italian
(3) Lui/Lei
He/she
mangi-a.
eat-3sg
He/she eats.
This is commonly taken to indicate that in Italian, the person/number marker is a true
agreement sufx, which redundantly marks the person/number features of the subject,
which may be pro (as e.g. in (1)). Now consider the following examples from Macushi.
Macushi
(4) t-ekn
re-pet.abs
erama-p
see-past
paaka
cow
esa-ya.
owner-erg
The owner of the cow saw his own pet.
Macushi
(5) *uur-ya
I-erg
i-koneka-p-i-ya
3sg-make-past-3sg-erg
He made it.
(4) shows that the person/number markers disappear in the presence of overt DP ar-
guments. Moreover, (5) shows that the co-occurrence of an overt DP argument and
the corresponding person/number marker leads to ungrammaticality. The comple-
mentary distribution of person/number markers and overt DP arguments shows quite
conclusively that the person/number markers in (2) are not instances of agreement,
but rather clitic pronouns that have argument status and receive a -role from the verb.
In this way, the distribution of person/number markers and overt DP arguments pro-
vides a rst rationale for deciding whether a given element is a clitic or an agreement
marker. Unfortunately, the picture is often less clear cut than in the Italian/Macushi
case. On the one hand, there are languages such as Welsh, in which the presence of
full-edged person/number markers is compatible only with pronominal subjects (cf.
Chapter 1 above). On the other hand, there are many languages where clitics generally
may co-occur with another DP argument, giving rise to instances of clitic doubling, as
in the following examples from French (subject clitic doubling), Spanish (object clitic
doubling), and Italian (clitic left dislocation) (the clitic and the double are marked by
underlining):
French
(6) Pierre
Pierre
il
he
mange
eats
une
an
pomme.
apple
As for Pierre, he eats an apple.
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.4 (239-319)
Spanish
(7) La
her
llamaron
call-3pl
a ella.
to her
They called her. (Suer 1988: 394)
Italian
(8) Credo
think-1sg
che
that
il tuo libro,
your book
loro
they
lo
it
apprezzerebbero
appreciate-subjnc
molto.
much
I think that they would appreciate your book very much. (Rizzi 1997: 288)
In general, instances of clitic doubling resemble the agreement relation in that there is a
free phrasal element and a bound element which share -features. This abets a reanal-
ysis in which the clitic is misinterpreted as a mere agreement marker, with the double
turning into the true argument.
2
Accordingly, instances of doubling are particularly
interesting in the context of this study. However, due to the similarities between dou-
bling and agreement congurations, it is not an easy task to decide whether the clitic
has already developed into an agreement marker. Hence, we must look for further cri-
teria that can be used to decide on the categorial status of the clitic in clitic doubling
structures. In principle, these criteria may relate to (i) properties of the double, (ii)
properties of the clitic, or (iii) the distribution of the doubling construction as a whole.
Let me rst consider some factors which relate to the distribution of the doubling
construction as a whole. Recall that it is a characteristic of agreement that it is always
obligatory, that is, the verb always agrees with the subject, independent of the con-
text. In contrast, the doubling construction is often linked to certain stylistic purposes
and therefore conned to certain contexts. For example, the speaker may choose to
add a full DP for reasons of emphasis, to reinforce the prosodically weak clitic, which
cannot bear stress, similar to examples where pro is replaced by an overt pronoun
in languages such as Italian. As a consequence, the doubling construction is charac-
teristically restricted to certain contexts where the speaker wishes to emphasize the
argument expressed by the clitic. An example of this usage of the doubling construc-
tion is Standard French, where the clitic alone sufces to express the subject. Crucially,
a full DP/pronoun can optionally be added to lay stress on the subject, cf.
(9) (Moi)
me
je
clit.1sg
porte
carry
la
the
table.
table
As for me, I carry the table.
Therefore, doubling is an optional mechanism that is restricted to contexts where it
serves certain communicative functions in French. A similar pattern can be observed
in Standard Italian, where clitic doubling is obligatory only to resume a topic which has
been moved into the left clausal periphery, as in (8) above. These contextual restric-
tions clearly show that the clitic element is a true pronominal clitic that has not yet
developed into an agreement marker. However, if doubling loses its stylistic force (pre-
sumably due to an over-use), it may gain a wider distribution and eventually become
obligatory in all contexts. This can then be taken to indicate that the doubling congu-
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.5 (319-375)
ration has undergone a grammaticalization process and that the clitic should actually
be analyzed as an agreement marker. Well-known examples of this diachronic develop-
ment are Non-Standard French, where subject clitic doubling is obligatory for 1sg, 2sg
and 1pl forms (Lambrecht 1981; Auger 1993, 1994; Gerlach 2002), and Spanish, which
exhibits obligatory object clitic doubling (cf. e.g. Suer 1988). Accordingly, I will use
the distribution of clitic doubling as a diagnostic for the categorial status of the clitic:
optional doubling signals that the clitic is still a pronominal element, while obligatory
doubling indicates that the clitic probably has been reanalyzed as an agreement marker
(see Chapter 5 on a set of Swiss Rhaeto-Romance dialects).
3
Another set of potential diagnostics has to do with properties of the dou-
ble. Uriagereka (1995) notes that clitics usually receive an interpretation as de-
nite/specic. Likewise, in clitic doubling constructions, the full nominal usually must
be denite/specic. This observation can be turned into a diagnostic for agreement
in cases where we have to decide whether a certain conguration is an instance of
clitic doubling or predicate-argument agreement: if it is possible for the full nomi-
nal to be indenite/non-specic, then it is rather clear that the relevant construction
represents an agreement relation (cf. Brandi & Cordin 1989; Suer 1992; Haiman
1991; Haiman &Beninc 1992 for similar considerations). For example, indenite NPs
such as nobody cannot occur as topics in clitic left dislocation structures in Standard
Italian, cf.
(10) *Nessuno,
nobody
lo
him
conosco
know-1pl
in
in
questa
this
citt.
city
*Nobody, I know him in this city. (Brandi & Cordin 1989: 118)
However, in a set of Northern Italian dialects this restriction does not hold any longer.
The following examples show that indenites such as no one can co-occur with the
clitic subject pronoun. This is commonly taken to suggest that the clitic is actually an
agreement marker (Brandi & Cordin 1989: 118):
(11) a. Nisun
nobody
lha
he-has
dit
said
niente.
nothing
(Trentino)
Nobody said anything.
b. Nessun
nobody
glha
he-has
detto
said
nulla.
nothing
(Fiorentino)
Nobody said anything.
Finally, we may look for properties of the clitic itself that give us a clue concerning its
categorial status. A set of relevant syntactic criteria is discussed in Rizzi (1986), Brandi
and Cordin (1989), Haiman (1991), Haiman and Beninc (1992), Suer (1992), and
Poletto (2000). First, it is sometimes suggested that anti-agreement or partial agree-
ment are properties which are found solely with real agreement markers (e.g., in
a number of Northern Italian dialects, Irish and Standard Arabic), but never with
pronominal elements, which always show complete -feature identity with their an-
tecedent (cf. Haiman & Beninc 1992). Under this assumption, anti-agreement be-
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.6 (375-451)
tween the clitic and its double can be taken to indicate that the clitic should be analyzed
as an instance of agreement marking, as in the following examples from a variety of
Northern Italian/Rhaeto-Romance dialects (takenfrom Haiman &Beninc 1992: 193):
(12) a. L
he
e
is
venju
come-masc.sg
la
the
vivano.
witch.fem
(Fassa)
There came the witch.
b. Chi
who
e
is
lo
he
po
then
i
the
ozitegn?
Occitans-masc.pl
(Moena)
Who are the Occitans?
c. Da
of
doman
morning
vegn
becomes
l
he
oshore
fed-masc.sg
les
the
vatges.
cows-fem.pl
(Badiot)
The cows are fed in the morning.
d. Gl
he
e
is
venuto
come-masc.sg
delle
some
ragazze.
girls-fem.pl
(Fiorentino)
There came some girls.
e. U
he
vene
comes
a
the
Katajning.
Catherine
(Zeneyze)
Catherine is coming.
f. Al
he
era
was
una
one
volta
time
una
a
fameja.
family-fem.sg
(Furlan)
There was once a family.
Another criterion relates to contexts with a subject gap, that is, subject relatives and
conjoined sentences, where a coreferential subject is often deleted in the second con-
junct. If the subject clitic is preserved in these contexts, so the argument goes, then it
should rather be treated as an agreement marker, since agreement markers are obliga-
torily present in relative clauses and conjunction contexts. Again, the diagnostic can
be illustrated with data from Northern Italian dialects (examples taken from Haiman
1991: 141):
4
Relative clauses
(13) a. dut
all
ce
that
che
which
al
he
e
is
gno
mine
(Friulian)
All that which is mine
b. doj
two
peres
poor
pitshuj
little-boys
ke
who
i
they
me
me
prea
ask
l
the
arberk
shelter
(Fassa)
Two poor little boys who ask me for shelter
Conjoined clauses
(14) a. Al
he
vent
sells
dut
everything
e
and
al
he
va
goes
lontan.
far away.
(Friulian)
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.7 (451-508)
b. Quant
when
ch al
that-he
vigni
came
a
to
se
himself
e
and
al
he
cognosse
recognized
(Friulian)
l
the
predi.
priest
(Haiman & Beninc 1992: 194)
In other words, if we come upon a subject clitic in circumstances where we would
normally expect a subject gap, as in (13) and (14), it seems likely that the clitic has
actually turned into an agreement marker. The syntactic diagnostics discussed so far
are summarized in (15).
5
(15) Indications for an analysis of clitics as agreement markers
a. The clitic is obligatory (doubling is not contextually restricted).
b. The double in clitic doubling structures may be indenite/non-specic.
c. Anti-agreement between the clitic and the double.
d. The clitic shows up in subject gap environments.
In the next section, I will discuss a couple of morphological properties that may be
used as further diagnostics for telling apart clitics and agreement markers.
.. Morphological criteria
Zwicky and Pullum (1983) develop a set of morphosyntactic criteria which help to
make a principled distinction between clitics and inectional afxes. The criteria de-
rive from the insight that word-clitic combinability is largely governed by syntactic
considerations. The conditions governing the combinability of stems with afxes are
of quite a different sort: they are morphological and/or lexical in character, being
concerned with the substructure of a nite set of words. (p. 503). In this section, I will
review some of their criteria and show how they can be applied to separate clitics from
agreement markers.
The rst criterion proposed by Zwicky and Pullum relates to the well-known ob-
servation that afxes usually impose stricter selectional requirements with respect to
their lexical host than clitics:
(16) Clitics can exhibit a low degree of selection with respect to their hosts, while
afxes exhibit a high degree of selection with respect to their stems.
Zwicky and Pullum illustrate this criterion with differences between clitics and afxes
in English. They observe that clitics such as s is, s has, and ve have can attach to
words of virtually any category (p. 504):
(17) a. preposition:
The person I was talking tos going to be angry with me.
b. verb:
The ball you hits just broken my dining room window.
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.8 (508-573)
c. adjective:
Any answer not entirely rights going to be marked as an error.
d. adverb:
The drive home tonights been really easy.
In contrast, inectional afxes in general display quite specic selectional properties
(as to which element they attach to): in English, plural /-z/ attaches only to nouns, past
/-d/ only to verbs, and the superlative /-est/ only to adjectives or adverbs. If applied
to the grammaticalization of agreement markers from clitic pronouns, this criterion
should lead us to expect that the change in question goes hand in hand with tighter
restrictions on the class of elements the clitic/agreement marker can attach to. This
development can be observed in a number of Uto-Aztecan languages (cf. Steele 1977).
Many Uto-Aztecan languages exhibit a set of enclitic pronouns that move to second
position and attach to the rst word or constituent of the clause independent of its
syntactic category:
(18) hunwuti=pum
bear:object=clit.3pl
seiwun.
are:shooting
They are shooting a bear. (Luiseo; Steele 1977: 539)
(19) semati=ne
nice=clit.1sg
napaha
shirt
rarimea
gonna:buy
ar.
probably
I am probably going to buy a nice shirt. (Tarahumara; Steele 1977: 554)
(20) tuk=ne
yesterday=clit.1sg
antnyta
Antonio
bicak.
saw
Yesterday, I saw Antonio. (Yaqui; Steele 1977: 554)
Interestingly, in some languages such as Classical Aztec, the clitics uniformly attach to
the left of the nite verb:
(21) an-teecLasoLa.
2pl-love-us
You love us. (Classical Aztec; Steele 1977: 539)
Accordingly, the criterion in (16) suggests that former clitics have developed into
agreement prexes on the verb in Classical Aztec (see Section 4.5 below for an outline
of the relevant syntactic scenario and Chapter 5 for detailed discussion).
The following criterion makes use of the observation that the morphophonolog-
ical results of attaching a clitic to a certain lexical host are usually highly regular, that
is, there are no cases where the host-clitic combination yields an unexpected form. In
contrast, irregular forms, suppletion etc. are hallmarks of inectional morphology.
(22) Morphophonological idiosyncrasies are more characteristic of afxed words
than of clitic groups.
In connection with (22), Zwicky and Pullum (p. 505) note that this principle should
also be taken to capture the fact that the attachment of clitics typically leaves the lexical
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.9 (573-634)
host unaffected. Thus, if we can observe that an alleged clitic has an effect on the shape
or morphological structure of its host, this should cast some doubt on its status as a
clitic. Consider the following examples from the Northern Italian dialect Fiorentino
(Brandi & Cordin 1989):
(23) a. Icch
what
glhanno
clit.3pl=have-3pl
fatto?
done
b. Icch
what
ha(e)-gli-no
have-(clit.)3pl-3pl
fatto?
done
(Brandi & Cordin 1989: 132)
In inversion contexts, the subject clitic may optionally intervene between the verb stem
and the original agreement ending, as in (23b). Under the above assumptions, this
possibility suggests that the 3pl form gli is no longer a clitic, but has developed into a
marker of verbal agreement. In a similar vein, I will treat cases where an apparent clitic
replaces the usual verbal agreement morphology as an indication that the clitic has
in fact developed into an inectional afx (see Chapter 5 for further examples from
German and Swiss Rhaeto-Romance varieties).
Another criterion concerns gaps in the paradigm, which are much more common
with inectional afxes than with pronominal elements (cf. e.g. Rizzi 1986):
(24) Arbitrary gaps in the set of combinations are more characteristic of afxed
words than of clitic groups.
For example, Roberts and Roussou (2003: 180) argue that leveling of person/number
distinctions in the pronoun paradigm of a set of Northern Italian dialects (Veneto)
which show identical forms for 1sg, 1pl and 2pl promoted the reanalysis of these
pronominal clitics as realizations of AgrS.
Finally, Zwicky and Pullum propose two further criteria that follow from the as-
sumption that cliticization applies after syntax (i.e., no syntactic operations apply
after syntax, p. 504), which is also adopted in the present work (see Chapter 2 above):
(25) Syntactic rules can affect afxed words, but cannot affect clitic groups.
(26) Clitics can attach to material already containing clitics, but afxes cannot.
Applied to cases like (23b) from Fiorentino, principle (26) further supports the above
conclusion that the clitics in this dialects have evolved into inectional afxes, since we
would not expect an inectional afx to be able to attach to the right of a verb-clitic
combination.
A good example of the application of (25) comes from differences with respect to
the behavior of clitics in Standard and Non-Standard French.
6
It is a well known fact
that the subject clitics of Standard French are sensitive to syntactic context, that is, they
precede the verb in declaratives, but must follow it in matrix questions with wh-ex-situ
(Kayne 1975):
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.10 (634-696)
(27) a. Tu
you
attends
await
qui?
who
b. Qui
who
attends-tu?
await-you
Who are you waiting for?
The kind of subject-verb inversion illustrated in (27b) is commonly analyzed as an
instance of I-to-C movement of the nite verb which crosses the position of the sub-
ject clitic (Kayne 1984; Rizzi & Roberts 1989). That is, the subject clitics of Standard
French are not affected by syntactic rules that affect their host, in the sense that they do
not accompany verb movement to C. In contrast, the subject clitics of Non-Standard
French fail to undergo subject-verb inversion in matrix questions, as can be seen from
the example (28), taken from Lambrecht (1981: 6):
(28) O
where
tu
you
vas?
go
Where are you going?
In other words, the preverbal clitics of colloquial French must accompany the verb if
the latter is affected by movement operations. This can be taken to suggest that they
are already a xed part of verbal morphology, namely prexal agreement markers (for
further arguments against an analysis as pronominal elements cf. Lambrecht 1981;
Haiman 1991; Auger 1993, 1994; Friedemann 1995; Gerlach 2002).
Finally, and related to the latter considerations, we can construe a last criterion
which has to do with the observation that inectional elements tend to occur in a xed
position, while clitics can be reordered with respect to other clitics:
(29) Clitics can be reordered within the clitic complex, while afxes occupy xed
positions.
Concerning the historical development of agreement markers, we can observe that
free pronouns can occupy a variety of different positions (e.g., IP-internal, clause-
initial etc.), while clitics are usually conned to certain structural positions (e.g., the
so-called Wackernagel position) or certain hosts (most often the nite verb), but still
may switch positions inside the clitic complex. In contrast, agreement markers typ-
ically show no variation at all, occupying xed positions/slots inside the inectional
domain of the verb.
In this section, I have reviewed a number of diagnostics which may help to decide
whether a given element should be analyzed as a clitic or rather an inectional afx.
The syntactic criteria developed in Section 4.2.1 pay special attention to the properties
of clitic doubling congurations such as obligatory character, specicity of the double
etc. In contrast, the morphological criteria outlined in 4.2.2 focus on properties of
the clitic alone (basically following Zwicky & Pullum 1983), for example selectional
attributes, morphophonological effects on the host and distribution with respect to
other clitics and afxes. In the remainder of this book, the criteria developed in the
present section will be used as diagnostics for the categorial status of pronominal clitics
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.11 (696-758)
that are subject to a grammaticalization process which eventually gives rise to new
verbal agreement markers.
It is important to note that the above criteria not only constitute a tool for linguis-
tic analysis. In addition, they may also be taken to reect cues that signal the learner
the categorial status of a clitic element during language acquisition (in the sense of
Lightfoot 1999, see Chapter 2 above). For example, if the learner can identify clear
contextual restrictions on the use of clitic doubling, he/she will presumably conclude
that the element in question is actually a clitic and not an agreement marker. Con-
versely, obligatory presence of a clitic can be seen as a syntactic cue for agreement
status, while effects on the morphophonological shape of the host act as morphological
cues that equally signal agreement status. I will come back to this issue in connection
with concrete examples in Chapters 5 and 6.
. Syntactic preconditions for the rise of agreement
It is fairly clear that there is a number of general, theory-independent preconditions on
the historical development of agreement markers from pronominal elements, includ-
ing for example the availability of a series of non-stressed clitic pronouns (see Chapter
2 above), or a number of the properties discussed in the previous section, which might
be associated with the notion of cues, in the sense of Lightfoot (1999). In addition, each
restrictive theory of agreement imposes a set of further restrictions on the reanalysis in
question, the exact nature of which depends on the set of assumptions that constitute
the theory. In this section, I will discuss the implications of the theoretical approach to
agreement outlined in Chapter 3 for the diachronic development of agreement mark-
ers from pronouns. It is shown that the theory of agreement proposed in Chapter
3 imposes certain restrictions on the categorial reanalysis of clitic pronouns, making
a set of predictions with respect to the syntactic congurations where the change in
question may take place. In addition, I discuss a set of further syntactic conditions im-
posed by -theory and derivational/representational economy, which shape the course
of the reanalysis of clitic pronouns.
In Chapter 3, I have argued for a theory of agreement that is based on the following
assumptions:
(i) Agreement morphemes do not head their own projection in the syntax. Rather,
they are parasitic on other functional heads.
(ii) Agreement morphemes may combine with contentful functional categories such
as C, T, or (and maybe others) in one of the following ways. First, the agreement
morpheme may attach to its functional host prior to the insertion of that host into
the syntactic derivation. Second, the agreement morpheme may be added post-
syntactically as a dissociated Agr-morpheme, as in the case of complementizer
agreement in Germanic.
(iii) Canonical Agr-morphemes are valued by an appropriate controller under closest
c-command (Agree) during the syntactic derivation. Dissociated Agr-morphemes
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.12 (758-823)
are valued under structural adjacency with an Agr-morpheme that has been
licensed in the syntax.
(iv) Bound Agr-morphemes must combine with their host prior to Vocabulary In-
sertion, either by syntactic verb movement or Morphological Merger under
adjacency at MS.
The assumptions (i)(ii) actually do not impose any strong restrictions on the reanaly-
sis of clitics as agreement heads. Compared with previous analyses in which agreement
is assumed to occupy a unique, well-dened structural position in the structure of the
clause, it appears that the present analysis is actually more liberal, since in principle
agreement morphemes can come into existence as part of any contentful functional
category if the learner can detect the presence of an Agr-morpheme on a certain func-
tional head onthe basis of the evidence available to him/her. This assumption is further
exploited in Chapter 5, where it is argued that in V2 languages, new verbal agree-
ment markers may develop via an initial stage where the learner posits an agreement
morpheme attached to C.
Let us now turn to the impact of the theoretical assumptions (iii)(iv), which re-
quire that the feature content of Agr be valued and that Agr-morphemes combine with
the verb prior to Vocabulary Insertion, respectively. The latter consideration correlates
with the perhaps most obvious syntactic restriction on the reanalysis of pronouns as
verbal agreement afxes: a clitic pronoun can only be reinterpreted as an inectional
afx if it is adjacent to the verb.
7
(30) Adjacency requirement
A clitic pronoun can be reanalyzed as a bound agreement afx on the verb
only if the clitic is string-adjacent to the verb.
Under the assumption that agreement morphemes are hosted by other functional
heads, the intuition captured by (30) must be reformulated in more formal terms by
saying that a pronominal clitic can only be reanalyzed as an agreement morpheme on a
functional head X if X combines with the verb prior to Vocabulary Insertion. This can
be accomplished either by verb movement to X in the overt syntax or by post-syntactic
operations such as Morphological Merger, which require that X and the verb be ad-
jacent at MS. Since this requirement concerns the process of word formation, that is,
the creation of complex morphosyntactic units either in the syntactic component or
at MS, I choose to call it the Word building constraint:
(31) Word building constraint
The reanalysis of a clitic adjacent to the verb as a (bound) verbal agreement
marker on a functional head X requires that X combines with the verb prior
to Vocabulary Insertion.
In other words, one might say that the presence of the nite verb in a functional head
X (as a result of overt movement) signals the learner that X is capable of hosting verbal
agreement morphemes/features.
8
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.13 (823-894)
In addition, the feature content of the newly created agreement morpheme must
be licensed (or, valued, using the terminology of Chomsky 2000, 2001a) in a local
structural conguration. If UG allows for two kinds of Agr-morphemes (i.e., syntactic
and dissociated Agr-morphemes), which are subject to different licensing mechanisms
(see Chapter 3), then the reanalysis of pronominal elements is restricted by the follow-
ing condition:
(32) Identication of feature content
The reanalysis of a clitic pronoun is licit only if the resulting agreement mor-
pheme is licensed
(i) in the syntax by a local Agree relation with a matching set of interpretable
-features, or
(ii) at MS as a dissociated morpheme under structural adjacency with a syn-
tactically licensed Agr-morpheme.
Apart from the syntactic conditions that follow from the approach to agreement
adopted in this work, the reanalysis of pronouns must arguably fulll at least two fur-
ther requirements imposed by -theory and general economy principles which shape
the course of language change (see Chapter 2 above).
A rather obvious condition on the categorial reanalysis of pronouns is imposed by
-theory. Since a pronoun usually carries a thematic role, it must be ensured that the
role in question can still be assigned when the pronoun is reanalyzed as an agreement
marker and disappears from the set of arguments in a given clause:
(33) Preservation of argument structure
The reanalysis of a pronoun as an agreement marker must preserve the pred-
icates argument structure.
In other words, a reanalysis of a subject pronoun as an agreement marker can only take
place if the relevant thematic role can be assigned to another element, either another
overt DP (e.g., a former topic in a clitic left dislocation structure, cf. Givn 1976) or a
covert pronominal element, that is, pro. This leads to some interesting issues concern-
ing the interaction between the development of verbal agreement and the pro-drop
property. More specically, it seems that pro-drop may be either a precondition for or
the result of the development of (new) verbal agreement marking. In other words, a
pro-drop grammar facilitates the reanalysis of pronouns as agreement markers, since
the learner has at his/her disposal an empty pronoun which may be inserted to receive
the -role assigned to the pronoun in the target grammar. Conversely, the reanalysis
of pronouns may introduce new pro-drop properties into a grammar that previously
lacked pro-drop (for a related statement, cf. Roberts & Roussou 2003: 185f.).
In generative approaches to language change it is often assumed that a reanalysis
must conform to general economy principles, in the sense that the learner assigns a
given input string the most simple or economic structure/derivation that is compat-
ible with the evidence (cf. e.g. Clark & Roberts 1993). The preference for simplicity
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.14 (894-945)
may trigger a reanalysis of a given structure if the trigger for a more complex target
grammar has been obscured by phonological erosion or independent change processes
(see Chapter 2 above). In other words, if the evidence available to the learner is am-
biguous, he/she will opt for the most economic structural option. Recently, Roberts
and Roussou (2003) have argued that this kind of structural simplication is a key
feature of grammaticalization, showing that grammaticalization processes generally
lead to the loss of syntactic operations (either overt movement or an Agree relation
between a higher and a lower head). The loss of movement operations fullls the
requirement of structural simplication rather straightforwardly: if a lexical element
previously moved to a functional head is reanalyzed as being merged in this functional
head position, the change in question instantiates a case where a rather costly Move
operation is replaced by the more basic operation Merge. This is exemplied by the
following change which according to Poletto (1995, 2000) has taken place in a number
of Northern Italian dialects such as Venetian. In (34), a pronominal element previ-
ously moved to SpecAgrSP is reanalyzed as an instantiation of AgrS, leading to the loss
of DP-movement (Roberts & Roussou 2003: 178):
9
(34) [
AgrSP
DP
i
[
AgrS
V ] [
TP
. . . [
VP
t
i
... [
AgrSP
[
AgrS
D+V ] ...
However, the structural simplication associated with the loss of an Agree relation is
less straightforward (e.g., it cannot be captured by counting nodes or syntactic objects
present in the derivation). Roberts and Roussou (2003: 201) propose that all instances
of structural simplication can be captured by the following simplicity metric (cf.
Longobardi 2001: 294f. for a related proposal):
(35) A structural representation R for a substring of input text S is simpler
than an alternative representation R iff R contains fewer formal feature-
syncretisms than R.
Feature syncretism is dened as the presence of more than one formal feature in a
given structural position (Roberts &Roussou 2003: 201). For example, if a given func-
tional head H triggers movement of an element X, this always implies the presence of
an additional formal (EPP-) feature F in H that attracts X (in addition to the set of
features that licenses Merge of H). If X is reanalyzed as being merged in the position
of H, this leads to the disappearance of F and therefore to a structure with less feature
syncretisms.
10
This approach works in a parallel fashion if the reanalysis leads to the
disappearance of an Agree-relation, implying the loss of the formal feature that origi-
nally triggered the Agree operation. In other words, the most economical structure is a
structure with the least occurrences of multiple features on single positions (Roberts
& Roussou 2003: 201), where ideally each syntactic head is spelled out by an element
that has been merged as this head. Note that the rationale behind (35) is therefore quite
similar to the traditional (structuralist) idea that ideally, there should be an one-to-one
relation between underlying grammatical features and overt grammatical morphemes
(the so-called agglutinative ideal, cf. Spencer 1991 for critical discussion; cf. Julien 2002
for a related proposal).
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.15 (945-1007)
Concerning the development of agreement markers from pronominal elements,
we should thus expect that the relevant reanalysis is only possible if the resulting struc-
ture involves less movement or Agree operations than the target structure. In general
this seems to be true, for example in cases where a subject pronoun previously moved
to SpecTP is reanalyzed as an agreement morpheme on T. Below, however, we will
see that the reanalysis of pronouns often leads to structures that are morphologically
more complex than the target structure, as in cases where a pronoun is reanalyzed as a
dissociated Agr-morpheme that is inserted by a morphological operation at MS.
The assumption that the reanalysis frompronouns to agreement markers is shaped
by the syntactic conditions discussed so far predicts that the change under consid-
eration can only take place in a certain set of syntactic environments where these
restrictions are satised. In the following, I give an outline of (a selection of) the
relevant syntactic congurations, focusing on the development of subject-verb agree-
ment. The exposition is arranged according to the well-known distinction between
In-oriented clitics (Section 4.4) and C-oriented clitics (4.5).
11
. Paths toward agreement I: In-oriented clitics
Let me begin with the conguration discussed by Givn (1976), that is, topic left dis-
location structures where the topic becomes the new subject while the resumptive
pronoun is reanalyzed as a subject agreement marker. Following Uriagereka (1995),
Cecchetto (2000), Kayne (2002), and Grewendorf (2002), I assume that the resump-
tive pronoun and the dislocated topic are merged together as a big DP, undergoing
movement to SpecTP for Case/EPP reasons. In a subsequent step, the topic further
moves into the left clausal periphery.
12
(36) [
CP
topic
j
... [
TP
[
DP
t
j
pronoun]
i
[
T
T [
P
t
i
...
[
CP
[
TP
subject
i
[
T
Agr+T [
P
t
i
...
Structural simplication: loss of topic movement,
loss of big DP
Preservation of argument structure: -role previously assigned
to big DP is now assigned
to previous topic
Identication of feature content: via Agree with the
subject in SpecP
Word building constraint: verb is located in T (via
movement) or adjacent to
T at MS (enabling
Morphological Merger)
As already indicated in (36), the reanalysis in question satises the relevant constraints
in a more or less straightforward way. Only the status of the Word building constraint
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.16 (1007-1058)
is somewhat unresolved in this structure and contingent on other properties of the
grammar. That is, the language in question must already have had V-to-T movement
prior to the reanalysis. Alternatively, T and the verb (which stays in situ) must be ad-
jacent at MS. The latter option is arguably only available in an SVO grammar similar
to English, but not in an SOV structure, where the object regularly intervenes between
T and the (VP-internal) verb. Note that this predicts that the reanalysis in (36) should
only be possible in (i) SVO languages or (ii) SOV languages with verb movement to
head-initial T. Possible examples of this type of change include English- and French-
based Creoles (Givn 1976), a number of Bantu languages (Givn 1976; Bresnan &
Mchombo 1987), and perhaps a number of Northern Italian Rhaeto-Romance dialects
such as Friulian, Fassan, or Ampezzan (Haiman & Beninc 1992: 187ff.; but see the
discussion around (37) below).
13
In Chapter 1, I have argued that (36) does not constitute a viable source of 1st and
2nd person agreement, since in many languages, clitic left dislocation of 1st and 2nd
person pronouns leads to highly marked, often ungrammatical results. Accordingly,
we must look for a different historical source of 1st and 2nd person agreement. Here,
I assume that verbal agreement markers may develop in a somewhat similar construc-
tion that also involves some form of clitic doubling (i.e., a big DP), where for reasons
of emphasis a full DP double is added to reinforce a weak/clitic pronoun that cannot
bear stress (cf. Simpson & Wu 2002 for a related idea):
(37) [
CP
[
TP
[
DP
double [
D
pronoun]]
i
[
T
T [
P
t
i
...
[
CP
[
TP
subject
i
[
T
Agr+T [
P
t
i
...
Structural simplication: loss of big DP
to previous double
subject in SpecP
T at MS (enabling
At rst sight, this kind of change is very similar to the reanalysis of clitic left disloca-
tion illustrated in (36). Still, it crucially differs from (36) in that it does not involve
(left) dislocation of the double. Because of this difference, (37) may give rise to 1st
and 2nd person agreement formatives, while (36) is restricted to 3rd person forms.
Thus, it seems reasonable to suppose that languages which license the kind of dou-
bling arising from a big DP structure may employ both strategies for the creation of
newverbal agreement markers from former clitic pronouns. Relevant examples of (37)
come from languages such as Non-Standard French, where 1sg, 2sg, and 1pl clitics are
obligatory and are therefore presumably better analyzed as prexal agreement markers
(Auger 1993, 1994; Gerlach 2002), and a number of Rhaeto-Romance dialects spoken
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.17 (1058-1122)
in Northern Italy, which exhibit obligatory clitic doubling with 1st and 2nd person
forms (Haiman & Beninc 1992).
14
Furthermore, it seems that with some slight mod-
ications, (37) can also account for the rise of object agreement from clitic pronouns
in the recent history of Spanish (cf. e.g. Suer 1988).
15
Another possible syntactic source of verbal agreement morphology, which in con-
trast to the previous scenarios does not involve an intermediate stage of clitic doubling,
consists of the direct reanalysis of In-oriented weak/clitic pronouns in languages
that either were already pro-drop languages prior to the reanalysis or developed pro-
drop properties as a result of it.
(38) [
CP
... [
TP
pronoun
i
[
T
T [
P
t
i
...
[
CP
... [
TP
[
T
Agr+T [
P
pro ...
Structural simplication: overt movement replaced
by an Agree-relation
to overt pronoun is now
assigned to pro
subject (pro) in SpecP
movement)
16
Examples of this reanalysis come from a variety of Northern Italian dialects such as
Venetian (Poletto 1995, 2000; Roberts & Roussou 2003), where pronominal elements
moved to SpecAgrSP/SpecTP are reanalyzed as realizations of AgrS. As already men-
tioned in Note 9 above, the reanalysis described in (38) raises some questions with
respect to the notion of structural simplication. Roberts and Roussou (2003) as-
sume that the newly created AgrS-head receives the subjects -role, which renders
an additional pro unnecessary and ensures that the outcome of the change under con-
sideration is simpler than the structure in the target grammar. However, I prefer to
assume that the reanalysis in question leads to the result shown in (38), where a refer-
ential pro is added to the structure, which is more in line with standard assumptions
concerning the assignment of -roles. The requirement of structural simplication is
then met by replacing overt syntactic movement to SpecTP by an Agree relation be-
tween the T+Agr complex and the empty pronominal subject, which stays behind in
its -position SpecP. In other words, (38) satises the requirement that the output of
the reanalysis should involve less feature syncretisms than the target structure (cf. (35)
above) by the loss of an EPP feature in T. Assuming further that the EPP actually has a
PF-avor in the sense that it simply requires that SpecTP is lled by visible material at
PF (cf. e.g. Holmberg 2000), it is actually expected that the reanalysis in (38) leads to
the loss of an EPP feature in T: due to the fact that the resulting empty subject pronoun
is not associated with phonological features, it cannot satisfy the EPP and is therefore
banned from moving to SpecTP.
17
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.18 (1122-1178)
Finally, new subject agreement markers may develop in VSO languages via a re-
analysis of pronouns that follow the nite verb, which has undergone movement to a
higher functional head. Under the assumption that in many VSO languages, the verb
moves to an inectional head such as T, while the subject occupies an intermediate po-
sition between T and P/VP (simply labeled FP here; cf. e.g. McCloskey 1996, 1997 for
the Celtic VSO languages, in particular Modern Irish) the relevant structural change
can be described as follows:
18
(39) [
CP
[
TP
[
T
V+T [
FP
pronoun
i
[
P
t
i
...
[
CP
[
TP
[
T
V+T+Agr [
FP
... [
P
pro ...
assigned to pro
movement)
Similar to (38) above, a pronoun is directly reanalyzed as a marker of verbal agree-
ment, while the relevant -role is assigned to a phonologically empty pronoun that
stays behind in SpecP and is accessed by an Agree operation initiated by the T+Agr
complex. Again, it is assumed that the requirement of structural simplication is met
by replacing overt pronoun movement with an Agree relation (= loss of an EPP feature
in T). The reanalysis in question has presumably taken place in the history of Welsh
(cf. Roberts & Roussou 2003) and might be underway in Irish, cf. e.g. McCloskey and
Hale (1984).
Summing up, this section has introduced a set of syntactic environments that li-
cense the reanalysis of In-oriented clitics as (bound) agreement markers. It has been
shown that the individual scenarios all comply with the set of syntactic conditions
outlined in Section 4.3, albeit each in different ways. For example, the requirement
of structural simplication may be satised by the loss of movement (as in (38) and
(39)), the loss of structural complexity (as in (37)), or both, as in (36). Other points
of variation concern the element which receives the -role previously assigned to the
pronominal clitic. The vacant -role may be assigned to a full DP that originally served
(i) to double the clitic in topic left dislocation structures (as in (36)) or (ii) to reinforce
the clitic for reasons of emphasis, as in (37). Alternatively, the clitic may be directly
reanalyzed as an agreement marker, with the subject -role being assigned to an empty
pronoun (as in (38) and (39)).
The discussion in this section has shown that In-oriented clitics are a viable
source of new agreement morphology. However, it appears that the structural changes
in question primarily involve head-initial structures. In the next section, it is argued
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.19 (1178-1251)
that the reanalysis of C-oriented clitics provides alternative paths toward agreement
which can be utilized to coin new inectional markers in SOV languages.
. Paths toward agreement II: C-oriented clitics
In the following, it is shown that languages with C-oriented clitics exhibit some inter-
esting variations on the above patterns. More specically, I argue that the V2 syntax
of many (Germanic/Romance) languages facilitates the reanalysis of C-oriented clitics.
Furthermore, it is shown that the reanalysis in questioninvolves the creation of a disso-
ciated Agr-morpheme on C which is inserted post-syntatically (see Chapter 3 above),
at least in languages that already had an existing paradigm of verbal agreement mark-
ers prior to the reanalysis in question. In addition, I suggest that there are at least two
further congurations where C-oriented clitics may turn into new agreement markers
in languages lacking the V2 property. Note that the various changes are only briey
outlined in the present section. They are discussed in more detail in Chapter 5, which
provides a number of more detailed case-studies of languages where C-oriented clitics
developed into new markers of verbal agreement.
First, we can observe that in V2 languages, new verbal agreement markers may
arise out of clitic doubling structures where a full nominal is added to the clitic
pronoun for reasons of emphasis (similar to (37) above). In the course of time, the
enclitic, which undergoes cliticization movement to C (either syntactically or at MS),
is reanalyzed as an agreement morpheme on C:
(40) [
CP
XP [
C
C+V+pronoun
j
[
TP
[
DP
double [
D
t
j
]]
i
[
T
T [
P
t
i
...
[
CP
XP [
C
C+V+Agr [
TP
[
TP
subject
i
[
T
T [
P
t
i
...
Structural simplication: loss of big DP (+ loss of
cliticization)
to previous double
Identication of feature content: dissociated Agr-
morpheme licensed under
structural adjacency with
Agr-on-T
Word building constraint: verb is located in C (via
movement)
With respect to the requirements of structural simplication and Preservation of ar-
gument structure, (40) is quite similar to (37).
19
However, it differs in two important
ways from the examples with In-oriented clitics discussed above. First, the reanalysis
in (40) is dependent on the V2 property and can only take place in inversion contexts,
since this is the only context where the reanalysis in questioncan satisfy the Word build-
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.20 (1251-1316)
ing constraint (at least in SOV languages, C-oriented clitics fail to be adjacent to the
verb in embedded clauses). Second, the present reanalysis gives rise to an agreement
morpheme on C, whereas it has been assumed that the reanalysis of In-oriented
clitics leads to a new canonical agreement morpheme on T. This typically gives rise
to instances of multiple agreement where the subjects -set is not only reected by
the existing verbal morphology (associated with Agr-on-T), but also shows up on the
newly created agreement morpheme on C. In Chapter 3 above, it has been argued that
this state of affairs results from the post-syntactic insertion of a dissociated agreement
morpheme at MS. In the present context, this entails that the pronoun is reanalyzed as
a dissociated Agr-morpheme on C that is licensed under structural adjacency with Agr-
on-T, the content of which has been evaluated in the syntactic derivation via Agree.
In Chapter 5, it is argued that this reanalysis is apparently underway in a number of
(Swiss) Rhaeto-Romance dialects.
Second, in inversion contexts, C-oriented clitic pronouns may be directly reana-
lyzed as agreement markers on C, giving rise to pro-drop properties without involving
a prior stage of clitic doubling (similar to (38) above):
(41) [
CP
XP [
C
C+V+pronoun
i
[
TP
t
i
[
T
T [
P
t
i
...
[
CP
XP [
C
C+V+Agr [
TP
T [
P
pro ...
assigned to pro
Identication of feature content: dissociated Agr-
morpheme licensed under
structural adjacency with
Agr-on-T
movement)
Again, this change is made possible by the special V2 syntax of the languages in ques-
tion, which guarantees adjacency between the pronoun and the verb and thereby
facilitates a reanalysis of the enclitic as a dissociated agreement morpheme on C. Sim-
ilar to (38) above, the resulting structure involves the loss of overt movement, since
the newly created empty subject pronoun stays behind in SpecP and is accessed by an
Agree operation initiated by the complex T+Agr (prior to movement of the complex
V+T+Agr to C). Relevant examples for this structural change come from a variety of
German and Rhaeto-Romance dialects, see Chapter 5 for details.
In addition, C-oriented clitics can evolve into verbal agreement markers in lan-
guages that lack the V2 property. Here, I will mention only three relevant structural
congurations, which presumably do not exhaust the range of possibilities.
20
For ex-
ample, true second position clitics which occupy a position after the rst prosodic
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.21 (1316-1370)
phrase or word may turn into prexal agreement markers in languages where they are
adjacent to the nite verb (presumably located in T):
(42) [
CP
XP [
C
C+pronoun
i
[
TP
t
i
[
T
V+T [
P
t
i
...
[
CP
XP [
C
C [
TP
Agr+V+T [
P
pro ...
assigned to pro
movement)
In (42), the reanalysis of the pronoun as a prexal agreement marker requires that the
verb is adjacent to second position. Thus, the change in question is possible only in
VO languages, in contexts where nothing intervenes between the C-oriented clitic and
the nite verb. As in (38) above, the outcome of the reanalysis satises the requirement
of structural simplication by the loss of overt movement (pro stays behind in SpecP
and is accessed by an Agree operation initiated by the complex T+Agr).
Similar to V2 languages, there exists an alternative scenario which involves a dou-
bling conguration. In Chapter 5, it is argued that this grammaticalization process
can be observed in a number of Uto-Aztecan languages (Steele 1977), where sec-
ond position clitics develop into prexal agreement markers, attaching to the nite
verb in T:
(43) [
CP
XP [
C
C+pronoun
i
[
TP
[
T
V+T [
P
[
DP
double [
D
t
i
]] ...
[
CP
XP [
C
C [
TP
Agr+V+T [
P
subject ...
to the former double
subject in SpecP
movement)
Similar to (40), this reanalysis leads to structural simplication via losing the big DP
structure underlying clitic doubling in the target grammar. Preservation of argument
structure is warranted by assigning the -role to the former double. The newly cre-
ated agreement morpheme is valued by initiating an Agree operation that accesses the
subject in SpecP, which by assumption may stay in situ in the Uto-Aztecan languages
under consideration (see Chapter 5 for details).
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.22 (1370-1430)
In Chapter 5, I will also discuss a set of Mongolian SOV languages (including
Buryat, Dagur, Moghol, Kalmyk; Poppe 1960; Comrie 1980) which developed sufxal
agreement markers in their recent recorded history. It is argued that these agreement
formatives originated from a structure where enclitic pronouns attach to C while a
constituent (presumably TP) containing the rest of the clause moves to a specier in
the C-domain (see Julien 2002 for a related proposal):
(44) [
CP
[
TP
t
j
[
T
[
P
t
j
[
VP
object]] V+T]]
i
[
C
C+pronoun
j
t
TP
]]
[
CP
[
TP
[
T
[
P
pro [
VP
object]] V+T+Agr]]]
Structural simplication: loss of TP movement;
subject movement
replaced by an Agree-
relation
assigned to pro
T at MS (enabling
In this way, (apparent) clause-nal clitics may evolve into verbal agreement sufxes in
an SOV language. The reanalysis depicted in (44) leads to a major structural reorgani-
zation in which TP movement and overt subject movement to SpecTP are lost, a rather
clear-cut case of structural simplication. Note that the change in question represents
another way (apart from inversion contexts) in which OV-languages can develop suf-
xal verbal inection. Furthermore, in OV-languages there exists always the possibility
that the verb stays in situ and combines with the resulting agreement morpheme (on
T) via Morphological Merger at MS, due to the fact that in a strict OV grammar, the
verb is always string-adjacent to the set of head-nal inectional heads (still, I opt for
a structure with overt verb movement in (44), see Chapter 5 for discussion).
To sum up, this section has introduced a set of environments where C-oriented
clitics may develop into new markers of verbal agreement. First, it has been suggested
that the V2 property provides a path for the grammaticalization of agreement markers
in SOV languages, where clitics are otherwise not adjacent to the verb. The relevant
change may involve either doubling structures where the double turns into the new
subject or a direct reanalysis of the clitic into an agreement morpheme, giving rise to
new pro-drop properties. Importantly, it has been argued that the change in ques-
tion proceeds via an initial stage where the pronoun is reanalyzed as a dissociated
Agr-morpheme on C, resulting in multiple agreement congurations, similar to the
phenomenon of complementizer agreement discussed in Chapter 3 above. In addi-
tion, I have outlined two further scenarios for the categorial reanalysis of C-oriented
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.23 (1430-1482)
clitics in languages that lack the V2 property. First, true second position clitics may
evolve into new prexal agreement markers in languages where the nite verb is right-
adjacent to the clitic. Second, I have argued that clitics which appear in (apparent)
clause-nal position in SOV languages (e.g., Mongolian) may be reanalyzed as agree-
ment sufxes on the left-adjacent verb which has undergone movement to SpecCP as
part of a fronted TP.
. Summary
In this chapter, I have argued that the reanalysis of pronouns as agreement markers
is not conned to a unique syntactic environment in contrast to common beliefs
in the literature. I suggested that this change can be triggered in a number of con-
texts where a set of syntactic restrictions on the reanalysis of pronominal elements
is fullled. At least in part, the relevant syntactic restrictions were derived from the
synchronic theory of agreement developed in Chapter 3.
In addition, I specied a number of syntactic and morphological diagnostics for
distinguishing between clitics and agreement markers which can also be conceived
of as cues for the learner during language acquisition. The relevant syntactic criteria
reviewed in 4.2.1 are repeated in (45).
(45) Indications for an analysis of clitics as agreement markers
a. The clitic is obligatory (doubling is not contextually restricted).
b. The double in clitic doubling structures may be indenite/non-specic.
c. Anti-agreement between the clitic and the double.
d. The clitic appears in subject gap environments.
In addition, Section 4.2.2 presented a number of morphological criteria that can also
be made use of for the present purpose, basically following Zwicky and Pullum (1983).
The relevant criteria included the type of selectional restrictions imposed by the ele-
ment under consideration (rather loose in the case of clitics, rather strict in the case of
afxes), morphophonological effects on the host, and the distribution with respect to
other clitics and afxes.
In Section 4.3, I argued that the theory of agreement developed in Chapter 3
imposes a set of syntactic restrictions on the reanalysis of pronouns as agreement
markers. More precisely, it was claimed that the change in question is only possible
if it (i) preserves the predicates argument structure (either by assigning the relevant
-role to the double in a clitic doubling conguration or to a newly introduced pro),
(ii) provides a means to value the feature content of the resulting agreement mor-
pheme (either in the syntax or at MS), and (iii) warrants that (bound) agreement
morphemes can combine with the verb prior to Vocabulary Insertion. In addition, I
followed Roberts and Roussou (2003) in assuming that the reanalysis must conform
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.24 (1482-1549)
to a simplicity metric (formulated in terms of feature syncretism) which requires that
the resulting structure is less complex than the target structure.
Subsequently, I suggested that these conditions can be met in a variety of different
syntactic environments. In Section 4.4, I focused on the reanalysis of In-oriented
clitics in SVO and VSO grammars. It was shown that viable historical scenarios for the
change in question are divided into two groups. In the rst group, the -role previ-
ously assigned to the clitic is transmitted to a full DP that originally served to double
the clitic in (i) clitic left dislocation structures, or (ii) doubling congurations where
a full nominal is added to the clitic for reasons of emphasis. By assumption, both
scenarios involve the reanalysis of a former big DP which contains the clitic and the
double. In the second group, the clitic is directly reanalyzed as an agreement marker,
with the subject -role being assigned to an empty pronoun which is newly added to
the structure. It was demonstrated that all these environments meet the requirement
of structural simplication either via loss of phrase structure or loss of movement
operations.
In Section 4.5, I argued that languages with C-oriented clitics provide a number of
further environments where pronominal elements may be reanalyzed as verbal agree-
ment markers. First, I suggested that the reanalysis of C-oriented enclitics provides an
alternative path to new verbal agreement morphology in languages that exhibit the V2
property. Again, the relevant reanalysis may involve either doubling structures where
the double turns into the new subject or a direct reanalysis of the clitic into an agree-
ment morpheme, giving rise to new pro-drop properties. In line with the analysis of
agreement presented in Chapter 3 above, I argued that the change in question may lead
to the presence of a dissociated Agr-morpheme on C, giving rise to instances of multi-
ple agreement. In addition, I claimed that C-oriented clitics may also give rise to new
verbal agreement markers in languages lacking the V2 property. By assumption, new
prexal agreement markers may evolve from true second position clitics in languages
where the nite verb is right-adjacent to the clitic. In addition, it was proposed that
clause-nal clitics in SOV languages such as Mongolian may give rise to new verbal
inection if it is assumed that the clitics in question attach to right of C which hosts a
fronted TP in its specier.
At this point let me address the question of whether it is possible to establish a
connection between the sufxing preference (cf. Chapter 1) and the set of historical
scenarios discussed in the present chapter. At rst sight, it seems that this not the case.
Of the nine syntactic environments presented above, ve give rise to prexal agree-
ment, while only four result in agreement sufxes. Thus, it seems that there exists in
fact a slight preference for the development of prexal agreement. However, note that
I have identied only two scenarios for the development of new agreement markers in
SOVlanguages (inversion and TP-fronting), which both give rise to sufxal agreement.
This captures the fact that SOV languages are predominantly sufxing (cf. Chapter 1).
In addition, recall that SOV is the most common word order across the worlds lan-
guages, so the two scenarios in question are presumably much more frequent than
other pathways which may lead to prexal agreement. Moreover, the fact that postver-
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.25 (1549-1613)
bal material afxes more readily to the verb than preverbal material is presumably also
inuenced by morphophonological factors, to the effect that enclitics are more likely to
fuse with the verb than proclitics (cf. e.g. Cutler, Hawkins, & Gilligan 1985; Hawkins &
Gilligan 1988; Hall 1988; Bybee et al. 1990; Kirby 1999; Siewierska 2004 for discussion).
Tentatively, I therefore suggest that the sufxing preference should receive a historical
explanation, in the sense that it results at least partially from the way agreement mark-
ers come into existence historically. In other words, it seems that cross-linguistically,
the grammaticalization process in question is more likely to be triggered in syntactic
environments which give rise to sufxal agreement morphology.
21
In the following chapter, I will discuss the historical scenarios involving the re-
analysis of C-oriented clitics in some more detail, drawing on examples from German
and Rhaeto-Romance dialects, Uto-Aztecan, and Mongolian languages.
Notes
. Note that in contrast to Italian, Macushi exhibits an absolutive/ergative organization. The
absolutive argument is cross-referenced on the verb by a prex, while the ergative argument is
marked by a sufx which follows the tense marker.
. For the claimthat subject clitic doubling constitutes the historical source of verbal agreement
cf. e.g. Givn (1976), Haiman(1991), Haimanand Beninc (1992), Simpsonand Wu (2002). For
an analysis of the subject clitics of Northern Italian dialects as agreement heads, cf. Rizzi (1986),
Brandi and Cordin (1989), Suer (1992), Poletto (1995, 1999, 2000), among many others. For
the claimthat object clitic doubling in Spanish in fact constitutes an instance of object agreement
cf. e.g. Givn (1976), Borer (1984), Suer (1988).
. Note that this relates to the observation that it is a characteristic of grammaticalization pro-
cesses that the relevant element becomes more and more obligatory (see Chapter 2 above for
discussion).
. Similar facts can be observed in Non-Standard French (cf. Lambrecht 1981):
(i) a. Il
he
mange
eats
et
and
boit
drinks
comme
like
un
a
cochon.
pig
(Standard)
b. I
he
mange
eats
et
and
i
he
boit
drinks
comme
like
un
a
cochon.
pig
(Non-Standard)
. Other criteria which have been proposed in the literature concern for example the placement
of clitics relative to other elements such as negation (Haiman1991; Haiman&Beninc 1992, and
in particular Poletto 2000). See Gerlach (2002: Ch. 2) for further criteria and in-depth discussion
and Ingram (1978) for a set of discourse related diagnostics.
. According to Kayne (1984), the decient subject pronouns of Standard French behave like
NPs in the syntax, but attach to the verb as clitics at PF.
. Here, the question arises of whether adjacency must hold without exceptions for a reanalysis
to be possible or whether it is sufcient if adjacency is given in the majority of cases. Note that
in many cases this creates no problems, since adjacency of the verb and the clitic is usually
obligatory.
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.26 (1613-1675)
. Note that (31) must be formulated in more general terms to account for cases where the
agreement marker does not attach to the verb, but rather to other (free) inectional morphemes
such as tense. Relevant examples come from Australian languages such as Wambaya (showing a
clitic Agrs/Agro/T marker, Nordlinger 1995), the Mayan language Jacaltec (Craig 1977), where
Agrs attaches to a clause-initial completive marker, and Finnish, where Agrs may attach to the
negation.
(i) Word building constraint (alternative)
The reanalysis of a clitic as a (bound) agreement marker on a functional head X requires
that X combines with the potential host of agreement prior to Vocabulary Insertion.
. To make sure that (34) fullls the requirement of structural simplication, Roberts and
Roussou (2003: 183) assume that the newly created AgrS-head receives the subjects -role, in
contrast to standard assumptions concerning the realization of argument structure. Otherwise,
one would have to assume that the reanalysis of (34) requires the presence of a referential pro
which receives the -role previously assigned to the pronoun. This, however, would annul the
structural simplication shown in (34) if pro moves to SpecTP for Case/EPP reasons, similar to
the overt pronoun. See below for some discussion and an alternative proposal.
. Note that this interpretation of the workings of the simplicity metric differs somewhat from
the proposals in Roberts and Roussou (2003), where feature syncretism and simplication is
associated with the PF-realization of features. More speccally, Roberts and Roussou assume
that a lexical item L that moves to a functional position F (and is spelled-out there) realizes both
its own features and the formal features of F. Structural simplication is then understood in
terms of the PF realization of these features, in the sense that a lexical item which PF-realizes
two separate features X and Y is more complex than a lexical item which realizes only a single
feature.
. In this book, I will not deal with instances where newsubject (absolutive) agreement markers
develop from former object pronouns. This often involves the reanalysis of passives as actives,
which in addition may give rise to an ergative alignment in the languages under investigation
(cf. e.g. Anderson 1980).
. Under a big DP analysis, Givns proposals might lead to sufxal agreement if it is assumed
that the resumptive clitic may remain in situ (i.e., SpecP), escaping the Case lter via cliticiza-
tion onto the verb in T. In general, absence of subject movement to SpecTP may open up further
possibilities with respect to syntactic environments where subject agreement can arise.
. For more examples cf. Givn (1976) and the references therein. According to Givn, Non-
Standard French constitutes a further instance of this reanalysis. However, it seems equally
possible that the change affecting French involves the structures in (37).
. Note that only (36) is associated with certain person restrictions. In contrast, (37) may
give rise to a full paradigm of agreement formatives. Thus, one might speculate that the cases
commonly attributed to (36) can be subsumed under (37). I leave that point open for future
research.
. In cases of Spanish object clitic doubling such as (i) (Suer 1988: 394), the double stays in
a position to the right of the verb, while the clitic undergoes movement (either in the syntax
or at PF/MS) to the left of the verb. This order may then be reanalyzed as involving a prexal
agreement marker and a postverbal object (presumably in situ):
JB[v.20020404] Prn:3/08/2005; 10:20 F: LA8104.tex / p.27 (1675-1741)
(i) La
her
llamaron
call-3pl
a ella.
to her
They called her.
. It is standardly assumed that the verb has to undergo movement in pro-drop languages
to license the empty subject pronoun. Thus, the late combination of the verb and the Agr+T
complex via Morphological Merger is presumably not available.
. As already mentioned above in Chapter 2 (Note 28), the reanalysis in (38) may lead to the
apparent multiple realization of agreement features in cases where the verb already carries an
agreement ending, as in the Northern Italian dialects discussed by Poletto (1995, 2000) and
Roberts and Roussou (2003). Note that the analysis of agreement proposed in Chapter 3 above
makes available an alternative solution to the problem of multiple agreement. More precisely, I
suggest that in cases where the verb already carries (sufxal) agreement morphology, the pro-
noun is reanalyzed as a dissociated agreement morpheme that adjoins post-syntactically to a
functional head directly above the head that licenses verbal agreement (presumably Agr-on-T).
Recall that the dissociated Agr-morpheme is then licensed under structural adjacency with Agr-
on-T (to which the verb moves). Thus, I propose the following structures for the reanalysis in
question (with the host of the dissociated Agr-morpheme simply labeled F):
(i) [
TP
DP
i
[
T
V+T+Agr ] ... [
VP
t
i
... [
FP
F+Agr [
F
[
TP
[
T
V+T+Agr ] ... [
VP
...
. A similar reanalysis, albeit one step further up in the structure, may take place if VSO
languages are analyzed as involving verb movement to C, with the subject pronoun in SpecTP.
. In V2 languages, one might expect that there exists an analogue of (36) in which C-oriented
resumptive enclitics turn into agreement markers in left dislocation contexts. However, in all
Germanic languages, the relevant structures involve a resumptive demonstrative which precedes
the verb, giving rise to V3 orders as for example in German (cf. Grewendorf 2002):
(i) Der
the
Peter,
Peter
der
dem
hat
has
Maria
Mary
geksst.
kissed.
As for Peter, he kissed Mary.
. For example, as already mentioned above in Note 18, C-oriented clitics may turn into new
agreement markers in verb-initial languages where the verb moves to C.
. Furthermore, note that many present-day SVO languages were originally SOV languages.
In contrast, the change from SVO to SOV seems to be quite rare cross-linguistically (cf. e.g.
Kiparsky 1996). Moreover, it is a well-known fact that the sufxal agreement morphology of
many present-day SVOlanguages is of considerable antiquity. Taken together, these facts suggest
that in many present-day SVO languages, the sufxal agreement formatives actually developed
when the language in question was SOV. In contrast, the converse development leading to agree-
ment prexes in a present-day SOV language which was previously SVO is much less likely.
This presumably provides a diachronic explanation of the fact that there are much more SVO
languages with agreement sufxes than SOV languages with agreement prexes.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.1 (45-122)
Chapter 5
The reanalysis of C-oriented clitics
. Introduction
This chapter deals with a particular subset of the structural reanalyses outlined in
Chapter 4, focusing on instances where C-oriented subject clitics develop into new
verbal agreement markers. Relevant examples of this historical scenario come from a
variety of V2 languages such as Bavarian, German dialects spoken in Northern Italy
(Cimbrian, Walser German), and a number of Rhaeto-Romance dialects spoken in
Switzerland. In addition, it is shown that related changes can be observed in languages
lacking the V2 property such as Uto-Aztecan and Mongolian. Based on the insights
reached in Chapter 4, I argue that the different historical pathways leading to new ver-
bal agreement markers share a well-dened set of properties, reecting the universal
nature of the abstract syntactic conditions which shape the grammaticalization process
in question across languages.
The chapter is organized as follows. Section 5.2 deals with the grammaticalization
of newverbal agreement markers from C-oriented enclitics in the history of Germanic,
focusing on Bavarian, where new agreement endings developed for 2nd person forms
and 1st plural. Section 5.2.1 describes the course of the grammaticalization process
while Section 5.2.2 presents a new account of the relevant changes based on the idea
that enclitics were reanalyzed as exponents of a dissociated Agr-morpheme on C. In
Section 5.2.3, I discuss further examples of this particular grammaticalization process
which can be observed in the dialects of German speaking minorities in Northern Italy,
drawing on data from Cimbrian and Walser German. Section 5.3 is concerned with
historical developments in a set of Rhaeto-Romance dialects spoken in Switzerland.
After providing a grammatical sketch of the languages under consideration in Sec-
tion 5.3.1, I show in Section 5.3.2 that similar to Bavarian, the Swiss Rhaeto-Romance
dialects were affected by a change in which C-oriented clitics were reanalyzed as en-
largements of existing agreement endings in inversion contexts. Section 5.3.3 shows
that some of these dialects exhibit another grammaticalization process in which C-
oriented clitics are reanalyzed as agreement markers in clitic doubling congurations.
Section 5.3.4 provides an analysis of this change which is again based on the idea that
clitics are reanalyzed as exponents of a dissociated morpheme on C. Section 5.4 deals
with alternative paths to new agreement markers in languages lacking the V2 property.
Section 5.4.1 focuses on second position clitics that turn into prexal agreement in a
number of Uto-Aztecan languages. Section 5.4.2 deals with the Mongolian SOV lan-
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.2 (122-197)
guage Buryat, arguing that the reanalysis in question targets C-oriented clitics which
show up in clause-nal position due to TP-fronting.
. Bavarian
In contrast to Standard German, Bavarian shows two series of pronominal elements:
(i) a paradigm of full forms that may bear stress, and (ii) a set of reduced, atonic
forms. The latter are C-oriented enclitics which attach to the right of elements located
in the C-system.
Table 1. Nominative pronominal forms of Bavarian (Bayer 1984: 230).
Full form Enclitic
1sg I /i/ -a, -e /a/, /e/
2sg du /du/ -(s)t /st/
3sg.masc er, der (demonst.) /er/, /der/ -a /a/
3sg.fem sie, die (demonst.) /si/, /di/ -s /s/
3sg.neut es, des (demonst.) /es/, /des/ -s /s/
1pl mir /mir/ -ma /mer/
2pl ihr, es /ir/, /s/ -(t)s /ts/
3pl sie, die (demonst.) /si/, /di/ -s /s/
The enclitic forms of the 2nd person nominative pronouns (2sg -st, 2pl -ts) exhibit
special properties which set them apart from the other pronominal clitics. The partic-
ular status of the 2nd person forms has received quite some attention in the literature
(cf. e.g. Pfalz 1918; Merkle 1975; Bayer 1984; Altmann 1984; Zehetner 1985; Werner
1988; Harnisch 1989; Wiesinger 1989; Nbling 1992; Rowley 1994; Abraham 1995;
Wei 1998, 2002). The special behavior of the 2nd person clitics is commonly taken
to indicate that they are actually not pronominal elements. Instead, they are often an-
alyzed as pieces of inection which attach to other elements in the left periphery of
the clause.
1
The particular properties of the 2nd person clitics are illustrated in the
following.
First, there are a number of syntactic (i.e., distributional) differences between 2nd
person forms and the other pronominal clitics. In contrast to the latter, the 2nd person
forms are obligatory in all contexts, that is, they cannot be simply replaced by the
relevant full forms:
(1) a. ob-st
whether-2sg
noch
to
Minga
Munich
kumm-st
come-2sg
b. *ob
whether
du
you.sg
noch
to
Minga
Munich
kumm-st
come-2sg
Whether you come to Munich
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.3 (197-278)
Chapter 5. The reanalysis of C-oriented clitics
(2) a. ob-ts
whether-2pl
noch
to
Minga
Munich
kumm-ts
come-2pl
b. *ob
whether
es/ihr
you-pl
noch
to
Minga
Munich
kumm-ts
come-2pl
In addition, a full 2nd person subject pronoun is only acceptable if it co-occurs with
the relevant clitic form, giving rise to instances of subject clitic doubling:
(3) a. ob-st
whether-2sg
du
you.sg
noch
to
Minga
Munich
kumm-st
come-2sg
b. ob-ts
whether-2pl
es/ihr
you.pl
noch
to
Minga
Munich
kumm-ts
come-2pl
This contrasts with the behavior of the other subject enclitics, which are not obligatory
and which cannot be doubled by full pronouns:
2
(4) a. ob=e
whether=clit.1sg
(*i)
I
noch
to
Minga
Munich
kumm
come-1sg
b. ob
whether
i
I
noch
to
Minga
Munich
kumm
come-1sg
Whether I come to Munich
Furthermore, Altmann (1984) notes another difference between the subject enclitics
for 2nd person and the other enclitic forms, having to do with the way the presence of
a clitic affects the distinction between V1 and V2 structures. When a 1st or 3rd per-
son subject enclitic follows a clause-initial nite verb, the structure is unambiguously
interpreted as a V1 clause:
(5) k
h
enn=6
know=clit.1sg
enngg
you.pl
w.
modprt
I know you guys. (Altmann 1984: 206)
However, if a 2nd person form is attached to a clause-initial verb, a sentence such as
(6a) is ambiguous between a V1 and a V2 structure. This ambiguity can be resolved if
a full pronoun is added, as in (6b):
(6) a. hfbbd-s
have-2pl
k
h
!
no
geId
money
nImm!.
not-anymore
You have no money anymore.
b. hfbbd-s
have-2pl
e:s
you.pl
k
h
!
no
geId
money
nImm!.
not-anymore
You have no money anymore. (Altmann 1984: 207)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.4 (278-345)
Table 2. 2nd person full/reduced pronominal forms and verbal agreement.
Full pronoun Enclitic pronoun Verbal agreement
2sg du -st -st
2pl es/ihr -ts -ts
Altmann concludes that the different behavior of 2nd persons forms follows directly
fromtheir inectional character: since they are always obligatorily present, they cannot
serve to distinguish between different word order or clause types.
Apart from these syntactic differences, we can also observe some crucial mor-
phophonological differences that underline the special status of 2nd person forms,
cf. Table 2. First, in contrast to the other clitics, the 2nd person forms cannot be derived
from the relevant full pronouns via phonological reduction. Second, the 2nd person
clitics are identical with the relevant verbal agreement sufxes, which already suggests
that they are in fact inectional elements. Accordingly, Pfalz (1918), Bayer (1984),
Altmann (1984), Zwart (1993b), and Wei (1998, 2002, to appear) (among others)
propose that the special properties of the 2nd person subject clitics can be explained
on the assumption that these elements are in fact some form of inection located in
C. This view is also adopted in the present work. In line with the analysis proposed in
Chapter 3, the 2nd person enclitics are analyzed as exponents of a dissociated agree-
ment morpheme that attaches post-syntactically to C (henceforth Agr-on-C). Note,
however, that this assumption leaves an open question, namely why complementizer
agreement is restricted to 2sg, 2pl forms. This question will be addressed in some more
detail in Chapter 6.
Another consequence of this analysis is that Bavarian is considered a partial pro-
drop language (Bayer 1984; Wei 1998): sentences which contain only a 2nd person
clitic have to be analyzed as instances of pro-drop (according to Bayer 1984, the overt
manifestation of agreement in C (with 2sg, 2pl) serves to license referential pro in
present-day Bavarian):
(7) a. Kumm-st
come-2sg
pro noch
to
Minga,
Munich
dann
then
muas-st
must-2sg
pro me
me
bsuacha.
visit
b. Kumm-ts
come-2pl
pro noch
to
Minga,
Munich
dann
then
mass-ts
must-2pl
pro me
me
bsuacha.
visit
(8) a. *Kumm
come-1sg
pro noch
to
Minga ...
Munich
If I come to Munich, ...
b. *Kumm-t
come-3sg
pro noch
to
Minga?
Munich
Will he/she/it come to Munich?
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.5 (345-408)
Note that in examples such as (7), there is no double afxation of agreement morphol-
ogy to verbs that move to C via T, that is, forms such as kummst-st come-2sg-(clit)2sg
are ungrammatical. On the assumption that -st and -ts are clitics, one has to resort to
some sort of phonological lter to rule out the sequence st-st (and ts-ts) that should
be possible after cliticization to an inected verb has taken place. However, a more
principled (and interesting) explanation becomes available if the 2nd person encli-
tics are treated as agreement markers, as suggested. Then, the absence of forms such
as kummst-st can be analyzed in terms of a general morphological condition that re-
stricts the number of overtly realized (identical) Agr-morphemes on the verb to one,
see Chapter 3 above. This lends further support to an analysis of the 2nd person clitics
as pieces of inection.
3
Again, the fact that pro-drop is restricted to 2nd person contexts is somewhat mys-
terious. For example, it cannot be attributed to some special morphological property
of the 2nd person agreement sufxes, in the sense that 2nd person forms are morpho-
logically stronger or more distinctive than 1st or 3rd person forms, which enables
only the former but not the latter to license pro:
Table 3. Verbal agreement paradigm
(pres. indic.) of Bavarian.
Verbal agreement
1sg -
2sg -st
3sg -t
1pl -an
2pl -ts
3pl -an
Table 3 shows that the only non-distinctive (i.e., homophonous) agreement endings
are 1pl and 3pl, whereas all other endings serve to unambiguously signal person and
number of the subject. Thus, it appears that the restrictions on complementizer agree-
ment and pro-drop observed in Bavarian escape a deeper synchronic account and
can only be captured by a stipulation that simply rules out the non-existing cases (cf.
Bayer 1984).
In this work, I argue that the restrictions in question can be more fully under-
stood if diachronic evidence is taken into account. More specically, it is claimed that
the restrictions on pro-drop and complementizer agreement follow from a set of syn-
tactic and morphological factors that determined the reanalysis of subject clitics as
verbal agreement markers in the history of Bavarian (see Chapter 6 for an explanation
of the person/number restrictions in terms of morphological factors which shaped
the development of Agr-on-C). In this section, it is shown that this reanalysis could
only take place in contexts with V-to-C movement, forcing the learner to assume the
existence of (i) pro-drop and (ii) an agreement morpheme on C, leading to com-
plementizer agreement. This analysis of the diachronic facts is based on the analysis
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.6 (408-494)
of complementizer agreement developed in Chapter 3, where the presence of inec-
tion in the C-domain was attributed to the post-syntactic insertion of a dissociated
Agr-morpheme at Morphological Structure.
.. The diachronic development of Agr-on-C in Bavarian
In all varieties of Bavarian, C-oriented enclitics were reanalyzed as enlargements of
existing agreement markers, giving rise to new agreement sufxes for 2nd person (cf.
Bayer 1984; Wiesinger 1989). Table 4 below lists the inherited verbal agreement end-
ings and the forms resulting from the reanalysis of the pronominal clitics for 2sg, t(hu),
and 2pl, (e)s (originally a dual that developed into a 2pl pronoun in Bavarian).
Interestingly, it can be shown that this grammaticalization process was initially
conned to certain syntactic environments. More specically, the relevant data suggest
that the historical development of the new 2nd person agreement sufxes affected rst
nite verbs in C and spread later to other verbal positions.
... 2sg /-st/
The development of 2sg /-st/ began already in early Old High German (OHG), pre-
sumably during the 9th century AD. The resulting new ending is found in all modern
German varieties. It is commonly assumed (cf. Brinkmann 1931; Braune 1950: 252;
Sommer 1994) that the agreement formative 2sg -s+t resulted from a reanalysis of the
combination of verb and clitic pronoun t(hu) in inversion contexts, possibly on the
analogy of the preterite-presents, which already showed /-st/ for the 2sg present in-
dicative (kanst can, tarst dare, muost must, weist know etc.), and the 2sg of be bist,
which resulted from an independent and earlier development (cf. Lhr 1984).
4
Exam-
ples such as (9) from the early OHG Tatian seem to suggest that the new agreement
ending was initially conned to verbs that have undergone V-to-C movement:
(9) Ih
I
forahta,
feared
[
CP
uuanta
since
thu
you
grim
grim
man
man
bist,
are
[
C
nimi-st
take-2sg
[
IP
thaz
that
thu
you
ni
neg
szto-s]]]
plant-2sg
inti
and
[
CP
[
C
arno-st
earn-2sg
[
IP
thaz
that
thu
you
ni
not
sto-s.]]]
sow-2sg
Since you are a grim man, I feared that you take what you havent planted
and earn what you havent sowed. (Tatian 151,7; Sievers 1961: 228)
In (9), we can see that the new ending shows up on verbs that occupy the C posi-
tion in V2 contexts, whereas nite verbs in clause-nal position still exhibit the older
ending /-s/.
5
However, a second look at the relevant data reveals that the picture is ac-
Table 4. Old and new agreement sufxes for 2nd person in Bavarian.
Old inherited ending New enlarged ending Lexical source
2sg -s -s+t t(hu)
2pl -t -t+s ( e)s
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.7 (494-590)
tually more complex. In the following example, also taken from the OHG Tatian, the
nite verb (ges go-2sg) does not show the enlargement /-t/ despite the fact that it has
undergone inversion with the subject pronoun thu:
6
(10) Ni
not
ges
go
th
you
thanan
from-there
z,
out
r thanne
before
thu
you
giltis
return
then
the
iungiston
newest
scz.
treasure
You will never get out until you have paid the last penny. (Tatian, , 27.3;
Sievers 1961: 51)
In general, it appears that the OHG Tatian reects an early stage of the develop-
ment of the 2sg formative /-st/. The following table shows that the old ending /-s/
by far outnumbers the enlarged form /-st/, even in inversion contexts (cf. Sommer
1994: 328):
Table 5. 2sg endings in the OHG Tatian
(without bist be.2sg).
Total
-s 166
-s thu 15
-st 26
-st thu 4
-st(t)u 9
As noted in Chapter 2 above, this kind of variation can be analyzed as an instance of
grammar competition (Kroch 1989), where speakers have access to more than one in-
ternalized grammar. In the case at hand, the relevant grammars presumably differ with
respect to the presence of an agreement morpheme in the C domain (giving rise to the
ending /-st/). Of course, settling this matter requires a detailed quantitative analysis
of the OHG texts in question (which is a topic for future research; see Chapter 6 for
morphological aspects of the development of 2sg /-st/ in OHG/Bavarian). Neverthe-
less, it seems reasonable to suppose that the historical development of the new verbal
agreement sufx /-st/ proceeded via the following stages:
(11) a. V + enclitic V+Agr/inversion contexts (reanalysis of clitics as Agr-on-
C)
b. extension to other elements located in the C-system such as complemen-
tizers, relative pronouns etc. (conned to Bavarian)
c. extension to verbs in clause-nal position
Note that at least with respect to the development of 2sg /-st/ there is not much empir-
ical data available which can be used to evaluate the hypothesis in (11), due to the fact
that the change in question predates the earliest Bavarian records. Still, even with no
relevant empirical evidence available, the order of events assumed in (11) seems to be
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.8 (590-616)
well-motivated from a conceptual point of view. First, it is rather uncontroversial that
sentences with inversion provided the only context where a pronoun could be reana-
lyzed as part of the verbal agreement morphology, since only in inversion contexts, the
pronoun was adjacent to the nite verb in early Bavarian/OHG. Thus, it is fairly clear
that the new ending could not develop rst on clause-nal verbs. In other words, stage
(11a) must precede stage (11c). What about the chronological ordering of (11a) and
(11b)? One might assume that enclitics were directly reanalyzed as agreement mark-
ers on complementizers, claiming that stage (11b) in fact might have preceded (11a)
or developed simultaneously. However, I think that it is more natural to assume that
agreement afxes spread from verbs (in inversion contexts) to complementizers rather
than the other way around (cf. Altmann 1984: 200 for a similar view). In general we
should expect new patterns of verbal agreement to develop rst on verbs. It would be
surprising if complementizers were leading the charge.
7
This is also in line with the
idea that during language acquisition, the relevant trigger experience comes from sim-
ple non-embedded contexts only (so-called Degree-0-Learnability, Lightfoot 1991). As
a consequence, properties of embedded contexts must be deducible from properties of
non-embedded contexts (however, see below for some qualifying remarks). Therefore,
we should not expect the development of complementizer agreement to predate the
development of new verbal agreement in inversion contexts. Next, it is shown that the
set of hypotheses stated in (11) receives further support by later developments that
affected the agreement markers for 2pl and 1pl in the history of Bavarian.
... 2pl /-ts/
As already noted above, in Bavarian, the original inherited ending for 2pl, /-t/, (which
is still found today in most other German varieties) was replaced by /-ts/ via a rein-
terpretation of the enclitic form (-s) of the pronoun es, originally a dual. It is usu-
ally assumed that this grammaticalization process began in the 13th century (cf. e.g.
Wiesinger 1989). Again, it seems that the new ending appeared rst on nite verbs in
inversion contexts. Moreover, in the case of 2pl /-ts/, there are further indications sup-
porting the sequence of historical stages assumed in (11) above. In other words, it can
be shown that after the initial development of a new agreement marker which shows
up on the nite verb in inversion contexts, the new ending rst spread to other poten-
tial hosts in the C-domain such as complementizers, wh-phrases, or relative pronouns
before it was generalized to verbs in clause-nal position. Pfalz (1918: 232) observes
that in some northeastern Bavarian dialects (spoken in the Lauterbach and Sanger-
berg area), the new ending for 2pl (/-ts/) still attaches only to conjunctions and verbs
in C, but crucially not to verbs which occur in clause-nal position:
(12) wou-ts
where-2pl
teham
at-home
sat
are
...where you are at home
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.9 (616-688)
(13) [
CP
wei-ts
when-2pl
iw6
over
tpruk
the-bridge
khumt
come-2pl
[
C
sea-ts
see-2pl
[
IP
swi6tshaus]]]
the-tavern
When you cross the bridge, you see the tavern.
These facts suggest that the new agreement marker was rst grammaticalized as a re-
alization of the feature content of C (attaching to verbs and other elements located
in the C-system) and spread later to verbs in clause-nal position.
8
Further sup-
port for the claim that the change in question proceeded along these lines (i.e., via
a transitional stage where the new ending was conned to the C-domain) comes from
Lower Bavarian dialects which exhibit a related grammaticalization process affecting
1pl forms.
... 1pl /-ma/ in Lower Bavarian
In a set of Lower Bavarian dialects, the 1st person plural subject enclitic -ma devel-
oped in a similar way as the 2nd person enclitics (cf. Pfalz 1918; Merkle 1975; Bayer
1984; Altmann 1984; Kollmer 1987; Wiesinger 1989; Rowley 1994; Abraham 1995;
Wei 1998, 2002, to appear). That is, the pronominal enclitic 1pl -ma shows a sim-
ilar behavior as the 2nd person forms: it is obligatory in all contexts, it can be doubled
by full forms for emphatic reasons, and 1pl contexts license pro-drop in these dialects
(Bayer 1984: 252):
9
(14) a. wem-ma
when-1pl
aaf
to
Minga
Munich
fon
drive
b. wem-ma
when-1pl
mia
we
aaf
to
Minga
Munich
fon
drive
c. *wem
when
mia
we
aaf
to
Minga
Munich
fon
drive
When we drive to Munich (Wei 2002: 9)
(15) a. Mia
we
fom-ma
drive-1pl
hoam.
home
We go home. (Wei 2002: 9)
b. *Mia
we
fon
drive
hoam.
home
We go home. (Helmut Wei, p.c.)
(16) Fom-ma
drive-1pl
pro noch
to
Minga?
Munich
Will we go to Munich?
Therefore, it is plausible to assume that /-ma/ developed into an additional instance of
Agr-on-C(cf. Bayer 1984; Wei 1998, 2002, to appear). This claimis further supported
by the fact that in V2 clauses, /-ma/ replaces the original agreement ending in bisyllabic
verbs such as laffa to run, gengan to go, soucha(n) to seek etc. (cf. Kollmer 1987;
Wei 2002):
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.10 (688-753)
(17) a. Mia
we
laff-ma/*laff-a
ran-1pl/ran-1pl
hoam.
home
We are running home.
b. Mia
we
gem-ma/*geng-an
go-1pl/go-1pl
hoam.
home
We are going home.
Crucially, no such replacement is possible when the nite verb shows up in sentence-
nal position:
(18) wa-ma
because-1pl
hoam
home
laff-a/*laff-ma.
go-1pl
Because we are going home
In other words, the dialects in question show a complementary distribution of the
new sufx /-ma/ and the old ending for 1pl, /-a(n)/ (cf. Kollmer 1987: I, 357): /-ma/
systematically appears on verbs in V2 clauses (main and embedded contexts), cf. (17),
whereas verbs in sentence-nal position maintain the old ending /-a(n)/, cf. (18).
As already noted above in Note 3, a similar development is exhibited by a number
of Carinthian dialects (Lessiak 1963; Wiesinger 1989; Abraham 1995). In the following
examples, we can observe a multiple realization of 1pl subject marking: (i) via the end-
ing -mr/-m6 on the verb, (ii) via a C-oriented clitic, and (iii) via an overt full pronoun
(similar to West Flemish, cf. Chapter 3, Section 3.7):
(19) a. Kher-m6=mr
belong-1pl=clit.1pl
wir
we
a aufn?
up
Do we belong at the top?
b. wi6-mr=mr
as-1pl=clit.1pl
wir
we
wln
want
The way we want (it) (Pernegg, Lessiak 1963: 204)
This clearly shows that the ending -mr/-m6 cannot be a clitic it would be quite odd
to have two clitics in these sentences and accordingly must also be analyzed as an
agreement marker, that is, another instance of Agr-on-C.
10
Furthermore, it seems that
at least in some dialects, the new agreement ending has spread to verbs in clause-nal
position, replacing the original agreement ending in all contexts:
(20) ... wos
what
m#r
we
wl-m#r
want-1pl
(Wiesinger 1989: 38)
Interestingly, an analogous development has apparently taken place in a subset of the
Lower Bavarian dialects as well. In some varieties spoken in the Bavarian Forest,
11
/-ma/ appears on clause-nal frequently used (light) verbs such as have and do
(Kollmer 1987: I, 357; Wiesinger 1989: 38; Wei 2002: 9). Note that /-ma/ must be an-
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.11 (753-777)
alyzed as an agreement marker in the following examples, since enclitics cannot attach
to clause-nal verbs in Bavarian.
(21) a. dass-ma
that-1pl
(mia)
we
ko
no
geid
money
ned
not
h-ma [instead of 1pl h-n]
have-1pl
That we have no money (Kollmer 1987: I, 362)
b. we-ma
if-1pl
(mia)
we
des
that
ned
not
dou-ma ... [instead of 1pl dou-n]
do-1pl
If we dont do that... (Kollmer 1987: I, 358)
The above discussion has shown that in Bavarian, new verbal agreement morphol-
ogy developed from enclitic subject pronouns in inversion contexts. With respect to
2nd person forms, the available evidence suggests that the new 2nd person agreement
markers developed rst on verbs located in C and spread to other verbal positions in a
subsequent development. The latter claim is supported by data from Lower Bavarian
dialects where the new agreement marker for 1pl, /-ma/, is obligatory on elements in
C (i.e., on verbs and complementizers), but still excluded on most clause-nal verbs.
.. Clitics, V2, and the rise of agreement
The basic proposal for the rise of new verbal agreement markers in Bavarian is illus-
trated by the phrase markers in (22) and (23). That is, I assume that the clitic pronoun
that attaches to C in the target grammar, cf. (22), is reanalyzed as the realization of an
Agr-morpheme adjoined to C in the learners grammar, as shown in (23):
(22)
C
T
TP C + V
fin
DP
i
P T
t
i
CP
Topic
D
clit.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.12 (777-834)
(23)
C
T Agr
TP C
C + V
fin
P T
pro
CP
Topic
In the following, I will discuss the changes leading to the structure in (23) in some
more detail, focusing on the set of syntactic conditions on this kind of structural
reanalysis argued for in Chapter 4 (briey repeated in (i)(iv) below).
(i) Structural simplication the resulting structure must be less complex than the
target structure.
(ii) Preservation of argument structure the -role previously assigned to the pronom-
inal clitic must be assigned to another element after the reanalysis.
(iii) Identication of feature content the content of the newly created Agr-morpheme
must be valued.
(iv) Word building constraint a bound Agr-morpheme must be able to combine with
the verb prior to Vocabulary Insertion.
As shown in (23) above, I assume that the reanalysis in question satises the re-
striction imposed by -theory (Preservation of argument structure) by assigning the
subject/agent -role previously assigned to the clitic pronoun to a referential pro that
is inserted into SpecP. This also serves to meet the requirement of structural sim-
plication, since the empty subject pronoun stays behind in SpecP throughout the
derivation.
12
The referential pro is Case-licensed by an Agree operation initiated by the
X
0
-complex T+Agr (prior to movement of V+T+Agr to C). This operation also serves
to value the non-interpretable -set of the canonical verbal agreement morpheme
which is part of to T
0
.
13
As already noted above, the rise of verbal agreement from
former pronominal elements may in this way give rise to new pro-drop properties for-
merly absent in the target grammar. In short, the reanalysis of the former subject clitic
as an agreement marker forced the learner to assume the presence of a referential pro in
the subject position, which is the historical source of the limited pro-drop properties
of present-day Bavarian (cf. Bayer 1984 on pro-drop in Bavarian; Wei 2002 for a re-
lated proposal). The rise of referential pro-drop was presumably facilitated by the fact
that Bavarian is not a strict non-pro-drop language. For example, the subject position
remains empty in impersonal passives, as illustrated in (24). In other words, Bavarian
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.13 (834-890)
does not require the insertion of an expletive subject pronoun in these contexts in
contrast to English, but similar to Standard German (cf. Grewendorf 1989).
(24) Heit
today
wird
is-pass
geoabeid.
worked
(Gnther Grewendorf, p.c.)
Note that in a V2 language, the reanalysis in question trivially fullls the requirement
that the resulting Agr-morpheme must combine with the verb prior to Vocabulary
Insertion (the Word building constraint), since subject enclitics are always adjacent to
the verb in inversion contexts. Moreover, it seems that the V2 property in fact provides
the only pathway to new verbal agreement morphology in languages with C-oriented
clitics and basic SOV order like Bavarian: in an SOV+V2 grammar, a C-oriented clitic
cannot be reanalyzed as the realization of an Agr-morpheme on C that combines via
Morphological Merger with a verb that occupies a lower head position, since these
heads are usually not adjacent at MS (e.g., the subject regularly intervenes between C
and T). Hence, overt movement of the nite verb to C is a necessary precondition for a
reanalysis that leads to the existence of verbal agreement features on C (in this way, the
presence of the nite verb in C can be said to signal that C is capable of hosting an Agr-
morpheme). To conclude, the rise of new agreement formatives (in C) appears to be
intimately linked to the V2 property. This corresponds to the observation that cross-
linguistically, complementizer agreement/Agr-on-C is a marked phenomenon which
goes hand in hand with another marked syntactic property, V2.
The V2 property arguably plays also an important role in extending the domain
of the new endings to other verbal positions. Recall that the change in question did
not immediately lead to a wholesale replacement of the old ending (i.e., the phonolog-
ical exponent of Agr-on-T) by the new one. Rather, it proceeded via a stage where the
new agreement ending was conned to C (as the phonological realization of Agr-on-
C), presumably due to the fact that embedded clauses (without V-to-C) still provided
enough evidence for the old ending. Thus, initially, the exponent of Agr-on-C replaced
the old agreement ending only in inversion contexts.
14
The new inection then grad-
ually spread to other verbal positions when the learner reanalyzed the exponent of
Agr-on-C as the canonical verbal agreement ending, that is, as the exponent of Agr-
on-T. The latter reanalysis depended on the V2 property as well: only instances where
the original agreement ending was replaced by the exponent of Agr-on-C could feed
a possible misinterpretation of the latter as the realization of Agr-on-T. Schematically,
this later development is illustrated for the change from 2pl /-t/ to /-ts/ in (25):
(25) [2pl/Agr-on-C] /-ts/ [2pl/Agr-on-T] /-ts/
(in inversion contexts)
Thus, the phonological exponent of Agr-on-C is eventually reanalyzed as the exponent
of the canonical agreement morpheme, Agr-on-T. It is important to note that this later
development is a counterexample to the claim that grammaticalization always leads to
new exponents for higher functional heads (Roberts & Roussou 2003, see Chapter 2
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.14 (890-961)
above). Instead, it seems that in V2 languages, the exponent of a higher (dissociated)
Agr-morpheme may be reanalyzed as a marker of canonical subject-verb agreement,
that is, as an exponent of a lower agreement morpheme, Agr-on-T.
So far, we have not addressed the question of how the content of the newly created
Agr-morpheme on C is identied, that is, how the reanalysis in question can satisfy
the condition that the feature content of the resulting Agr-morpheme must be li-
censed (Identication of feature content). Recall that in Chapter 3, I have provided a
number of arguments against an analysis in which the agreement morpheme on C is
valued by accessing the subjects -set via an Agree operation. Instead, I presented an
analysis where Agr-on-C is inserted post-syntactically as a dissociated Agr-morpheme
that is licensed under structural adjacency with an agreement morpheme on T which
has been valued in the syntactic derivation (via Agree). Accordingly, I assume that
the subject clitic is reanalyzed as the realization of a dissociated Agr-morpheme that
is post-syntactically adjoined to C and licensed under structural adjacency with Agr-
on-T, thereby fullling the condition Identication of feature content. In this way, the
post-syntactic insertion of dissociated morphemes provides an alternative way of rean-
alyzing a former movement dependency in terms of a less costly syntactic structure,
in addition to external Merge (Roberts & Roussou 2003). The various aspects of the
change in question can be summarized as follows:
(26) [
CP
XP [
C
C+V+pronoun
i
[
TP
t
i
[
T
T [
P
t
i
...
[
CP
XP [
C
C+V+Agr [
TP
T [
P
pro ...
assigned to pro
Identication of feature content: dissociated Agr-morpheme
licensed under structural
adjacency with Agr-on-T
movement)
This analysis of the grammaticalization of new agreement markers in the history of
Bavarian raises the question of whether the reanalysis of clitics as dissociated Agr-
morphemes constitutes a necessary intermediate stage of the development of new
verbal agreement formatives (i.e., exponents of Agr-on-T) in languages that already
exhibit a series of verbal agreement markers. The presence of more than one Agr-
morpheme should then give rise to instances of multiple agreement where the subjects
-set is reected in more than one position (see also Chapter 2, Note 28, and Chapter
4, Note 17, on the development of new agreement markers from In-oriented clitics
in Northern Italian dialects).
Initial support for this hypothesis comes from the diachronic facts observed in
Bavarian, where the new agreement endings started on C as exponents of a disso-
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.15 (961-1010)
ciated morpheme and replaced the original exponents of Agr-on-T in a subsequent
development. In addition, it appears that the difference between clitics and dissociated
morphemes is rather subtle and therefore easily missed during language acquisition.
Note that dissociated Agr-morphemes share basic properties with clitics such as post-
syntactic positioning (for post-syntactic accounts of clitic placement cf. Bonet 1991;
Halpern 1992; Schtze 1994, among others). In Bavarian, the clitic-like behavior of
Agr-on-C is also evident in cases where the agreement morphology does not attach
to an element in C, but rather to an XP located in SpecCP. The following examples
illustrate that in the absence of a lled C-head, the inection can attach to any element
that occurs in the left periphery of the clause such as DPs (27a), adjectives (27b), or
adverbs (27c):
(27) a. Du
you
soll-st
should-2sg
song
say
[
CP
[ an wichan
which
Schuah]-st
shoe-2sg
[
IP
du
you
wui-st]]].
want-2sg
You should say which shoe you want.
b. [
CP
[ Wia
how
oit]-ts
old-2pl
[
IP
ihr/es
you
sei-ts]]
are-2pl
is
is
mir
for-me
wurscht.
not-important
How old you are makes no difference to me.
c. [
CP
[ Wia
how
schni]-ts
fast-2pl
[
IP
ihr/es
you.pl
fahr-ts ]]!
drive-2pl
How fast you drive! (Bayer 1984: 235)
Note that this behavior of Agr-on-C in Bavarian seems to contrast with the properties
of well-behaved inectional afxes, which usually select for a unique host/syntactic
category they attach to.
15
The systematic differences between dissociated Agr-morphemes
and Agr-morphemes which are present in the syntactic derivation (adjacency effects,
weak morphological selection, interaction with PF-processes such as deletion or clitic
placement) can be used as a diagnostics for telling apart the two types of agreement.
The similarities between clitics and dissociated morphemes presumably abet a reanal-
ysis of the former as the latter, since they blur the boundary between these elements.
Thus, it seems that at least in Bavarian, the development of new agreement end-
ings proceeded via an initial reanalysis of subject clitics as exponents of a dissociated
Agr-morpheme (on C). Whether this change represents a necessary stage on the di-
achronic pathway toward canonical argument-predicate agreement across languages is
discussed in more detail in the sections to come.
.. Developments in other German varieties
In a number of German dialects spoken in minority enclaves surrounded by Italian-
speaking communities, we can observe grammaticalization processes that clearly re-
semble the historical developments in Bavarian. In the following, I will present data
from Cimbrian, a variety of Bavarian spoken in a small number of villages in the Ital-
ian Alps (southeast of Trento) and a set of Alemannic dialects spoken in Northern
Italy. Without giving further arguments, I will assume that the relevant developments
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.16 (1010-1087)
can be analyzed on a par with the Bavarian examples discussed in the previous section
(with some qualications added where it seems necessary). Thus, the following data
are taken to represent further examples of enclitics turning into instances of Agr-on-C
in inversion contexts.
... Cimbrian
Cimbrian is a Germanic (Bavarian-Austrian) language spoken by about 2,000 speak-
ers in Northeast Italy in Lusern, the Sette and Tredici Comuni (Sieben and Dreizehn
Gemainde, southeast of Trento), Giazza (Glietzen, Ljetzen), and Roana (Rabam). Cim-
brian is related to Bavarian and shows many archaic features which date back to the
time when the area in question was settled (in the 11th and 12th century). Note,
however, that there are many differences between present-day Bavarian and Cimbrian
concerning grammar, lexicon and pronunciation which justify a classication of Cim-
brian as a separate Germanic language in its own right. The following discussion is
based on the unpublished Zimbrische Gesamtgrammatik by Bruno Schweizer (1952,
Vol. 5: Syntax der zimbrischen Dialekte in Oberitalien), and Tyroller (2003).
Some Cimbrian varieties such as Lusern are characterized by the V2 property,
that is, the subject inverts with the nite verb if a non-subject is fronted, as shown
in example (28).
Lusern
(28) disan
this
libar
book
hat
has
ar
he
gelest,
read,
dar
the
pua.
boy
The boy has read this book. (Tyroller 2003: 226)
In general, however, the Cimbrian V2 dialects allow more variation concerning the
placement of the nite verb, including V1 and V3 order in contexts where other Ger-
man varieties require V2 order (see Schweizer 1952; Tyroller 2003 for details). In other
varieties of Cimbrian, the V2 property has been (at least partially) lost, probably due
to Italian inuence, cf. Heller (1975) on the Sette Comuni and Schweizer (1952: 25)
on the Tredici Comuni. In contrast to German, all Cimbrian dialects exhibit basic
VO word order. This is illustrated by the following example from the variety spoken
in Lusern:
Lusern
(29) I
I
hn
have
s
it
as
you.pl
kht,
said
ke
that
dar
you.pl
mucht
must
machan
make
lautrar
more-uid
di
the
malt.
mortar
I told you that you that the mortar must be more uid. (Tyroller 2003: 234)
All Cimbrian varieties exhibit a series of enclitic pronouns that attach to the right of
the nite verb in inversion contexts and to the complementizer (or other elements of
the C-domain) in embedded clauses (cf. Schweizer 1952: 26, 225). Schweizer (1952: 4,
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.17 (1087-1150)
26) observes that the enclitics enter into a close relation with the nite verb, which
sometimes gives the impression of amalgamated forms:
Roana
(30) a. morgen
tomorrow
vr
morning
ge=x
go=clit.1sg
bar
over
van
from
Roan.
Roana
Tomorrow morning, I will cross over from Roana.
b. ba
where
pis=to
are=clit.2sg
gabest
been
hte?
today
Where have you been today? (Schweizer 1952: 26)
Interestingly, the clitic pronouns that attach to the inverted nite verb can be doubled
by full pronominal forms. According to Schweizer (1952: 4, 28), this should be taken
to indicate that the enclitics have developed into a new form of verbal inection which
he calls Pronominalkonjugation:
16
Roana
(31) a. ix
I
gan
to
ux
you.pl
kim-ex
come-1sg
nemmear.
never again
I will never come to you again.
b. bazt
What
s-te
do-2sg
du?
you
What are you doing?
c. denne
then
miz-er
must-3sg
gen
go
ear.
he
Then he must leave.
d. hat-er
hat-3sg
gahat
had
tzo
to
megelan
marry
ear
he
ox.
too
He had to marry, too.
e. bir
we
t-ber
do-1pl
nixt.
nothing
We do nothing.
Giazza
(32) a. pin-i
am-1sg
nixt
not
ploax
pale
haute
today
i?
I
Am I not pale today?
b. has-to
have-2sg
du
you
nixt
not
gahoart
heard
lautan?
ring
Havent you heard the bells ring?
Lusern
(33) dopo
then
mchten-ber
wish-1pl
essan
eat
biar.
we
(Dal Negro 2004: 179)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.18 (1150-1217)
Importantly, it seems that this form of clitic doubling is restricted to non-embedded
clauses (cf. Schweizer 1952: 26ff.), which suggests that the various Cimbrian dialects
exhibit an early stage of the grammaticalization of new verbal agreement markers
where the newly created agreement morpheme on C is still conned to matrix clauses
with subject-verb inversion, that is, the environments where the reanalysis of cli-
tics originally took place. In other words, Cimbrian exhibits another instance of the
structural reanalysis schematized in (26) above. In contrast to Bavarian, however, the
relevant grammaticalization process is still conned to stage (34a):
(34) a. V + enclitic V+Agr/inversion contexts
(reanalysis of clitics as Agr-on-C)
b. extension to other elements located in the C-system such as complemen-
tizers, relative pronouns etc.
c. extension to verbs in clause-nal positions
The hypothesis that Cimbrian exhibits an early stage of the change in question is fur-
ther supported by the fact that the clitics are not yet obligatory. Thus, in contrast to
Bavarian (see above), clitic pronouns can still be replaced by the relevant full forms,
cf. the following examples from Roana and Lusern:
Roana
(35) a. ix
I
pin
am
bolaibet
stayed
segen...
to-see
I stayed in order to see... (Schweizer 1952: 18)
b. han
have
ix
I
gavunt
found
an
a
hasen
rabbit
I found a rabbit. (Schweizer 1952: 24)
Lusern
(36) a. I
I
pin
am
gest
stayed
bachant
awake
da
the
gnz
whole
nacht.
night
I stayed awake during the whole night.
b. Bar
we
hm
have
gepitet
waited
viar
four
urn.
hours
We waited for four hours. (Tyroller 2003: 215)
The variation between new enlarged forms resulting from the reanalysis of pronouns
and old, non-enlarged endings presumably represents an instance of grammar com-
petition (cf. Kroch 1989). That is, the speakers have command over two internalized
grammars which are characterized by the absence vs. presence of a dissociated Agr-
morpheme on C, respectively. Over time, one of the grammatical options will win out
over the other. If the change proceeds in a fashion parallel to the developments in
Bavarian, this might eventually lead to a grammar in which the exponent of Agr-on-C
is generalized to all verbal positions.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.19 (1217-1286)
... Walser dialects
Hotzenkcherle (1971), Nbling (1992: 257ff.), and Dal Negro (2004) report on a
number of Walser German (Alemannic) varieties spoken in Northern Italy where en-
clitic pronouns apparently have developed into new markers of verbal agreement. For
example, in the local dialects of Rima and Rimella, the grammaticalization process
in question has led to new agreement endings for 1pl (-wer/-war) and 2pl (-(ed)er/-
ar), which are obligatory and clearly resemble the relevant pronominal forms, cf. the
following examples from Rimella (taken from Nbling 1992: 257):
17
Rimella
(37) a. endsch
we
andre
others
ber-wer ...
carry-1pl
We carry...
b. ier
you.pl
andru
others
ber-eder ...
carry-1pl
You (pl) carry...
Dal Negro (2004: 161) notes that similar examples of doubling can be observed in
other person/number combinations as well. However, in contrast to 1pl and 2pl forms,
the enclitics are not obligatory here. Still, they presumably represent early instances
of a grammaticalization process in which the enclitic is reanalyzed as an agreement
morpheme on C:
Rimella
(38) ix
I
tjn-ex
do-1sg
setsu.
sit
I am sitting.
Rima
(39) sch
they
gn-sch.
go-3pl
They go.
Apparently, the new endings developed rst on highly frequent, short (often mono-
syllabic) verbs such as do, go, have, sit etc.
18
A related development, albeit conned
to 1pl forms, took place in another Walser dialect, Bosco Gurin, spoken in Switzerland.
Again, the new agreement ending for 1pl (-w) is clearly derived from the correspond-
ing pronoun (wir):
Bosco Gurin
(40) wir
we
scht-w
stand-1pl
dsch
the.gen
morgn-dsch
morning-gen
f,
up
um
at
fy.
ve
We get up at ve in the morning. (Dal Negro 2004: 162)
Recently, Dal Negro (2004) has demonstrated that a similar development can be ob-
served in Pomattertitsch, a Walser dialect spoken in Formazza (Italy/Piedmont) which
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.20 (1286-1340)
is closely related to the Walser varieties discussed above. Similar to Bavarian, Cimbrian,
and the other Walser dialects, the grammaticalization process in question is apparently
triggered in inversion contexts, where enclitic pronouns are reanalyzed as agreement
sufxes on the verb. As a result, the clitics are obligatory, which suggests that the cl-
itics are presumably better analyzed as new sufxal agreement markers. Optionally,
full pronouns can be added, giving rise to instances of apparent doubling.
19
Similar
to the other Walser dialects, the change seems to affect highly frequent, short (often
monosyllabic) verbs (sometimes called Kurzverben short verbs) rst:
20
Pomattertitsch
(41) a. dets
this
bin-i
am-1sg
ich.
I
This is me. (Dal Negro 2004: 165)
b. d
the
mter
mother
set-sch ...
says-3sg
Her mother says ... (Dal Negro 2004: 171)
Dal Negro (2004) considers declaratives with V1 order in which the clitic attaches to
the right of the clause-initial verb as further instances where the clitic has turned into
a verbal agreement marker (presumably in analogy to V1-sequences in Italian), cf. the
following examples taken from Dal Negro (2004: 170f.):
(42) a. bin=i
am=I
deheima
home
gsi.
been
I have been home.
b. chan=der
can=he
no
yet
nit
not
heirate.
marry
cannot get married yet.
c. si=wer
are=we
so.
so
We are like this.
d. tn=tsch
do=they
in
in
titsch
German
zell.
speak
They speak German.
However, it is not entirely clear to me whether these examples actually can be taken
as evidence that the clitic has turned into an agreement marker. It is equally possi-
ble that V1-orders of this kind represent an inherited feature, which is reminiscent of
the frequent occurrence of V1-declaratives in older stages of Germanic. As Dal Negro
(2004: 39) herself notes, the German speaking enclaves in Northern Italy have pre-
served a number of archaic traits that in some cases date back to Old High German.
So it is quite likely that these orders in fact do not constitute an innovation, in contrast
to Dal Negros claims.
21
According to Dal Negro, the grammaticalization process in question is intimately
linked to the pro-drop property. Thus, even in contexts that do not show the new
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.21 (1340-1404)
agreement endings, overt pronouns can be omitted (probably the result of Italian in-
uence), which abets a reinterpretation of enclitics as verbal agreement sufxes (with
the relevant -role assigned to pro):
22
(43) a. tsch
do-2sg
leiga
lie
wi
like
a
tiful.
devil
You are just lying.
b. h
have-1sg
tischut
changed
ds
the
wrch.
work
I have a new job. (Dal Negro 2004: 167)
To sum up, it appears that a set of Walser dialects is affected by a change very simi-
lar to that observed in Bavarian, in which enclitic pronouns are reanalyzed as verbal
agreement sufxes in inversion contexts. Note, however, that the grammaticalization
process that affects the Walser dialects apparently differs in two important respects
from the changes in Bavarian. First, it seems that in contrast to Bavarian, the catego-
rial reanalysis of clitics does not give rise to new pro-drop properties in the Walser
dialects. Instead, it appears that the dialects in question already exhibited extensive
pro-drop features prior to the change in question. This supports the idea put forward
in Chapter 4 that pro-drop may either be a result of the rise of agreement morphology
(as in Bavarian) or constitute a factor which promotes the reanalysis of clitic pronouns,
as in the Walser dialects. Second, it seems that in Pomattertitsch, the development of
Agr-on-Chas not led to the rise of complementizer agreement, in contrast to Bavarian.
To account for this difference, I assume that the newly created agreement morpheme
on C is still restricted to the contexts where the reanalysis of clitic pronouns has taken
place (i.e., inversion environments where the nite verb occupies C) and has not been
generalized to all instances of C.
23
.. Section summary
In this section, I have argued that there is a diachronic link between complemen-
tizer agreement, pro-drop and the development of new verbal agreement markers in
Bavarian. More precisely, it has been shown that the development of new verbal agree-
ment markers from clitic pronouns could only take place in inversion contexts. This
change forced the learner to assume that the subject position is occupied by referen-
tial pro, in accordance with the requirement that the reanalysis preserves the verbs
argument structure by assigning the former pronouns -role to another element. In
addition, we have seen that the change in question is also consistent with the other
syntactic conditions on the grammaticalization of new agreement markers proposed
in Chapter 4, leading to a simplied structure in which the content of the newly cre-
ated agreement morpheme is identied under structural adjacency with Agr-on-T.
Note that this diachronic analysis is based on the synchronic account of complemen-
tizer agreement developed in Chapter 3, making use of the post-syntactic insertion of
a dissociated Agr-morpheme. On these assumptions, the diachronic development in
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.22 (1404-1459)
Bavarian has been shown to proceed via a stage where the learner assumes the existence
of a dissociated Agr-morpheme which is initially conned to C (Agr-on-C). In a sub-
sequent change, the phonological exponent of Agr-on-C is reanalyzed as the exponent
of the canonical subject agreement morpheme, Agr-on-T, and spreads to other ver-
bal positions. Furthermore, it has been demonstrated that similar grammaticalization
processes can also be observed in a set of Germanic dialects spoken in Northern Italy.
Both Cimbrian and Walser varieties have been shown to exhibit an early stage of the
change in question, where enclitics are reanalyzed as instances of Agr-on-C in inver-
sion contexts. In contrast to Bavarian, however, the newly created agreement markers
are still conned to the environments where the reanalysis originally took place.
Importantly, the developments observed in this section contradict the claim that
grammaticalization always proceeds in an upwards fashion, leading to a new exponent
of a higher functional head, as proposed by Roberts and Roussou (2003) (see Chapter
2 above). Rather, in Bavarian the exponent of a higher functional morpheme (Agr-
on-C) is reanalyzed as the realization of a lower functional morpheme (Agr-on-T),
which seems to be a typical grammaticalization path for the rise of agreement in V2
languages. Schematically, this can be depicted as in (44), where /-x/, the exponent of X
(= Agr-on-C) is reanalyzed as the exponent of Y (= Agr-on-T).
(44) XP
YP
Y ...
X
/-y/
/-x/
XP
YP
Y ...
X
/-x/
/-x/
Present-day Bavarian still shows a residue of these diachronic processes, namely a
dissociated Agr-morpheme on C, which gives rise to the phenomenon of comple-
mentizer agreement. Thus, pro-drop and complementizer agreement are reexes of
the historical development of new verbal agreement morphology in the history of
Bavarian, which had to proceed via C in inversion contexts.
24
In addition, the above
discussion has suggested that the grammaticalization of agreement markers from clitic
pronouns generally proceeds via a stage where the clitic is reanalyzed as a dissociated
Agr-morpheme before it turns into a syntactic Agr-morpheme in a subsequent devel-
opment. This hypothesis is illustrated by the following grammaticalization path for
verbal agreement markers, which is a more ne-grained description of the relevant
historical developments than the traditional model depicted in (46).
(45) free pronoun weak pronoun clitic pronoun dissociated Agr-
morpheme syntactic Agr-morpheme
(46) free pronoun weak pronoun clitic pronoun (bound) agreement
marker
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.23 (1459-1508)
However, although we have managed to establish a diachronic connection between
complementizer agreement and pro-drop in Bavarian, the investigation of syntactic
aspects has not provided an answer to the question of why complementizer agreement
and pro-drop are restricted to 2nd person contexts in present-day Bavarian. In Chap-
ter 6, it is demonstrated that this restriction is a result of morphological conditions
that shaped the development of new verbal agreement morphology in the history of
Bavarian. In the next section, it is shown that C-oriented subject enclitics were repeat-
edly reanalyzed as agreement markers in a number of Swiss Rhaeto-Romance dialects,
which are also characterized by the V2 property, but differ from Bavarian with re-
spect to the particular syntactic environments where the reanalysis of subject clitics is
triggered.
. Rhaeto-Romance
The Rhaeto-Romance languages are particularly interesting with respect to the present
research context, since it can be shown that some of them repeatedly underwent a
change in which a former enclitic is reanalyzed as an agreement ending. This section
focuses on a set of Swiss Rhaeto-Romance dialects spoken in Graubnden, drawing on
data from Gartner (1883, 1910), Widmer (1959), Haiman (1971, 1974), Linder (1987),
Oetzel (1992), Haiman and Beninc (1992), Ebneter (1994), and Liver (1999). Among
the Swiss varieties of RR (sometimes referred to as Romansh) ve major dialects are
usually distinguished (each with a number of local varieties): Puter (spoken in the
higher Engadin), Vallader (in the lower Engadin and Mnstertal valleys), Surmeiran
(in Oberhalbstein, Surmeir valley, and in Albulatal valley), Surselvan (in the higher
Rhin or Vorderrheintal region), and Sutselvan (in the lower Rhin or Hinterrheintal,
and in Nidwald). It is demonstrated that clitics repeatedly developed into new markers
of verbal agreement in the recorded history of these dialects. Interestingly, it seems
that the individual grammaticalization processes are triggered in different syntactic
contexts. First, we can observe a development similar to Bavarian where C-oriented
clitics are directly reanalyzed as agreement markers on C, giving rise to a limited
amount of pro-drop. In addition, it seems that in the present-day dialects, a different
development is underway in which new agreement formatives are created from subject
enclitics that occur in clitic doubling structures.
.. A grammatical sketch of the Swiss Rhaeto-Romance dialects
Before turning to the diachronic processes that led to the development of a set of new
agreement markers in the Swiss RR dialects, this section briey introduces some basic
syntactic characteristics of the RR dialects under investigation, focusing on basic word
order, the syntax and morphology of pronominal forms, and pro-drop properties.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.24 (1508-1586)
... Word order: V2 and SVO
All Swiss varieties of RR are characterized by the V2 property (cf. e.g. Haiman 1974;
Linder 1987; Haiman & Beninc 1992; Oetzel 1992). Usually the subject follows the
nite verb in main clauses if the clause-initial position is occupied by another XP (an
adverbial, a PP, an object, a wh-phrase etc.). This is illustrated by the following ex-
amples which exemplify inversion in main declaratives and wh-questions for the ve
major Swiss RR dialects:
Puter
(47) a. Las
the
chavras
goats
charget
put
el
he
aint
into
illa
the
chaista...
box...
He put the goats into the box... (Linder 1987: 17)
b. A
to
chi
whom
ho
has
lhomin
the
sulvedi
dwarf
preschanto
introduced
na
a
giuvna?
young-girl
To whom did the dwarf introduce a young girl? (Oetzel 1992: 7)
Vallader
(48) a. A
to
quella
her
sto
must
el
he
imprometter
promise
da
to
nun
never
ir
go
m
prt
pl
again
davent.
away
He must promise to her that he will never go away again.
(Oetzel 1992: 28)
b. Da
of
che
what
discuorra
talks
el
he
tuotta
all
pezza?
time
What does he talk about all the time? (Ebneter 1994: 798)
Surmeiran
(49) a. A
to
quests
these
pleds
words
suonda
followed
ena
a
parfetga
complete
quietezza.
silence
A complete silence followed these words. (Oetzel 1992: 29)
b. Tge
what
fo
makes
pia
then
ti
your
onda?
aunt
What is your aunt doing? (Gartner 1910: 17)
Surselvan
(50) a. La
the
brev
letter
ha
has
la
the
mumma
mother
scret.
written
The mother wrote the letter. (Liver 1999: 146)
b. Cura
when
eis
is
ella
she
morta?
died
When has she died? (Liver 1999: 145)
Sutselvan
(51) a. A
and
que
this
num
name
porta
bears
igl
the
cratsch
baby-of-the-family
nc
still
oz.
today
And the baby of the family still bears this name today. (Oetzel 1992: 29)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.25 (1586-1674)
b. Cura
when
vala
goes-3sg.fem
la
the
Nona?
grandma
When will grandma leave? (Linder 1987: 155)
However, note that the Swiss RR dialects are actually less strict with respect to the
V2 constraint than the neighboring Germanic V2 languages.
25
Oetzel (1992: 17ff.) ob-
serves a number of examples with non-inverted order in written records of Puter,
Vallader and, in particular, Surmeiran, cf. the following examples from Vallader and
Surmeiran:
26
Vallader
(52) In
in
quel istess
this
mumaint
moment
il
the
trid
ugly
uorsin
bear
as
re
transmdet
changed
in
into
n
a
bel
beautiful
prinz.
prince
In this moment, the ugly bear changed into a handsome prince.
Surmeiran
(53) La
the
seira
evening
anturn
against
las
the
nov
nine
Tina
Tina
sa
re
prepara
prepares
per
for
sorteir.
leaving
In the evening around nine, Tina gets ready for going out.
For the time being, I assume that V2 orders result from the nite verb moving to C,
with the clause-initial XP occupying the specier of CP (see Section 5.3.4.1 for some
modications). Note that the V2 property appears to be restricted to root clauses in
the Swiss RR dialects, in contrast to Icelandic or Yiddish, for example. In embedded
clauses, the nite verb apparently moves out of the VP into a position in the inec-
tional domain. This is suggested by following examples where the nite verb precedes
adverbials or the sentential negation:
27
Puter
(54) ch=i
when=it
gnit
came
dandettamaing
suddenly
n
a
ferm
great
temporal...
storm
When suddenly, a great storm came (Linder 1987: 136)
Vallader
(55) cha
that
il
the
tschl
sky
vuless
would
nalmantig
nally
as
re
sclerir
clear
That the sky would nally clear up (Linder 1987: 35)
Surmeiran
(56) ci
that
t
you
lesti
like-subjnc
ssa
now
n
a
marus
lover
That you would like to have a lover now (Linder 1987: 59)
Sutselvan
(57) perquei ca
since
la
the
furtga
gallows
era
was
betga
not
daliensch
far-away
Since the gallows was not far away (Linder 1987: 162)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.26 (1674-1778)
Table 6. Verbal agreement (present indicative) in ve Swiss RR dialects.
Puter Vallader Surmeiran Surselvan Sutselvan
1sg - - - -6l -
2sg -ast -ast -as -as -(a)s
3sg -a -a -a, -e -a -a
1pl -ains -ain(a) -(g)n -in, -ein -(g)n
2pl -ais -aivat -es, -as -is, -eis -(e)s, -(a)s
3pl -an -an -an -an -an
Moreover, under the assumption that there is a correlation between rich verbal in-
ection and overt verb movement (Kosmeijer 1986; Holmberg & Platzack 1988, 1991,
1995; Roberts 1993a; Bobaljik 1995; Rohrbacher 1999; Vikner 1994, 1995, 1997), the
Swiss RR dialects are actually expected to exhibit verb movement into the inectional
domain. Table 6 illustrates the agreement endings (present indicative) found in the RR
dialects under investigation.
28
If we take a look at the paradigms listed in Table 6, it becomes clear that the
verbal agreement morphology of the Swiss RR dialects qualies as rich under all
common criteria for richness. Apart from Surmeiran and Sutselvan, where 2sg and
2pl are (at least partially) homophonous, the dialects show a fully distinctive agree-
ment paradigm, which is expected to correlate with overt verb movement. For the sake
of concreteness, I assume that the functional head targeted by verb movement is T in
embedded clauses. Finally, the fact that the subject usually precedes the nite verb in
embedded clause suggests that the subject occupies SpecTP in the overt syntax (which
is standardly attributed to the presence of an EPP feature in T).
In contrast to the surrounding German dialects, the Swiss RR varieties exhibit
basic VO word order (which is a characteristic of RR in general). Thus, the non-nite
verb is followed by its complements in the unmarked word order. This is illustrated for
DP and PP complements with the following examples from Puter and Vallader taken
from Oetzel (1992: 7ff.):
29
Puter
(58) Lhomin
the
sulvedi
dwarf
ho
has
preschanto
introduced
[na
a
giuvna]
young-girl
[al
to-the
raig].
king
The dwarf has introduced a young girl to the king.
Vallader
(59) Ed
and
eu
I
nha
have
vis
seen
[davant
before
mai]
me
[ad
at
n
a
pitschen
little
homet].
man
And I saw a little man in front of me.
Note that these examples exemplify merely the basic word order of these dialects. In
addition, there is a variety of further ordering possibilities which can only be hinted at
in the present discussion (for an overview and some discussion cf. Linder 1987 and in
particular Oetzel 1992). As an illustration, consider the following embedded clauses.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.27 (1778-1858)
In (a) a PP has been scrambled to the left of the subject, while in (b), a quantied
object occurs in preverbal position:
30
(60) a. chi
that
[cun
with
quel]
this
els
they
nun
not
stetan
stay
pl...
any-longer
That they will not stay with this one any longer (Puter; Linder 1987: 31)
b. M
but
che
what
ans
us
gdigl
help-it
cur
if
nus
we
[quegl
that
tutt]
all
cretein?
believe
But how does it help us if we believe all that? (Sutselvan; Linder
1987: 182)
The next section focuses on the inventory and distribution of personal pronouns, a
topic which is particularly relevant for the diachronic analyses developed later (re-
call that pronouns commonly provide the lexical source for the grammaticalization of
agreement markers).
... The distribution of pronominal forms and pro-drop
The Swiss RR dialects exhibit three series of pronouns, one set of full pronouns that
may bear stress and two sets of atonic, reduced pronouns: (i) proclitics that attach
to the left of the nite verb in V2 clauses, and (ii) a set of enclitic forms which re-
semble the C-oriented clitics of Germanic (cf. Linder 1987). Similar to Bavarian, all
Swiss RR dialects are characterized by the lack of enclitic 2nd person pronouns (cf.
Haiman 1971; Linder 1987). In Puter and Surmeiran, we can observe a proclitic (Puter
a-, Surmeiran i-) which can be used for several person/number combinations (1sg,
3sg.neut, and all plural forms).
31
Table 5 illustrates the pronominal system found in
Puter:
32
Table 7. Pronominal forms (nominative) in Puter
(Haiman 1971; Linder 1987; Liver 1999).
Full pronoun Proclitic Enclitic
1sg eau a- -i, a
2sg t t(-)
3sg.masc el (e)l- -l
3sg.fem ella (el)la- -la
3sg.neut ad a- -a
1pl nus a- -a
2pl vus a-
3pl.masc els a- -e
3pl.fem ellas a- -e
Recall that in Chapter 4 above, it has been claimed that the presence of clitic pronouns
is a prerequisite for the grammaticalization of new agreement markers. In addition, it
appears that the Swiss RR dialects exhibit (limited) pro-drop, another feature which
by assumption plays an important role in the development of new verbal agreement
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.28 (1858-1930)
morphology, since it provides a means to preserve the verbs argument structure if
a pronominal argument is reanalyzed as an agreement marker (see Chapter 4 for
discussion). Linder (1987) lists a number of examples where a subject pronoun is
omitted. However, he notes that there is a signicant difference between proclitic and
enclitic pronouns with respect to pro-drop: the omission of proclitics (which appear
in clause-initial position) is quite rare and is apparently the result of Italian inuence
or an instance of topic-drop similar to German. Linder (p. 21) further notes that apart
from cases of topic-drop, the omission of proclitics is a phenomenon which is more
or less conned to the written language of the 19th century and no longer accepted
by present-day speakers. The following examples from Puter and Vallader illustrate
omission of proclitics, leading to V1 order:
(61) Puter
Giavschessans
wish-subjnc-1pl
da
to
preschenter
present
a
to
noss
our
lectuors
readers
ils
the
tips
typical
cracteristics
characters
da
of
nossistoria
our-story
engiadinaisa.
Engadinian
(We) would like to present our readers the typical characters of our Enga-
dinian story. (Chalender Ladin 1911; Linder 1987: 24)
(62) Vallader
Il
the
gigant
giant
es
is
gni
come
our
out
e
and
n
have-3pl
fat
made
la
the
partida...
game
The giant came out and (they) made a game... (Linder 1987: 31)
In contrast, enclitic pronouns can be more freely omitted (in all person/number com-
binations apart from 3sg, where no examples with pro-drop are attested, cf. Linder
1987: 34), even in the present-day language.
33
2nd person pronouns play a special role
in that they are normally dropped in inversion contexts in all Swiss RR dialects (cf. e.g.
Linder 1987: 53). This is illustrated by the following examples from Vallader and Puter
for 2sg, 2pl, respectively (recall that no enclitics are available for 2nd person forms).
(63) Vallader
Hoz
today
est
are
vaira
really
nervus,
nervous
Paul.
Paul
Today, (you) are really nervous, Paul. (Linder 1987: 35)
(64) Puter
Cu
how
fais
make-2pl
que?
that
How do (you) make that? (Linder 1987: 35)
So far, I have illustrated the essential word order properties of the Swiss RR dialects,
which are characterized by a V2 constraint that operates on a basic SVO syntax. In
addition, I have shown that the Swiss RR dialects exhibit an elaborate series of clitic
pronouns and thus satisfy a necessary condition for the development of new verbal
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.29 (1930-1991)
agreement markers from former pronouns. In addition, all dialects apparently exhibit
a limited amount of referential pro-drop, which still differs signicantly from the kind
of general pro-drop found in Italian, for example. Recall, however, that it has been
claimed above that the availability of pro-drop (even if limited) may facilitate the
grammaticalization of agreement markers by providing an alternative means to pre-
serve the predicates argument structure if a pronoun is reanalyzed as an agreement
morpheme. In the following, it is shown how the particular syntactic properties of the
Swiss RR dialects presented so far shaped the development of new verbal agreement
markers in the (recent) history of these languages.
.. Earlier grammaticalization processes affecting enclitic pronouns
As already noted above, we can observe a set of distinct grammaticalization processes
in the history of the Swiss RR dialects under investigation. This section focuses on
developments that took place in previous historical stages. It can be shown that the
agreement morphology found in the present-day Swiss RR dialects has been at least
partially shaped by the reanalysis of former enclitic subject pronouns as enlargements
of the existing verbal agreement endings, similar to the facts observed in Bavarian. In
the following, this is illustrated for 1st and 2nd person forms.
... 1st person forms
The Swiss RR dialects are characterized by the loss of the original Romance 1sg agree-
ment sufx -o, similar to many Northern Italian dialects. In some Italian Rhaeto-
Romance dialects, this loss was compensated by the development of new markers for
1sg: in present-day Ampezzan we again nd -o, presumably under inuence from the
nearby Italian dialects (e.g., Venetian). In some varieties of Friulian, a new form for
1sg (-e -i) can be observed.
34
A rather recent development took place in Surselvan,
which exhibits 1sg -6l, while Old Surselvan showed consistently - in this context.
However, from 1700 on, -6l appeared as a new ending for 1sg, which is still used in
the present-day language. The origin of the new ending is still unclear (cf. Haiman &
Beninc 1992: 93f. for some discussion).
In contrast, the origin of the present-day 1pl forms is more transparent. In Puter,
a nal /-s/ has been added to the older ending 1pl -(a)in.
35
It is commonly assumed
that the new agreement formative 1pl -ains developed via a reanalysis of an enclitic
1pl (no)s in inversion contexts (Gartner 1883; Widmer 1959; Linder 1987; Haiman &
Beninc 1992). Thus, similar to the Bavarian examples discussed above, the V2 prop-
erty seemingly provided an appropriate syntactic context for the development of new
agreement endings from former clitic pronouns. In one of the earliest Puter records,
a translation of the New Testament by Giachem Bifrun (dating from 1560), the new
enlarged form is still conned to inversion contexts, while the older ending -(a)in is
consistently found when the 1pl pronoun occupies a preverbal position (cf. Linder
1987: 70ff.).
36
This is illustrated by the following example:
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.30 (1991-2055)
(65) Mu
but
hauiand
having
la
the
spaisa,
food
&
and
a quellas
some
chioses
things
che
that
nus
we
nscuurin
us=put.on-1pl
cun
with
quaistes
these
chioses
things
dains
shall-1pl
esser
be
cuntains.
content
But having food and clothing, with these things we shall be content. (Bifrun,
Timothy I. 6,8; Linder 1987: 70)
In (65), the old ending -(a)in is still found on the verb cuurin which occurs in non-
inverted position inside a relative clause, while the enlarged form dains is used in a V2
clause with inversion, in which the clause-initial position is occupied by the PP cun
quaistes chioses. Note that this distribution is identical to the facts observed in connec-
tion with the grammaticalization of new agreement markers in the history Bavarian:
the new forms emerge rst on verbs that have moved to C and spread to other verbal
positions in a subsequent development. In the late 16th/early 17th century, nal -s is
already much more common. Linder (1987: 71) gives the following example:
(66) Il
the
Segner
Lord
vulains
want-1pl
luder,
praise
cha
that
nus
we
poassens
can-1pl
a
to
chiesa
home
turner.
turn
We want to praise the Lord, so that we can return to home.
Interestingly, in some dialects the 1pl subject enclitic a has undergone a similar de-
velopment, eventually giving rise to a second cycle of cliticization and reanalysis in
inversion contexts (Linder 1987). In spoken Surselvan, Sutselvan, and Surmeiran, the
clitic shows up obligatorily on oxytonic verbs (i.e., verbs that are stressed on the last
syllable) in inversion contexts, cf. the following examples from Surmeiran:
(67) a. Davent da l
from there
muntag-ns-a
climb-1pl-1pl
sen
onto
igls
the
Corns
peaks
greischs...
gray
From there, we climb the Gray Peaks.
b. Ea,
yes
alloura
then
lai-ns-a
want-1pl-1pl
parteir.
leave
Yes, then we want to leave. (Linder 1987: 76)
However, -a cannot be added to verbs that receive stress on the penultimate syllable
(paroxytonic verbs). Instead, the form with nal -s is retained:
(68) Finalmantg
nally
ischans
are-1pl
anc
still
ballos
driven
cugl
with-the
auto
car
segl
on-the
Klausen...
Klausen
Finally, we drove with the car onto the Klausen... (Linder 1987: 77)
This distribution is apparently governed by a stress rule which is characteristic of the
Swiss RR dialects. This rule, called penultimate stress target in Haiman (1971) and
three-syllable rule in Haiman and Beninc (1992), forbids antepenultimate stress on
verbs. If the verb receives stress on the nal syllable (as e.g. muntagns climb-1pl), nal
-a can be added to the existing verbal inection, since the resulting form complies with
the penultimate stress target, receiving stress on the penultimate syllable. In contrast,
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.31 (2055-2115)
with verbs such as ischans are-1pl, which already receive stress on the penultimate
syllable, the integration of -a into the verbal inection is impossible, since this would
lead to a verb with a stressed antepenultimate (therefore *ischans+a).
Linder (1987: 76) shows that in some dialects such as Pontresina (a variety of
Puter), the new ending -insa is used independent of word order: nal -a is completely
integrated into the verbal inection and no longer conned to inversion contexts. That
is, according to Linder the forms in (69) can also show up in embedded clauses:
(69) a. nus
we
curr-insa
run-1pl
b. nus
we
gia-insa
go-1pl
Thus, -(a)insa has been generalized as a new marker of 1pl verbal agreement. Again,
these facts clearly show that the new agreement formatives arose rst in inversion
contexts and spread to other verb positions in a subsequent development. However,
similar to the distribution of 1pl -a in Surmeiran (see above), the former clitic has only
been added to verbs that do not bear stress on the penultimate syllable. Elsewhere, the
old ending (-(a)ins) has been maintained:
(70) a. ad
we
es-ans(-*a)
are
b. a
we
durm-ivans(-*a)
were-sleeping
To summarize, in the Swiss RR dialects under investigation, the possibility of reinter-
preting a subject enclitic as a verbal agreement sufx is dependent on two factors: rst,
the reanalysis is only possible in inversion contexts. Hence, the V2 property plays an
important role in the change in question, similar to the Bavarian data discussed above.
In addition, the reanalysis depends on the syllable structure (and stress pattern) of
a given verb and can only be carried out if the resulting agreement ending does not
lead to a stress pattern where stress falls on the antepenultimate syllable. In the next
section, it is shown that quite similar observations can be made with respect to 2nd
person forms.
... 2nd person forms
Similar to many German varieties, Puter, Vallader and Surmeiran have replaced the
original verbal agreement ending 2sg /-s/ by the enlarged form /-st/.
37
This change is
commonly attributed to the reanalysis of the 2sg clitic -t(i) in inversion contexts (cf.
Widmer 1959; Linder 1987; Haiman & Beninc 1992). Widmer (1959: 64f.) shows that
the development of the new ending /-st/ proceeded via the following three historical
stages. In stage 1, the weak pronoun follows the nite verb, which still carries the old
ending /-s/ (inversion contexts):
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.32 (2115-2175)
(71) Or
from
tutta
all
quest
this
na
not
ps
can
ti
you
provar
prove
il
the
teu
your
fagg.
case
From all this, you cannot prove your case. (Nauli, 46; Widmer 1959: 64)
In Stage 2, the progress of the relevant grammaticalization process is reected by the
fact that the inverted verb and the enclitic t are written as one word:
(72) Mo
but
qui
here
sarest
will.be-you
leid...
sad
But here, you will be sad... (Wietzel, Prf., 32; Widmer 1959: 64)
Finally, in stage 3, the enclitic has been reanalyzed as part of the verbal agreement
morphology and can be accompanied by a full pronoun. In addition, the new end-
ing can appear on nite verbs in embedded clauses. This indicates that the clitic has
developed into a marker of verbal agreement, cf. the following dialogue taken from
Widmer (1959):
(73) A: Che
What
tscherchi-ast
search-2sg
t,
you,
juwen,
young man
in
in
quaista
this
strda?
street
What are you looking for?
B: Meis
my
frars
brothers
veng
come
eau
I
to
tscharchir,
search
s
if
t
you
msav-est
me=know-2sg
intraguidr.
to-guide
I come to search my brothers, if you know how to guide me (to them).
(Travers, 4; Widmer 1959: 65)
In one of the oldest Swiss RR text, the bible translation by Bifrun (dating from 1560),
nal /-t/ is still very frequently absent in subject-initial clauses, but consistently present
in inverted word order. Moreover, even at this very early stage there are a few examples
where the enlargement /-t/ co-occurs with an additional 2sg pronoun t and thus can-
not be analyzed as an enclitic (Linder 1987: 54, fn. 42). This resembles the present-day
usage (which according to Haiman & Beninc 1992: 95 was already established around
1700), where the agreement ending /-st/ has been generalized to all contexts (examples
taken from Haiman & Beninc 1992: 94f.):
(74) a. t
you
vaes
go
(John 14: 5)
b. t
you
nu
not
pous
can
(John 13: 36)
(75) innua
where
vaest
goest
t?
thou
(John 13: 36)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.33 (2175-2252)
Again, these observations emphasize the important role of the V2 property for the
reanalysis of C-oriented clitics, suggesting that in the Swiss RR dialects, the historical
development of 2sg /-st/ proceeded along similar lines as proposed for Bavarian (cf.
(11) above).
Interestingly, it can be observed that the RR dialects that developed 2sg /-st/ share
another characteristic: as noted above, there are usually no overt 2sg subject enclitics
present (full forms may be used when they receive stress). According to Linder (1987),
the present-day Swiss RR dialects in fact lack 2nd person enclitics completely. The dis-
cussion of Bavarian in Section 5.2 above suggests a diachronic explanation of this fact
(cf. Linder 1987: 53ff. for a related idea). More precisely, it seems reasonable to sup-
pose that similar to present-day Bavarian, the historical development of the agreement
formative /-st/ led to partial pro-drop for 2sg contexts in RR, due to the conversion
of former enclitics into verbal agreement sufxes.
38
According to Meyer-Lbke (1894) and Linder (1987: 58), the RR dialects spoken
in Graubnden show a related development for the agreement ending 2pl present in-
dicative (e.g., chant-ais sing-2pl), suggesting that nal /-s/ in the present-day sufx
-(a)is developed via cliticization of the 2pl pronoun (vo)s in inversion contexts.
39
Sim-
ilar to 2sg, I assume that this development is the historical source of pro-drop in 2pl
contexts.
40
We can therefore conclude that the Swiss RR dialects exhibit similar historical de-
velopments as Bavarian. Again, the V2 property provided the syntactic context for
the reanalysis of C-oriented enclitics as Agr-morphemes. Without giving further argu-
ments, I assume that these developments proceeded more or less completely analogous
to the changes that affected Bavarian (see (26) above). In other words, C-oriented en-
clitics were reanalyzed as dissociated Agr-morphemes the content of which is licensed
under structural adjacency with a valued Agr-morpheme on T (which by assumption
also hosts the nite verb in the overt syntax in RR, see above). This reanalysis depends
on the V2 property, which brings about adjacency between the clitic and the verb and
thereby licenses a reanalysis of the former as part of the verbal inection, in line with
the Word building constraint. The new agreement formatives were initially conned
to C and spread later to other verbal positions as a result of a change in which the
phonological exponent of Agr-on-C replaced the existing exponent of Agr-on-T. As a
by-product of this change, the Swiss RR dialects developed limited pro-drop proper-
ties due to the fact that the reanalysis of enclitic pronouns forced the learner to posit
the existence of a referential pro in the contexts where the reanalysis took place. On the
assumption that the empty pronoun is accessed by an Agree operation and thus does
not move to SpecTP, this change is also consistent with the requirement of structural
simplication:
41
(76) [
CP
XP [
C
C+V+pronoun
i
[
TP
t
i
[
T
T [
P
t
i
...
[
CP
XP [
C
C+V+Agr [
TP
T [
P
pro ...
In the next section, I discuss a more recent development in which the grammaticaliza-
tion of new agreement formatives proceeds from clitic doubling structures.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.34 (2252-2304)
.. Clitic doubling and the rise of agreement
In a number of Rhaeto-Romance (RR) dialects, we can observe subject clitic dou-
bling, the properties of which can be taken to indicate that the clitic elements in these
contexts undergo a grammaticalization process leading to new (verbal) agreement
markers.
42
The change in question exclusively affects C-oriented subject enclitics in
inverted position, similar to the data discussed so far. Again, it seems that the develop-
ment of new verbal agreement morphology proceeds via a transitional stage where the
new agreement morphemes are conned to the C-head, suggesting that the V2 prop-
erty plays an important role in the reanalysis of second position clitics (cf. Haiman
1991 for a similar claim). Crucially, however, the present development differs from the
phenomena discussed above in that it proceeds from a clitic doubling conguration
where an additional DP is present. This DP turns into the overt subject of the clause
when the clitic is reanalyzed as a marker of verbal agreement.
While the Swiss varieties of RR show a relatively uniform behavior with respect to
the basic syntactic properties illustrated above, the individual dialects exhibit some
important differences concerning the availability of subject clitic doubling. In the
remainder of this section, I will focus on these differences, arguing that they repre-
sent different stages of a new wave of enclitic pronouns developing into agreement
formatives.
In all RR dialects apart from Surselvan (which lacks subject enclitics, Linder
1987: 146), enclitic subject pronouns can be reinforced by full DPs or pronouns, which
then usually receive stress (cf. Linder 1987: 146), giving rise to instances of subject clitic
doubling. In principle, doubling is possible in all person/number combinations where
a subject enclitic is available, that is, 1sg, 1pl and 3sg, 3pl. Examples with 2nd per-
son subjects are not attested, since the dialects in question appear to lack 2nd person
subject enclitics (possibly due to the fact that they have been reanalyzed as agreement
markers in an earlier development, see above). Moreover, with 3rd person subjects,
clitic doubling occurs much more frequently with full DP subjects than with full tonic
pronouns.
43
It appears that doubling is conned to inversion contexts where both the
C-oriented enclitic and the full subject follow the nite verb occupying the C-head
(see Section 5.3.4.1 for some discussion). The additional full nominal is often adjacent
to the clitic (in particular with 1st person forms), but may also appear in a position
further to the right of the clause (at least in Puter, Vallader and Surmeiran). The con-
struction is optional in Puter and Vallader and occurs quite frequently in Surmeiran.
The following data illustrate (optional) subject clitic doubling in Puter, Vallader and
Surmeiran (examples taken from Linder 1987: 147ff.; the clitic and the corresponding
full form are marked by boldface):
44
Puter
(77) a. Dals
them
spisantr
to-feed
he=ia
have=clit.1sg
eau
I
grand
great
dallett.
pleasure
I have great pleasure in entertaining them to a meal.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.35 (2304-2370)
b. Et
and
per che
why
giain=s
go=clit.1pl
r
also
nus
we
in
in
mnchia
jeopardy
hura
hour
priuel?
every
And why do we also stand (lit. go) in jeopardy every hour?
c. Co
then
haun=e
have=clit.3pl
ells
they
crida
cried
tuotts
all
duos.
both
Then, they both cried.
Vallader
45
(78) a. Alura
then
=l
has-clit.3sg.masc
dit
said
il
the
gl: ...
son
Then, the son said: ...
b. Che
which
effet
effect
ha=la
has-clit.3sg.fem
gn
done
aint
in
il
the
pvel
people
la nouva predgia?
the new sermon
Which effect had the new sermon on the people?
c. Cur
when
chals
that=the
ns
ones
rivainan,
arrived
surdaivn=a
handed-over=clit.3pl
ils
the
oters
others
ad
to
els
them
la
the
scossa
herd
e
and
partivan.
left
When they arrived, the others handed over the herd and left.
Surmeiran
(79) a. Schi
if
te
you
ist
are
cuntaint,
content
vign=a
come=clit.1sg
ia
I
avant
before
mezde...
noon
If thats ne with you, I will come before noon.
b. Lagns=a
want=clit.1pl
nous
we
dus
two
betg
go
eir
also
dumang
tomorrow
ainten
in
enotra
another
vischnanca
village
a
to
messa?
mass
Should we go to mass in another village tomorrow?
c. Chelidea
this idea
veve=l
had=clit.3sg.masc
gia
had
igl
the
uestg
bishop
Ziegler
Ziegler
sez.
himself
Bishop Ziegler himself had had this idea.
d. Par
for
tema
fear
e=lla
is=clit.3sg.fem
la
the
femna
woman
curoida
run
ancheunter
to
Zorten.
Zorten
Because she was frightened, the woman ran to Zorten.
e. Dantant
meanwhile
n=igl
are=clit.3pl
rivos
arrived
igls
the
bernes
Bernese
e
and
turitges.
Zurich
Meanwhile, the Bernese and Zurich people have arrived.
As briey mentioned above, these instances of subject clitic doubling are restricted to
inversion contexts where both the C-oriented enclitic and the full formfollowthe nite
verb, which occupies C. In contrast, proclitics apparently do not allow doubling, that
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.36 (2370-2441)
is, a conguration such as *proclitic-nite verb-full DP is not attested in the data sam-
ple available to me. In addition, Haiman and Beninc (1992: 192) explicitly state that
doubling is possible only in inverted order. Furthermore, full pronouns/DPs and weak
forms/proclitics are in complementary distribution in clause-initial position. These
restrictions can be summarized as follows:
(80) Ordering restrictions on clitic doubling in the Swiss RR dialects
a. *full DP-nite verb-enclitic...
b. *proclitic-nite verb-full DP...
c. *full DP-proclitic-nite verb...
Furthermore, in Puter, Vallader, and Surmeiran, doubling exhibits a deniteness effect:
the full DP must be denite (or specic), that is, sentences such as yesterday arrived-
he a man are not possible (cf. the examples above). Together with the observation
that the full nominal often receives stress, this can be taken to indicate that in Puter,
Vallader, and Surmeiran, clitic doubling fullls certain stylistic (or, rather, discourse)
functions in which a full element is added to reinforce an enclitic pronoun for reasons
of emphasis.
Sutselvan shows a form of clitic doubling the properties of which differ from the
kind of (optional) doubling found in the other Swiss RR dialects. In Sutselvan, clitic
doubling appears to be a much more common, almost obligatory phenomenon that
has lost its function as a stylistically marked structural option. For example, in contrast
to the other dialects, the full nominal is not added for reasons of emphasis and does
not receive stress (cf. Linder 1987: 150). Relevant example sentences from Sutselvan
are shown in the following.
46
(81) 1st person singular
Egn
one
da
of
quels
those
lev-i
wanted-1sg
ear
also
jou.
I
I also wanted one of those. (Sutselvan; Linder 1987: 148)
(82) 1st person plural
Ascheia
so
vain-sa
have-1pl
nus
we
arviart
unlocked
igl
the
mulegn...
mill
So we have unlocked the mill... (Sutselvan, Linder 1987: 149)
(83) 3rd person singular masculine
Igl
it
fetschi
is
preaschas,
urgent
-l
has-3sg.masc
el
he
getg.
said
He said its urgent. (Sutselvan; Linder 1987: 153)
(84) 3rd person singular feminine
Cunquegl
since
cigl
it
eara
was
november,
November
vev-la
had-3sg.fem
la
the
scola
school
antschiat.
begun
Since it was November, the school had begun. (Sutselvan; Linder 1987: 155)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.37 (2441-2494)
(85) 3rd person plural
Natiral
of
vev-in
course
las
had-3pl
matns
the
radetg
girls
sei
brought up
mailenders.
Milans
Of course, the girls had brought up some Milans [pastries]. (Sutselvan;
Linder 1987: 161)
Note that the old agreement ending is replaced by the clitic in the examples (84) and
(85). In other words, subject verb agreement is solely marked by the clitic in these
sentences. This already suggests that the former enclitics have developed into some
form of agreement marking in Sutselvan. A further important difference between Sut-
selvan and the other Swiss RR dialects concerns the deniteness restriction, which
requires the full nominal to be denite or specic in Puter, Vallader and Surmeiran (see
above). In contrast, no such restriction can be observed in Sutselvan, where doubling
has spread to non-denite nominals as well:
(86) a. Qua
there
han-i
have-3pl
schon
already
blears
many
via
seen
da
all
tutta
kind
sorts...
of things
Many people have already seen all kind of things there...
b. Mo
but
igl
the
lungatg
language
da
of
la
the
dunnetta
little-woman
san-i
knows-3pl
nigns.
no-one-pl
But nobody knows the language of the little woman. (Sutselvan; Linder
1987: 162)
Under the assumption that the deniteness effect is a restriction imposed on clitic dou-
bling by its discourse function (similar to the well-known deniteness/specicity re-
striction on topics, cf. Givn 1976; Grewendorf 2002), these facts again suggest that the
doubling construction has lost a great deal of its stylistic force in Sutselvan. Further-
more, recall that the loss of deniteness/specicity effects is often understood as an in-
dication that the clitic has evolved into an agreement marking element (cf. Uriagereka
1995). If taken together, the almost obligatory status of clitic doubling, the fact that it is
not restricted to stressed and denite/specic full nominals, and the observation that
the clitics can replace the original agreement morphology in V2 contexts show rather
conclusively that the clitic has developed into a new marker of verbal agreement in
Sutselvan (cf. Linder 1987: 162; Oetzel 1992: 49, 51; cf. Haiman & Beninc 1992: 197ff.
for similar conclusions concerning RR varieties spoken in Northern Italy).
.. The reanalysis of emphatic doubling structures
In this section, I present an analysis of the developments affecting Sutselvan that is
based on the idea that new forms of agreement may result from a (formerly stylistic)
strategy in which a full DP/tonic pronoun is added to reinforce a phonologically de-
fective clitic, leading to clitic doubling. In the course of time, the construction loses
its stylistic force and the originally reinforcing element is eventually reanalyzed as the
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.38 (2494-2539)
real argument, whereas the former clitic is reinterpreted as an (verbal) agreement
marker.
47
Before we can explore this scenario further, let me introduce a set of theoreti-
cal assumptions that the present analysis is based on, focusing on the structure of
clitic doubling in the Swiss RR dialects under consideration. As noted above, the Swiss
RR dialects are characterized by the V2 property. In addition, we have seen that they
are characterized by a basic VO order and that the nite verb presumably occupies
T in the overt syntax. With respect to the C-oriented enclitics found in these di-
alects, I suggest that they should be analyzed as D-heads. Following Kayne (1994: 61),
Uriagereka (1995: 81), and Torrego (1995, 2002), I assume that the phrase structural
representation of pronominal clitics is as in (87).
(87) DP
NP D
clitic pro
In other words, a pronominal clitic enters the syntactic derivation as the determiner
of a phonetically empty NP which occupies the clitics complement position (and
presumably acts as the real argument in this conguration).
In the RRdialects that exhibit optional clitic doubling, full nominals may be added
for reasons of emphasis, as a reinforcement of the enclitic pronoun which cannot bear
stress. I assume that this state of affairs should be modeled by a structure where the
clitic D-head selects for a reinforcing full nominal (henceforth called the double) in its
specier (Uriagereka 1995; Kayne 2002). The two elements are then merged together
in a big DP:
48
(88) DP
D
D NP
DP
clitic pro
double
This complex DP is base-generated in SpecP, where it receives the -role for the exter-
nal argument (which can be assumed to percolate to the elements contained in the big
DP). Subsequently, the big DP moves to SpecTP for reasons of Case and EPP checking.
From there, the clitic adjoins to C (either at MS/PF or in the overt syntax), similar to
the C-oriented clitics found in the Germanic languages (e.g., German). Accordingly,
optional clitic doubling in the present-day Swiss RR dialects (i.e., Puter, Vallader and
Surmeiran) is analyzed as resulting from the following structure:
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.39 (2539-2580)
(89)
T
NP
pro
P T
t
i
D
clitic
VP
TP V
fin
+ C
C
CP
XP
DP
i
D DP
double
It seems reasonable to assume that (89) also represents the structure of clitic doubling
in a historical stage of Sutselvan prior to the changes that affected this construction
and the status of clitic pronouns in this variety of RR. The current status of clitic dou-
bling in Sutselvan (doubling is obligatory, the double is not necessarily emphasized and
does not obey the deniteness restriction observed in the other dialects) suggests that
at some point in the (recent) history of Sutselvan, the doubling strategy lost its stylis-
tic force (presumably due to an over-use). As a consequence, the doubling structure
(which is associated with a certain stylistic effect) failed to be cued during language
acquisition, giving way to a somewhat simpler structure:
49
(90)
C XP
T Agr
TP C
DP
i
V
fin
+ C
T P
t
i
VP
CP
In (90), the former clitic has been reanalyzed as an Agr-morpheme on C. The former
double assumes the argument role previously assigned to the big DP. On the syntactic
surface, this change gives the impression of obligatory clitic doubling. In the follow-
ing, I will take a closer look at the factors that were involved in the categorial reanalysis
of clitics, paying special attention to the set of syntactic conditions on the rise of new
agreement formatives proposed above.
50
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.40 (2580-2650)
Similar to the changes affecting Bavarian and earlier stages of Swiss RR, adjacency
between the C-oriented clitic and the nite verb is obligatory in inversion contexts
(where the reanalysis took place), so no problem arises with respect to the Word build-
ing constraint. The resulting Agr-morpheme on C attaches to the nite verb as a result
of syntactic V-to-C movement, satisfying the requirement that the two elements com-
bine prior to Vocabulary Insertion. In other words, the V2 property again provides a
syntactic environment where new agreement formatives may develop.
As noted above, the recent developments in Sutselvan differ from the changes that
affected Bavarian and earlier stages of RR in that there is an overt element present in
the structure which may assume the role of the subject. As a result of the reanalysis of
the clitic, the agent/subject -role is no longer assigned to the big DP which contained
both the clitic and its double. Preservation of argument structure is warranted by assign-
ing the relevant thematic role to the former double, which turns into the real subject
of the structure that results from the reanalysis in question. Note that this change also
leads to structural simplication due to the loss of the big DP structure and the rele-
vant selectional features that required the addition of a full DP double in the specier
of the big DP.
Finally, parallel to the analysis of the Bavarian data above, I suggest that the newly
created agreement morpheme attached to C is inserted post-syntactically as a dissoci-
ated Agr-morpheme which is licensed (under structural adjacency) by an agreement
head that has been valued in the syntactic derivation. Thus, the changes affecting clitic
doubling structures in Sutselvan can be summarized as follows (cf. Chapter 4):
(91) [
CP
XP [
C
C+V+pronoun
j
[
TP
[
DP
double [
D
t
j
...]]
i
[
T
T [
P
t
i
...
[
CP
XP [
C
C+V+Agr [
TP
[
TP
subject
i
[
T
T [
P
t
i
...
to previous double
Identication of feature content: dissociated Agr-morpheme
licensed under structural
adjacency with Agr-on-T
movement)
This analysis makes an interesting prediction: similar to the Germanic dialects which
exhibit complementizer agreement (i.e., West Flemish, Frisian, East Netherlandic,
Bavarian etc.), the presence (or, rather, valuation) of an Agr-morpheme on C should
require structural adjacency between C and T (hosting the syntactically valued Agr-
morpheme). Thus, on the assumption that the subject occupies SpecTP in the RR
dialects under investigation (see above), we should expect that no material (apart from
modal particles and clitics) may intervene between the Agr-morpheme on C and the
subject in SpecTP (see Chapter 3, Section 3.7). Note that this would then constitute
a further deviation from the behavior of optional clitic doubling in the other dialects,
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.41 (2650-2729)
where a number of elements may intervene between the clitic and the full subject DP,
cf. the following examples from Puter, Vallader and Surmeiran:
51
Puter
(92) Co
then
dalungia
immediately
fttan=e
were=clit.3pl
to
Seneca
Seneca
et
and
a
to
sa
his
duonna
wife
avertas
opened
s
up
las
the
avainas
veins
in
in
la
the
bratscha.
arms
Immediately, then, they slashed the wrists of Seneca and his wife. (Linder
1987: 160)
Vallader
(93) Che
which
effet
effect
ha=la
has=clit.3sg.fem
gn
done
aint
in
il
the
pvel
people
la
the
nouva
new
predgia?
sermon
Which effect had the new sermon on the people? (Linder 1987: 154)
Surmeiran
(94) an
in
quel
this
mumaint
moment
=la
is=clit.3sg.fem
riveda
arrived
er
also
la
the
mama.
mother
In this moment, the mother arrived too. (Linder 1987: 154)
Interestingly, this prediction seems to be borne out by the facts. Of the 42 examples of
clitic doubling in Sutselvan listed in Linder (1987: 148ff.), the subject is directly right-
adjacent to the clitic (i.e., the Agr-morpheme on C) in 34 cases (81%). Similar to
the adjacency requirement observed in connection with complementizer agreement in
Germanic, this can be attributed to the fact that the dissociated Agr-morpheme on C
(resulting from the reanalysis of a former clitic) requires structural adjacency between
C and T. However, note that there are ve examples where a modal particle or a short
adverb (marked by underlining) intervene between the clitic and the subject as for
example in the following sentences:
(95) a. Egn
one
da
of
quels
those
lev-i
wanted-1sg
ear
also
jou.
I
I also wanted one of those. (Linder 1987: 148)
b. Anzucuras
eventually
e-la
is-3sg.fem
lura
then
la
the
mort
death
gnida...
come
And eventually, the death then came... (Linder 1987: 155)
Among the intervening elements, I found only ear also (one example), (al)lura then
(three examples) and schon already (one example), that is, short, monomorphemic
particles and adverbs. By assumption, these elements are to be analyzed on a par with
the modal particles in languages such as Bavarian, which are base-generated as ad-
juncts on TP and therefore do not block structural adjacency between C and T (see
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.42 (2729-2786)
Chapter 3, Section 3.7). Finally, there are three examples where a participle intervenes
between C and the subject:
(96) a. Ah, nagn
go-1pl
a
to
tgea,
home
-l
has-3sg.masc
getg
said
igl
the
fumegl...
servant
Well, lets go home, said the servant... (Arquint-Felix; Linder 1987: 155)
b. Dalnga
immediately
en-i
are-3pl
svanidas
vanished
las
the
pusadas.
cutlery-pl
Immediately, the cutlery vanished. (Arquint-Felix; Linder 1987: 161)
c. Igl
the
mestar
master
e-l
is-3sg.masc
sto
been
igl
the
Mia
Mia
la
of
Freana.
Freana
The Mia of Freana was the master. (Linder 1987: 153)
Note that two of these examples originate from the same source (J.C. Arquint, Cuors
da Rumantsch. Adataziun sutsilvana da A.-L. Felix. Cuera, 1958). Thus, it might be
possible that (96ab) reect the grammar of a single speaker that has not yet completed
the reanalysis of clitics as Agr-morphemes on C. Alternatively, one might attribute
the placement of the participle in (96) to inuence from Italian, where the participle
always precedes the subject in inversion contexts.
Of course, the limited amount of available data does not allow for rm conclu-
sions concerning the status of clitic doubling in Sutselvan. This would require a survey
over a greater data set (and in particular, speakers judgments), which I leave open
as a research project for the future. Still, the evidence which is available suggests that
Sutselvan has undergone a change in which clitic pronouns in doubling structures
were reanalyzed as exponents of a dissociated agreement morpheme on C. Appar-
ently, this change led to a further difference between Sutselvan and the other Swiss
RR dialect which has gone unnoticed by Linder (1987), namely a requirement that C
and the subject (or TP) be adjacent. Note that this (somewhat preliminary) conclu-
sion lends further support to the hypothesis that the grammaticalization of agreement
markers proceeds via an intermediate stage where pronouns are reanalyzed as post-
syntactically inserted agreement morphemes (on C) which may eventually turn into
syntactic agreement morphemes in a subsequent development.
The analysis presented so far should lead us to expect that Sutselvan exhibits a
similar behavior as other languages that acquired Agr-on-C. In other words, it should
be possible that a full DP subject can co-occur with the agreement morpheme on C in
all environments, for example in embedded clauses or when the subject occupies the
clause initial positions, as in the following Bavarian sentences:
(97) a. ob-st
whether-2sg
du
you.sg
noch
to
Minga
Munich
kumm-st
come-2sg
b. Mia
we
fom-ma
drive-1pl
hoam.
home
We go home.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.43 (2786-2828)
Surprisingly, I have not managed to nd sentences parallel to (97) in the sources avail-
able to me. This might be taken to indicate that these orderings are ungrammatical
in Sutselvan. In fact, Haiman and Beninc (1992: 192) state that doubling (that is, the
presence of Agr-on-C) is only possible in inverted order in the Swiss RR dialects. How-
ever, it is unclear whether this generalization refers only to main clauses or whether it is
meant to exclude doubling in embedded contexts as well. Here, clearly more empirical
work is necessary to settle this matter.
52
In the following, I will focus on the apparently
clear-cut cases, that is, the constraint against the presence of an agreement morpheme
on C in main clauses that lack inversion.
... Syntactic restrictions on Agr-on-C
In this section, it is argued that the peculiar distribution of Agr-on-C in Sutselvan can
be traced back to restrictions on clitic doubling which hold in all Swiss RR dialects.
In other words, I claim that certain synchronic properties of Sutselvan should ulti-
mately receive a diachronic explanation, in the sense that the distribution of Agr-on-C
still reects restrictions on its historical source, clitic doubling. This suggests that the
change in question is a rather recent development and that Agr-on-C has not yet been
generalized to all instances of C.
Above, I have noted that in all Swiss RR dialects, doubling is restricted to inversion
contexts where both the enclitic and the full nominal follow the nite verb in C. In
contrast, weak pronouns/proclitics do not license doubling and are in complementary
distribution with full nominals:
(98) Ordering restrictions on clitic doubling in the Swiss RR dialects
a. *full DP-nite verb-enclitic...
b. *proclitic-nite verb-full DP...
c. *full DP-proclitic-nite verb...
Here, I want to suggest that these restrictions on doubling can be explained if the
RR dialects in question are analyzed as instances of asymmetric V2. According to
Travis (1984) and Zwart (1993a), supercially identical V2 orders are in fact the result
of two different underlying structures in the Germanic V2 languages (hence the no-
tion asymmetric V2):
53
the nite verb moves to C only in inversion contexts, where
SpecCP is occupied by a non-subject. In contrast, the verb stays behind in In/T in
subject-initial main clauses, with the subject occupying SpecIP/TP. Note that this anal-
ysis seems to be much more natural for the Swiss RR dialects under investigation than
for the Germanic OV languages for which this proposal was originally developed, due
to the SVO character of RR and the fact that the dialects discussed here show general
V-to-T in embedded clauses (see above). Combined with the analysis of clitic doubling
proposed above, an asymmetric analysis leads to the following structure with the big
DP in SpecTP, that is, in clause-initial position:
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.44 (2828-2886)
(99)
T
NP
pro
P V
fin
+ T
t
i
D
clitic
VP
TP C
C
*CP
DP
i
D DP
double
The non-availability of doubling in subject-initial main clauses can then be explained
in the following way. First, if we assume that the licensing of clitics is connected to
certain functional categories (which must provide an appropriate prosodic host), that
is, C in Germanic-type languages and In/T in many Romance languages, the or-
der (98a) cannot be generated on the basis of a structure like (99), since there is no
appropriate prosodic host available for the C-oriented subject enclitics of the Swiss
RR dialects. More precisely, the enclitic has to attach either to a nite verb (in C) or
to a complementizer. Therefore, the order (98a) (full DPnite verbenclitic) cannot
be generated by (i) lowering the enclitic to the right of the nite verb in T or (ii)
raising only the subject to SpecTP, while the enclitic stays behind in SpecP. In both
instances, the enclitic fails to be associated with C and thus does not meet its licens-
ing requirements. What about (98b) and (98c), which both involve a proclitic instead
of an enclitic pronoun? Tentatively, I assume that the capacity to license a specier
where the double is merged correlates with morphological properties of the clitic. Re-
call that in those dialects that exhibit a special series of proclitics (Puter, Surmeiran),
a single proclitic (Puter a-, Surmeiran i-) is used for several person/number combina-
tions (1sg, 3sg.neut, and all plural forms). Therefore, there is a signicant difference
between the proclitic forms, which are entirely homophonous and non-distinctive,
and the enclitics which serve to signal several person/number distinctions (see Table
7 above). By assumption, there is a correlation between this morphological difference
and the capability to project a specier to host the double (cf. Uriagereka 1995 for
related conclusions concerning the morphological strength of D and the capacity to
license a specier). As a consequence, doubling is only available with enclitic forms
which are morphologically strong enough to license a double in their specier.
54
Thus, it appears that at least some of the restrictions on clitic doubling observed
in the present-day Swiss RR dialects can be better understood if these dialects are an-
alyzed as instantiating a form of asymmetric V2 in the sense of Travis (1984) and
Zwart (1993a). Moreover, it seems reasonable to assume that similar restrictions on the
availability of clitic doubling were at work in historical stages of Sutselvan. This makes
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.45 (2886-2924)
available a diachronic explanation for the distribution of Agr-on-C in the present-day
language. When instances of clitic doubling were reanalyzed as instances of Agr-on-C,
the restrictions on clitic doubling made their way into the newgrammar as restrictions
on Agr-on-C. In other words, the restrictions on Agr-on-Cobserved in the present-day
language reect restrictions that formerly constrained clitic doubling: Agr-on-C could
not arise from clitic doubling in subject-initial clauses, since clitic doubling was not
licensed in these contexts in previous stages of Sutselvan. Furthermore, even after the
general development of Agr-on-C, doubling is not possible in subject-initial clauses,
since there is no lexical host available for the Agr-sufx in C (due to the asymmetric
nature of V2 in Sutselvan):
(100)
C
T Agr
TP C
DP
i
C
V
fin
+ T P
t
i
VP
*CP
In addition, the observation that doubling (i.e., Agr-on-C) seems to be limited to
non-embedded clauses in Sutselvan (see note 52) further supports the hypothesis
that Sutselvan represents an early stage of the grammaticalization process in ques-
tion, where Agr-on-C is still conned to the contexts where it originally arose, that
is, inversion in non-embedded clauses.
.. Section summary
This section has demonstrated that the Swiss RR dialects exhibit several waves of a
grammaticalization process in which C-oriented enclitic pronouns are reanalyzed as
agreement markers. I pointed out that we can identify two different syntactic con-
texts where this change took place. First, Section 5.3.2 showed that in a set of earlier
developments affecting 1st and 2nd person forms, enclitic pronouns were directly
reanalyzed as agreement morphemes, giving rise to a limited amount of pro-drop,
especially in 2nd person contexts. I argued that this diachronic process basically re-
sembles the changes in Bavarian discussed in more detail in Section 5.2 above. In other
words, enclitic pronouns were reanalyzed as realizations of a dissociated agreement
morpheme that attaches to the C-head at Morphological Structure. In the Swiss RR
dialects, this reanalysis has been further promoted by a stress rule (penultimate stress
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.46 (2924-2985)
target, Haiman 1971) which may trigger elision of the existing agreement ending if a
clitic attaches to the verb, ensuring that the antepenultimate syllable does not receive
stress. In an independent later development, the new endings are generalized to other
verbal positions, replacing the original agreement formatives, that is, the morpholog-
ical exponent of Agr-on-C is reanalyzed as the exponent of Agr-on-T, the canonical
agreement morpheme. Similar to Bavarian, these changes are made available by the V2
property, which creates linear adjacency between the clitic and the verb.
Moreover, I argued in Sections 5.3.3 and 5.3.4 that at least in Sutselvan, we can
detect a rather recent process in which new agreement formatives develop from clitic
doubling structures where an enclitic pronoun is reinforced by a full nominal for rea-
sons of stress and emphasis. This doubling conguration was analyzed in terms of a
structure where the clitic and the reinforcing DP are merged together as a big DP.
After the reanalysis of the clitic as an exponent of (dissociated) Agr-on-C, the former
double receives the -role previously assigned to the big DP, thereby preserving the
predicates argument structure. Due to the loss of the big DP structure, this change is
also in line with the requirement of structural simplication. Again, the reanalysis in
question can only take place in inversion contexts, where adjacency of the clitic and the
verb is guaranteed by the V2 property. The analysis presented in this section predicts
that the newly created Agr-on-C must be structurally adjacent with T in order to be
licensed. It was shown that this prediction appears to be borne out by the Sutselvan
data, constituting a further deviation from the behavior of (optional) clitic doubling
in the other Swiss RR dialects, which to the best of my knowledge has gone unnoticed
in previous work.
Finally, in Section 5.3.4.1, I claimed that a set of syntactic restrictions on Agr-on-C
in Sutselvan should ultimately receive a diachronic explanation, arguing that the distri-
bution of Agr-on-C reects independent restrictions on clitic doubling, the historical
source of the agreement morpheme on C (these restrictions can still be observed in
the other Swiss RR dialects). This diachronic explanation of the distribution of Agr-
on-C was based on an analysis of the Swiss RR dialects as asymmetric V2 languages,
where the nite verb moves to C only in inversion contexts, but stays behind in T in
subject-initial clauses.
Thus, even in a set of very closely related dialects, we can observe different path-
ways toward newverbal agreement morphology. This observation further supports the
hypothesis that the grammaticalization process in question may be triggered in a vari-
ety of different syntactic environments. In the next section, I will examine another set
of historical scenarios in which C-oriented clitics may turn into new verbal agreement
markers, drawing on data from languages lacking the V2 property.
. Reanalysis of C-oriented clitics in non-V2 languages
This section discusses two instances of a grammaticalization process in which C-
oriented clitics turn into agreement formatives in a non-V2 environment, drawing
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.47 (2985-3042)
on data from various Uto-Aztecan (Section 5.4.1) and Mongolian languages (Section
5.4.2). It is shown that the reanalysis of clitics takes place in different syntactic envi-
ronments in these languages, either under adjacency with a verb moved to T or under
adjacency with a nite verb that has undergone fronting as part of a TP moved into
the left clausal periphery. Importantly, the languages under consideration differ from
the cases discussed so far in that they lacked verbal agreement markers prior to the
reanalysis of subject clitics. It is argued that this difference gives rise to a change where
the clitic is directly reanalyzed as an instantiation of a syntactic Agr-morpheme on T.
Thus, the grammaticalization process does not involve an intermediate stage where the
clitic turns rst into a dissociated Agr-morpheme.
.. Uto-Aztecan
Steele (1977) shows that various Uto-Aztecan languages (spoken in Southern Cali-
fornia and Mexico) exhibit different stages of a development in which independent
pronouns rst become second position clitics and eventually develop into prexal ver-
bal agreement morphology. Unfortunately, there exist only very few written records
of previous historical stages of these languages. However, due to the fact that the rel-
evant change seems to affect the individual languages in more or less parallel fashion,
a comparative study of different Uto-Aztecan languages can be taken to provide im-
portant insights into the various historical stages of the grammaticalization process in
question. Hence, the following discussion is based on the assumption that the indi-
vidual Uto-Aztecan languages represent different stages of the development of prexal
agreement markers from former second position clitics (cf. Steele 1977).
Many Uto-Aztecan languages exhibit a set of enclitic pronouns which historically
derive from full tonic pronouns (cf. Steele 1977, 1978, 1995). These clitics occupy the
second position of the clause, that is, they move into the left periphery and attach to
the rst word or constituent:
Luiseo
(101) hunwuti=pum
bear.object=clit.3pl
seiwun.
are.shooting
They are shooting a bear. (Steele 1977: 539)
Interestingly, in some languages such as Classical Aztec, Huichol, and Pochutla, the
former clitics obligatorily attach to the left of the nite verb and are no longer conned
to second position. This suggests that they have developed into person/number (i.e.,
agreement) prexes on the verb (see below for further evidence in favor of an analysis
in terms of verbal inection):
Classical Aztec
(102) an-teecLasoLa.
2pl-love-us
You love us. (Steele 1977: 539)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.48 (3042-3133)
In Classical Aztec, the (historical) relation between agreement markers and inde-
pendent pronouns is still discernible. Table 8 shows that the onset of the individual
agreement markers is identical to the onset of the respective pronouns (apart from 3rd
person forms):
55
Table 8. Pronouns and agreement prexes of Classical Aztec
(Steele 1978: 611f.).
Verbal agreement prexes Independent pronouns
1sg ni- newaatl
2sg ti- tewaatl
3sg yewaatl
1pl ti- tewaan
2pl an- amwaan
3pl yewaan
In the following, I will discuss the course of the grammaticalization of prexal per-
son/number markers in Uto-Aztecan, drawing on data from a set of languages that by
assumption exhibit different developmental stages of the historical process in question.
According to Steele (1977), the change in question proceeded via an initial stage
where full pronouns attracted to second position turned into reduced clitic forms. In
a later development, these second position clitics were eventually reanalyzed as per-
son/number prexes on the verb. The early beginnings of the grammaticalization of
verbal inection can still be observed in languages such as Mono (Western Numic)
where full, non-reduced (subject) pronouns generally occur in second position:
Mono
(103) nopihweeh
to.home
nGG
I
miyawaih.
will.go
I shall go home. (Steele 1977: 541)
In Mono, the majority of the pronouns are still full forms which do not attach to the
clause-initial phrase/word. However, 1st dual and plural inclusive forms have devel-
oped into enclitics which lean on to the clause-initial element:
56
Mono
(104) nopihweeh=taa
to.home=clit-1pl
miyawaih.
will.go
We shall go home. (Steele 1977: 541)
Tubatulabal and Comanche present a somewhat advanced stage and exhibit a full
paradigm of second position clitics (in addition to full pronouns). The second po-
sition clitics are in complementary distribution with full pronouns (and other DPs),
which clearly shows that the clitics still have argument status. Note that all three lan-
guages that show early stages of cliticization (i.e., Mono, Comanche, and Tubatulabal)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.49 (3133-3195)
exhibit no evidence that topic left dislocation is involved in this process. For exam-
ple, clitics follow clause-initial elements such as innitival clauses and time adverbials,
which presumably do not constitute topics:
Tubatulabal
(105) aanayuwibG=gi
ght.subordinate=1sg
iimi.
went
Without ghting, I went. (Steele 1977: 551)
Tubatulabal
(106) pi=bum
then=2pl
pinahi
must.bring
tohiilin
deers
tuguwayin.
meat
Then you must bring the meat of the deer. (Steele 1977: 551)
Furthermore, if second position clitics originated from reduced resumptive pronouns
in topic left dislocation structures, we would expect the pronouns to co-occur with
topicalized DPs in the early stages of this development. But this is not the case. Neither
in Mono, nor in Comanche or Tubatulabal, clitic pronouns can co-occur with co-
referential topicalized DPs (i.e., there is no clitic doubling in these languages). Thus,
it seems that doubling phenomena are in fact a result, and not the original source of
the grammaticalization process in question. Instead, it seems reasonable to suppose
that clitic doubling came into existence when the former full pronouns underwent
further phonological reduction, presumably due to their placement in clausal second
position (see belowfor some discussion). For reasons of emphasis, the second position
clitics are then reinforced by a full pronoun that bears stress, giving rise to instances of
subject clitic doubling, presumably involving a big DPstructure similar to the Swiss RR
data discussed in Section 5.3 above. This can be observed in all Uto-Aztecan languages
except for Mono, Comanche, and Tubatulabal:
Luiseo
(107) wunaalum=pum
they=clit.3pl
hunwuti
bear.object
seiwun.
are.shooting
They are shooting the bear. (Steele 1977: 547)
Papago
(108) ai=
I=clit.1sg
Gok.
am.speaking
I am/was speaking. (Steele 1977: 547)
Serrano
(109) Gmi=ta=m
you=modal=2sg
payGka
away
miib.
will.go
You will go away. (Steele 1977: 547)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.50 (3195-3271)
Yaqui
(110) nepo
I
ka=ni
neg=clit.1sg
aman
there
nitik.
went
I didnt go there. (Steele 1977: 547)
Southern Paiute
(111) qacu=anga=ni
neg=object.clit=clit.1sg
nG
I
imincuxwavaangwainiaanga.
shall.give.him.to.you
I shall not give him to you. (Steele 1977: 547)
In (107) and (108) (from Luiseo and Papago), the clitic attaches directly to the right
of the double (in both cases a full pronoun), which occurs in clause-initial position. In
(109) and (110) (fromSerrano and Yaqui), the clause-initial position is again occupied
by the double. In contrast to the rst two examples, however, the clitic does not attach
directly to the full form, but rather to a free inectional marker (a modal in (109)
and a negation marker in (110)) which is placed in clausal second position.
57
Finally,
in (111) (Southern Paiute) the clitic attaches to a complex consisting of the negation
marker and an object clitic, followed by the 1sg full pronoun nG.
At rst sight, the examples (107) and (108) seem to suggest that the whole big
DP has moved into clause-initial position (presumably a specier in the left clausal
periphery). However, on the assumption that the relevant construction should receive
a uniform analysis across Uto-Aztecan, the examples in (109) and (110) show that this
cannot be correct: in both (109) and (110), an inectional head intervenes between
the double and the clitic. Therefore, it seems reasonable to suppose that the double
moves independently into the left periphery (an instance of A-movement), while the
clitic attaches to an appropriate host in a separate operation, presumably some kind of
prosodic repositioning which is part of MS or, more generally, the mapping to PF (cf.
e.g. Bonet 1991; Halpern 1992; Schtze 1994; Embick & Noyer 2001).
58
The clitic may
attach to the double (in, say, SpecCP) if no other material is located in an intervening
head position. For the time being, I therefore assume that the big DP moves to SpecTP
prior to fronting of the double (but see below for some modications). Then it seems
likely that the intervening inectional markers occupy a functional head in the C-
domain, which I will refer to simply as C.
In some languages such as Cora, Cupeo, Tarahumara, Tepecano, and Yaqui, the
second position clitics can co-occur with prexal person/number markers on the verb,
the shape of which clearly indicates that they developed from former pronominal ele-
ments (furthermore, these languages exhibit a series of full pronouns), cf. the following
examples from Tarahumara and Yaqui:
59
Tarahumara
(112) c=mu
wh=clit.2sg
ika
wh
k
neg
mu-nak
2sg-want
muh
you
ko
emph
ba?
emph
Why dont you want it? (Steele 1977: 553)
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.51 (3271-3355)
Yaqui
(113) kwarnta
forty
pso
peso
dyryota=ne
daily=clit.1sg
ne-kba
1sg-earn
ani
now
nine.
here
Now I make forty pesos a day here. (Steele 1977: 543)
Since it is rather implausible that both elements constitute pronominal clitics, these
data can be taken to indicate that the proclitic elements are better analyzed as prexal
agreement markers on the verb. This conclusion is further supported by the fact that
in (112), the example from Tarahumara, there is a full pronoun present in addition
to the second position clitic and the person/number marker on the verb.
60
The claim
that the prexal elements in these languages are in fact agreement markers is further
supported by the fact that in Cora (similar to Classical Aztec, see above), the prexal
element is not restricted to second position any longer. In the following example, the
person/number prex occurs on the left edge of a clause-initial complex verb which is
followed by the second position clitic n:
Cora
(114) n-aanain
y
ichee
1sg-was.going.to.build.a.re
n.
clit.1sg
I was going to build a re. (Steele 1977: 543)
We can thus conclude that the various Uto-Aztecan languages exhibit different stages
of the grammaticalization of new prexal agreement markers from former second po-
sition clitics. The evidence available to us suggests that the change proceeded via an
intermediate stage where the clitic is doubled by a full pronoun, presumably for rea-
sons of emphasis. In a subsequent development, these doubling congurations gave
rise to a reanalysis where the clitic is reanalyzed as a verbal agreement prex.
61
The em-
pirical ndings with respect to the different stages of the grammaticalization process
in question are summarized in Table 9.
62
In stage 1 (exemplied by Mono), full pronouns become attracted to second posi-
tion.
63
Presumably, the occurrence in an intonational gap after the clause-initial phrase
accelerates the phonological reduction of the pronouns (stage 2). At some point, the
Table 9. Different stages of the grammaticalization of person/number prexes in Uto-
Aztecan.
Grammaticalization process Languages
Stage 1: Full pronouns are attracted to second
position
Mono
Stage 2: Reduction to second position enclitics Tubatulabal, Comanche
Stage 3: Second position enclitics are doubled
by a full nominal
Luiseo, Papago, Serrano, South-
ern Paiute
Stage 4: Second position enclitics have turned
into verbal prexes
Classical Aztec, Cora, Huichol,
Pochutla, Tarahumara, Tepecano,
Yaqui
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.52 (3355-3429)
second position clitic is reinforced by a full DP (presumably for reasons of emphasis),
leading to the big DP structure proposed above (stage 3). At some point, the result-
ing clitic doubling structures lose their stylistic force and the clitics are reanalyzed as
agreement morphemes on T, leading to stage 4. The transition from stage 3 to stage 4
which leads to the presence of an agreement prex on the verb is schematized in (115).
(115) [
CP
XP [
C
C+pronoun
i
[
TP
[
T
V+T [
P
[
DP
double [
D
t
i
]] ...
[
CP
XP [
C
C [
TP
Agr+V+T [
P
subject ...
to the former double
subject in SpecP
movement)
Similar to the change affecting Sutselvan (see Section 5.3.4), this reanalysis leads to
structural simplication via losing the big DP structure underlying clitic doubling in
the target grammar. Preservation of argument structure is warranted by assigning the
-role to the former double. The newly created agreement morpheme is valued by
initiating an Agree operation that accesses the subject in SpecP, which by assumption
may stay in situ (cf. Note 63).
Finally, the question arises of how the Word building constraint can be satised.
More generally speaking, it is fairly clear that the reanalysis in (115) is possible only
if the former clitic is adjacent to the verb (or an auxiliary-like element base-generated
in T, see above). Interestingly, Steele (1977: 556) notes that the change from second
position clitics to verbal prexes took place only in languages that fail to exhibit the
basic SOV order characteristic of most Uto-Aztecan languages, but show instead SVO
or VSO/VOS word order.
64
This is exactly what is expected under the Word building
constraint: the change described in (115) can only take place if the verb has moved
high enough in the structure to warrant adjacency to the C-oriented (second position)
clitic. By assumption, the target of verb movement is T in these contexts, which also
carries the agreement morpheme resulting from the reanalysis.
65
If the verb stays in
a lower position it is not adjacent to the clitic, which renders a reanalysis of the clitic
impossible in SOV orders.
Of course, this section provided only a cursory discussion of the intriguing syn-
chronic and diachronic facts which can be observed in the various Uto-Aztecan lan-
guages, leaving many details open for future research. Still, it seems that it is possible
to subsume the individual grammaticalization processes under a single explanation
which is based on the assumption that the agreement prexes under consideration
evolved from second position clitics in clitic doubling congurations. The individual
steps taken by this process have been isolated by comparing a set of different languages
that by assumption exhibit different stages of the grammaticalization of agreement
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.53 (3429-3502)
markers from clitic pronouns. Interestingly, the grammaticalization process leading
to new verbal prexes apparently took place only in a subset of the Uto-Aztecan lan-
guages which is characterized by the lack of basic SOVorder. This was attributed to the
fact that only in this subset, the necessary adjacency between the clitic and the verb was
guaranteed. Note that the change in questiondiffers fromthe diachronic developments
discussed in Sections 5.2 and 5.3 in that it did not proceed via an initial stage where
the clitic is reanalyzed as a dissociated Agr-morpheme. Instead, the clitic is reanalyzed
as an instantiation of Agr-on-T which is valued during the syntactic derivation. This
derives from the assumption that dissociated agreement morphemes are parasitic on
a syntactically licensed Agr-morpheme. Accordingly, the latter path is not available in
languages that lack agreement altogether prior to the reanalysis. It follows that the re-
analysis of clitics does not necessarily proceed via an initial stage where the clitic turns
into a dissociated Agr-morpheme, in contrast to what has been hinted at in Section
5.2 above.
As mentioned above, basic SOV order represents an obstacle for the development
of verbal inection from clitic pronouns which apparently can only be overcome by
moving the verb to the left of the pronoun in V2 languages (as in the case of Bavar-
ian, see Section 5.2 above). Still, I have noted in Chapter 1 and Chapter 3 that most
SOV languages exhibit sufxal agreement morphology. Given the rareness of the V2
phenomenon across languages, we clearly cannot assume that all these languages de-
veloped agreement sufxes during a historical stage when they were all V2 languages.
In the next section, I will therefore consider an alternative path which leads to the
development of sufxal agreement markers in an SOV language.
.. Mongolian
In a number of Mongolian SOV languages (Buryat, Dagur, Kalmyk, Moghol, and
Oirat), we can observe a rather recent grammaticalization process that led to the de-
velopment of sufxal agreement markers. This section focuses on Buryat, spoken in
East Siberia and various parts of Mongolia (Poppe 1960; Saneev 1973; Comrie 1980).
Even a cursory look at table 3 reveals the similarities between nominative pronouns
and the subject agreement sufxes found on nite verbs. In addition, a comparison of
the last two columns shows that the series of possessive markers found in that language
apparently developed from the genitive pronouns in a similar fashion.
66
Recall that SOV languages such as Buryat raise an important issue concerning the
grammaticalization of agreement markers. Above, we have noted that basic SOV or-
der often hinders the development of agreement markers from subject clitics, since the
clitic is usually not adjacent to the nite verb. Even more problematic, it has been ob-
served in Chapter 1 that the vast majority of SOVlanguages exhibits sufxal inection.
The assumption that the position of agreement morphemes on the verb reects the
relative order between verb and clitic at the stage when clitics were reanalyzed as agree-
ment markers (Givn 1971, 1976) implies that the verb preceded the clitic when the
latter turned into a verbal agreement marker. As noted in Section 5.2 above, V+clitic
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.54 (3502-3549)
Table 10. Pronouns and agreement in Buryat (Poppe 1960; Comrie 1980).
Nominative pronouns Verbal Agr Genitive pronouns Possessive sufxes
1sg bi -b mi nii -m(ni), -ni
2sg i - i nii -(ni)
3sg - -(i) n
1pl bide -bdi manai -(m)nai
2pl ta -t(a) tanai -tnai
3pl -d -(i) n
orders may be the outcome of a V2 syntax. However, the fact that the V2 property is
actually quite rare among the worlds languages suggests that it is not very likely that
every SOV language with sufxal agreement went through a V2 stage at some point in
its history.
In this section, I will therefore give a sketch of an alternative scenario for the de-
velopment of sufxal agreement in SOV languages. Before we turn to the details of the
relevant proposal, let us reviewsome more data which bear on the syntax of Mongolian
and the development of sufxal agreement in particular.
In present-day Buryat, the morpheme order found on the verb does not corre-
spond to the position of free pronouns. Subject pronouns regularly precede the verb
(and the object, if present), while the agreement markers which still clearly resemble
the pronominal forms are realized as sufxes on the verb:
(116) a. bi
I
jaba-na-b.
go-pres-1sg
I am going (Comrie 1980: 89)
b. bide
we
jerexe-bdi.
shall.come-1pl
We shall come. (Poppe 1960: 123)
To account for the mismatch between word and afx order, one might assume that
the basic word order was VS at the historical stage when the pronouns were reana-
lyzed as agreement markers, followed by a subsequent change to SOV order. However,
it can be shown that Buryat has not undergone any major word order change in its
recorded history (cf. Comrie 1980). Moreover, according to Comrie (1980), the gram-
maticalization of agreement sufxes is in fact a rather recent development, since the
relevant sufxes are absent in earlier stages of the language, that is, Classical Mongo-
lian (attested in written records from the 13th century on, cf. Poppe 1954; Grnbech
& Krueger 1955).
67
Thus, an analysis of the diachronic developments in Buryat has to
address the question of how a language with basic SOV word order can develop agree-
ment sufxes on the verb. Given that subject pronouns regularly appear to the left of
the verb, we should rather expect the development of agreement prexes, contrary to
the facts. Moreover, a further problemarises if we acknowledge that adjacency is a nec-
essary precondition for the reanalysis of a (clitic) pronoun as a verbal afx. In an SOV
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.55 (3549-3619)
language like Buryat, this precondition is often not met, since the subject is regularly
separated from the verb by intervening objects or adverbs.
Interestingly, however, it can be shown that in written records of older stages of
Mongolian, pronouns can also show up in other positions, often following the verb,
cf. the following quote from Poppe (1954: 124):
In the pre-classical language, personal and demonstrative pronouns are often
placed after the nite verb. Sometimes a personal pronoun is placed, as usual,
before the predicate, but is repeated after the latter.
This marked word order option is illustrated by the following examples from (pre-)
Classical Mongolian (taken from Poppe 1954: 124f.; unfortunately, Poppe gives no
examples with doubling):
(117) a. tere
this
met
like
jalbarin
prayer
glemi
say
bi.
I
I am praying in that manner.
b. inegeldki-yi
laughter
yekin
why
tayalamu
like
ci?
you
Why do you like laughter?
Even more interesting, Poppe (1954) assumes that this word order option provided the
context for the grammaticalization of agreement sufxes on the verb:
68
This repetition of the pronoun after a verb has produced personal endings on
verbal forms in certain colloquial languages, e.g., in Buriat, Kalmuck, and others,
e.g., Buriat bi abanab I take, i abana you take, etc. (p. 125)
In other words, it seems fairly clear that the agreement sufxes found in present-day
Buryat originated from a marked word order option in which a weak unstressed pro-
noun followed the nite verb, while an additional full form apparently could be added
in preverbal position (presumably for reasons of emphasis). Thus, it is possible to as-
sume that similar to other instances of clitic doubling, this marked stylistic option lost
its communicative function over time, giving way to a reanalysis in terms of a neutral,
non-marked construction, where the postverbal, unstressed/clitic pronoun is reana-
lyzed as a verbal agreement sufx, while the preverbal pronoun turns into the true
subject of the clause. Unfortunately, I have not been able to track down instances of
doubling in the material available to me. As noted above, Poppe (1954) merely remarks
that doubling is possible, but does not give a relevant example. In other words, at least
for the time being it remains a speculation that instances of doubling were involved
in the diachronic process under investigation. However, note that Classical Mongolian
(Poppe 1954: 123; Grnbech & Krueger 1955: 39) as well as present-day Buryat (Poppe
1960: 119) seem to license pro-drop: similar to Chinese (and other Asian languages),
pronouns need not be realized overtly if their reference can be inferred from the dis-
course context. This observation suggests an alternative scenario for the development
of new agreement markers in Buryat. As shown above in this chapter (and Chapter 4),
the pro-drop property may facilitate the reanalysis of pronouns as agreement markers,
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.56 (3619-3685)
since the relevant -role can be readily assigned to an empty pronoun. Accordingly, I
assume that in the history of Buryat, the availability of pro-drop licensed a direct re-
analysis of postverbal pronouns as agreement markers. However, while this hypothesis
captures the relevant historical developments, it still leaves the exact structural prop-
erties of the reanalysis in question in the dark. For example, it is still left unclear how
weak subject pronouns can come to occupy a postverbal position in an SOV language
such as Buriat.
Comrie (1980: 92) suggests that the reanalysis in question affected pronouns that
were postposed to a position to the right of the verb. He claims that in the Mongolian
languages, this operation is necessary to avoid a sequence where an unstressed element
precedes its selecting head (which seems to be ruled out in Mongolian in general).
69
However, an analysis where the weak pronouns are assumed to undergo right-
ward movement seems to be problematic for the following two reasons. First, it is
widely accepted that weak/clitic pronouns tend to move into the inectional domain
(or even higher) for syntactic licensing (cf. e.g. Cardinaletti & Starke 1999). Second, it
is well-known that rightward movement (if it constitutes an option at all) is normally
conned to heavy elements, as in clausal extraposition or Heavy NP Shift. Hence,
extraposition of pronouns (even full pronouns that may receive stress) usually leads
to ungrammaticality (e.g., in the Germanic languages). Therefore, it is perhaps more
promising to assume that the position of the weak pronouns of (Classical) Mongo-
lian is the result of another syntactic process that guarantees that the pronoun ends
up in a (well-established) defocused position, namely clausal second position. More
specically, I suggest that the weak pronouns of Mongolian adjoin to C, similar to the
Wackernagel clitics in the other languages discussed so far. This cliticization move-
ment is then followed by fronting of a larger constituent, presumably TP, into SpecCP,
which creates the word order exhibited by examples such as (117) above (cf. Kayne
1994; Julien 2002 for similar analyses of SOV order; Julien 2002: 40 explicitly mentions
an analysis of Buryat along these lines).
70
(118)
C
clitic
i
t
TP
C
C
CP
TP
(S) OV t
i
On the assumption that in the history of Buryat, structures such as (118) provided the
syntactic environment for the reanalysis of postverbal pronouns as agreement sufxes
on the verb, the following scenario can be hypothesized for the grammaticalization
process in question:
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.57 (3685-3725)
(119) [
CP
[
TP
t
j
[
T
[
P
t
j
[
VP
object]] V+T]]
i
[
C
C+pronoun
j
t
TP
]]
[
CP
[
TP
[
T
[
P
pro [
VP
object]] V+T+Agr]]]
Structural simplication: loss of TP movement;
subject movement
replaced by an Agree-relation
assigned to pro
Identication of feature content: via Agree with the subject
(pro) in SpecP
T at MS (enabling
As already noted above, the reanalysis in question was presumably facilitated by the
pro-drop properties which already existed in the language prior to the structural
change in (119), providing a means to fulll Preservation of argument structure by the
insertion of an empty pronoun the vacant -role could be assigned to. The change
described in (119) leads to a major structural reorganization in which TP movement
and overt subject movement to SpecTP are lost. This seems to be the most likely out-
come of the reanalysis in question if the requirement of structural simplication is
understood as a pressure to posit the least complex structure compatible with the lin-
guistic evidence. Still, other less dramatic changes are also conceivable. For example,
we might assume that the reanalysis in question does not affect TP-movement, but
merely leads to the loss of overt subject movement (via insertion of pro in SpecP)
which sufces to meet the requirement of structural simplication. In any case, the
reanalysis in question differs from all other changes discussed so far in that the re-
sulting agreement morpheme is realized on a functional head different from the head
the clitic pronoun was attached to (thus, it seems that linear adjacency is sufcient
for the reanalysis to take place). Note that the verb and the pronoun do not form a
head complex in the target grammar. Instead, the verb is contained within TP (ei-
ther in situ or moved to T), while the pronoun has undergone movement to C. In
contrast, in the structure resulting from the reanalysis, the agreement morpheme is
not attached to C, but part of the verbs inectional domain, presumably adjoined to
T. Hence, similar to Uto-Aztecan, the reanalysis leads to a syntactic agreement mor-
pheme that enters the syntactic derivation as an adjunct to a contentful functional
head. Moreover, it must be ensured that the verb can combine with the agreement
morpheme prior to Spell-out. By assumption, this is achieved by verb movement in
(119). Note, however, that in principle, OV-languages always allow the verb to stay in
situ and combine with the agreement morpheme on T via Morphological Merger at
MS, due to the fact that in a strict OV grammar, the verb is always string-adjacent to
the set of right functional heads. In fact, this alternative seems to be more economi-
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.58 (3725-3783)
cal than the derivation involving verb movement. Still, I prefer to stick to the analysis
in (119), which is more in line with the well-established generalization that rich ver-
bal morphology goes hand in hand with overt verb movement (whatever the ultimate
reason for this correlation may be).
. Conclusion
In this chapter, I have taken a closer look at the reanalysis of C-oriented clitics as
markers of verbal agreement, showing that this change is shaped by a variety of dif-
ferent syntactic factors which (partially) derive from the way agreement is established
in natural languages. It has been demonstrated that the grammaticalization process
in question may take place in a number of different scenarios, depending on the spe-
cic syntactic properties of the languages under consideration. In the Germanic and
Rhaeto-Romance varieties discussed in this chapter, the V2 property has been shown
to play an important role in the reanalysis of C-oriented enclitics by creating the nec-
essary adjacency between subject clitic and nite verb, which uniformly gives rise to
sufxal agreement morphology. In non-V2 languages, the transition from C-oriented
clitics to verbal agreement markers may be triggered in alternative environments,
either leading to prexal agreement in SVOlanguages where the verb moves to T (Uto-
Aztecan) or sufxal agreement in SOV languages where by assumption the whole TP
is fronted to a position to the left of the clitic (Mongolian).
Another point of variation involves the realization of the -role assigned to the
pronominal clitic in the target grammar. On the one hand, the relevant -role may be
assigned to a phonologically empty pronoun. It has been argued that this change either
exploits a pro-drop strategy already available prior to the reanalysis (as in Mongolian)
or widens the range of pro-drop, leading to new pro-drop properties absent in the
target grammar (as in Bavarian). In this way, the rise of agreement is intimately linked
to the pro-drop phenomenon. On the other hand, doubling congurations provide
an alternative path to new agreement formatives by assigning the vacant -role to the
former double (as in the Swiss RR dialect Sutselvan).
Additionally, languages may differ with respect to the nature of the agreement
morpheme which results from the reanalysis of a clitic pronoun. In languages which
develop agreement from scratch, the grammaticalization process in question directly
results in a newsyntactic agreement morpheme (e.g., Agr-on-T), which by assumption
enters the syntactic derivation adjoined to a contentful functional head. In languages
with an existing verbal agreement paradigm, however, the creation of new agreement
formatives usually proceeds via an initial stage where a clitic is reanalyzed as the re-
alization of a post-syntactically inserted dissociated Agr-morpheme (e.g., Agr-on-C),
the licensing of which is parasitic on the presence of an Agr-morpheme that has been
valued during the syntactic derivation. In a subsequent development, the phonolog-
ical exponent of this newly created dissociated Agr-morpheme may then replace the
original exponent of the (hierarchically lower) syntactic agreement morpheme (e.g.,
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.59 (3783-3837)
Agr-on-T). As pointed out, this particular pathway also represents a counterexample
to the claim that grammaticalization always leads to new exponents of higher func-
tional heads, as proposed by Roberts and Roussou (2003). These ndings motivate the
following slightly modied grammaticalization cline which captures the distinction
between different types of agreement morphemes (see (45) above, repeated here for
convenience):
(120) free pronoun weak pronoun clitic pronoun (dissociated Agr- mor-
pheme) syntactic Agr-morpheme
Thus, even if we concentrate on changes affecting C-oriented clitics, it is evident that
the reanalysis of pronominal elements as agreement formatives can come about in a
variety of different syntactic environments. Accordingly, attempts to reduce the gram-
maticalization process in question to a single syntactic scenario seem to be misguided.
Still, I have demonstrated that the different historical pathways share a set of clearly
identiable common properties due to the universal nature of the restrictions the re-
analysis of clitics is subject to. That is, all structural changes discussed in this chapter
meet (and are shaped by) the requirements of Structural simplication (the outcome
must be less complex than the target structure), Preservation of argument structure
(the relevant -role must be assigned either to a former double or pro), Identication
of feature content (the content of the newly created Agr-morpheme must be valued),
and the Word building constraint (the Agr-morpheme must combine with the verb
prior to Vocabulary Insertion). Inasmuch as these conditions derive from the gen-
eral theory of agreement laid out in Chapter 3 above, the diachronic data examined
in this chapter provides further support for the particular approach to syntactic agree-
ment advocated in this book. In particular, the hypothesis that Agr-morphemes do not
occupy a unique position in the structure of the clause, but rather are parasitic on con-
tentful functional categories such as C, T, or , is corroborated by the observation that
in V2 languages such as Bavarian, Cimbrian, Alemannic and Rhaeto-Romance, new
agreement formatives are initially conned to the C head before they replace the orig-
inal exponent of the canonical subject agreement morpheme on T in a subsequent
development. In addition, the fact that this historical development typically leads to
instances of multiple agreement is successfully captured by an analysis according to
which the change in question proceeds via an intermediate stage where clitics are re-
analyzed as the realization of a dissociated morpheme which is inserted and valued
post-syntactically.
Finally, note that the particular analyses presented in this chapter can be taken
to alleviate the problems raised by the observation that cross-linguistically, agreement
markers tend to be realized as sufxes (the so-called sufxing preference, see Chapter
1): If we take a look at the scenarios discussed in this chapter, it appears that most
of them give rise to sufxal agreement markers. Thus, a study of the transition from
C-oriented clitics to agreement markers seems to support the conjecture (cf. Chapter
4) that the sufxing preference should receive a historical explanation, in the sense
that it results at least partially from the way agreement markers come into existence
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.60 (3837-3900)
historically. More specically, it seems that cross-linguistically, the grammaticalization
process in question is more likely to be triggered in syntactic environments which give
rise to sufxal agreement morphology.
At this point, it should again be stressed that the syntactic factors discussed in this
chapter are all taken to restrict the grammaticalization process in question. In other
words, they are necessary, but crucially not sufcient conditions for the change to take
place. This raises the question of whether it is possible to identify any causal factors
which lead the learner to reanalyze pronouns as agreement markers. This question is
dealt with in some more detail in the following chapter, where it is argued that the
ultimate trigger for the reanalysis of pronominal elements is morphological in nature.
Notes
. However, see Nbling (1992) for an analysis of the relevant elements as pronominal enclitics.
. In a number of dialects, the 1pl clitic -ma has developed similar properties as the 2nd person
forms, cf. Section 5.2.1.3.
. Thus, it is assumed that the forms 2sg /-st/, 2pl /-ts/ are solely agreement markers; in other
words, there are in fact no 2nd person clitics in most Bavarian dialects. Otherwise, the question
would arise of why it is not possible to add a clitic onto an inected verb or complementizer in
C, giving again rise to forms such as *kummts-ts or *ob-ts-ts. The lack of 2nd person clitics is
presumably the result of the reanalysis of these forms as part of the verbal agreement morphol-
ogy, see below. Note, however, that there are some Carinthian dialects (spoken e.g. in Pernegg,
cf. Lessiak 1963) where such forms seem to exist with 1pl mr (which apparently developed into
another instance of Agr-on-C in these dialects, see below):
(i) Kher-m6=mr
belong-1pl=clit.1pl
wir
we
a aufn?
up
Do we belong at the top?
(ii) wi6-mr=mr
the way-1pl=clit.1pl
wir
we
wln
want
The way we want (it) (Pernegg, Lessiak 1963: 204)
The multiple realization of -mr can be accounted for if we assume that in this dialect (in contrast
to other Bavarian varieties), the reanalysis of the clitic as an instance of Agr-on-C did not lead
to the loss of the clitic itself, but rather to a situation where the source and the result of the
grammaticalization process in question continue to exist side by side in the grammar (which is
actually a common feature of grammaticalization processes; cf. Chapter 2).
. The early OHG manuscripts written in the monastery of Fulda show this change in the
process of its development, cf. the Hildebrandslied (preserved in an early 9th century copy of
the original text dating from the late 8th century), the Basel Recipes (around 800), or the Ta-
tian (translated around 830840. This translation was then copied in the second half of the
9th century); see Brinkmann (1931), Moulton (1944), Sievers (1961), and Sommer (1994) for
details.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.61 (3900-3972)
Chapter 5. The reanalysis of C-oriented clitics i:
,. Note that verbs such as fear may optionally take V2 complement clauses in German. There-
fore, the verbs nimist take-2sg and arnost earn-2sg can be taken to occupy the C-head of an
embedded V2 clause.
o. Here, one might speculate that the enlargement -t in (9) is actually to be analyzed as a 2sg
enclitic. This would explain the absence of the enlargement in the presence of the full pronoun
thu in (10), on the assumption that clitics and full pronouns are in complementary distribution
in OHG. However, in the OHG Tatian, there is at least one example where a full pronoun and a
clitic form do co-occur:
(i) Eno
qprt
thu
you
bistu
are-clit.2sg
mera
more
unsaremo
our
fater
father
Iacobe...
Jacob
Are you more powerful than our father Jacob...? (Tatian, 87.3; Sievers 1961: 119)
. This was pointed out to me by Tom McFadden.
8. This predicts the existence of a historical stage of Bavarian that exhibits (apparent) doubling
phenomena in main and embedded clauses as a result of Agr-on-C, but crucially lacks the new
ending on verbs in clause-nal position.
. Rowley (1994: 494) reports that a similar development can be observed for 3pl in some
North Bavarian and East Franconian dialects. In the following examples, an overt 3pl subject
is accompanied by the relative marker wf which carries the (verbal) agreement ending 3pl -n:
(i) dez
that
wfu-n
what-3pl
fn6r!
others
vfdein
earn
moust-n
must-3pl
The things that others have to work for (dialect of Erkersreuth)
(ii) wf-n
what-3pl
z!
they
wl-n
want-3pl
What they want (dialect of Kastl)
:o. Recall that in contrast to Bavarian, the reanalysis that led to the rise of Agr-on-C did not
lead to the loss of the clitic, but gave rise to a situation where the source and the result of the
grammaticalization process continue to exist side by side in the new grammar, resulting in the
unusual form of triple subject marking in this dialect.
::. These dialects are spoken in an area the boundaries of which are (roughly) marked by Cham
in the west, Lam in the east, Furth i. W. in the north, and Ktzting in the south, cf. Kollmer
(1987, I).
:i. Note that even if pro moved to SpecTP, the requirement of structural simplication would
still be satised, due to the loss of clitic movement to C. However, recall that I have argued above
(in Chapter 4) that empty pronouns cannot undergo overt movement since they fail to provide
SpecTP with phonological content, which would be necessary to satisfy the EPP (cf. Holmberg
2000).
:. Thus, it appears that the change illustrated in (22)(23) apparently led to the loss of an EPP
feature in T in the contexts where the reanalysis took place. Given the often stated universality
of the EPP (cf. e.g. Chomsky 1995, 2000; Collins 1997), this raises the question of whether such
a change is possible at all. However, it is a well-known fact that subjects may remain in a lower
(presumably P-internal) position in German (cf. Grewendorf 1989; Haider 1993). This suggests
that T may lack an EPP-feature in German. Accordingly, I am led to assume that the reanalysis
in question was facilitated by this special trait of German.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.62 (3972-4052)
i:8 The Rise of Agreement
:|. Recall that in V2/V1 contexts, replacement of Agr-on-T by Agr-on-C is required by the as-
sumption that in a complex head adjunction structure, only the highest Agr-morpheme (of a
set of identical Agr-morphemes) is spelled-out (cf. Chapter 3, Note 62).
:,. However, one might argue that in examples such as (27ac), there is actually a complemen-
tizer dass that present to which the inectional ending attaches (cf. Harnisch 1989; Nbling
1992). On this assumption, one can maintain that the set of hosts which Agr-on-C selects for
is rather limited (only C
0
elements, i.e., complementizers and nite verbs). This analysis is sup-
ported by the fact that the complementizer can also be overtly present in the above examples, cf.
(i) Du
you
soll-st
should-2sg
song
say
[
CP
[ an wichan
which
Schuah]
shoe-2sg
(dass)-st
that-2sg
[
IP
du
you
wui-st ]]].
want-2sg
You should say which shoe you want.
:o. Interestingly, the same kind of doubling is found with objects, in contrast to Bavarian:
(i) a. vorsch=me
ask=clit.1sg.acc
net
not
miar.
me
Dont ask me.
b. dar
he
vorscht=de
asks=clit.2sg.acc
di.
you
He asks you. (Lusern; Tyroller 2003: 157)
(ii) a. du
you
pxes=mar
scold=clit.1sg.dat
miar.
me.dat
You scold me.
b. i
I
blta=ter
wished=clit.2sg.dat
dir
you.dat
contraren...
tell
Iwished to tell you... (Roana; Schweizer 1952: 28)
This might be taken to suggest that Cimbrian develops some form of object agreement previ-
ously absent in the grammar, similar to present-day Spanish (cf. e.g. Suer 1988).
:. This development goes hand in hand with the rise of new full pronominal forms which are
clearly inuenced by the neighboring Italian dialects: wir/endsch andre we others, ir/ouw andre
you.pl others.
:8. This is reminiscent of the expansion of 1pl -ma to verbs in clause-nal position in some
Lower Bavarian dialects, see Section 5.2.1.3 above.
:. The distribution of clitics and full pronouns parallels the distribution of pro and full pro-
nouns in Standard Italian (and the 2nd person forms in Bavarian): In Pomattertitsch stressed
subject pronouns (as well as nouns) mainly occur in emphatic, focused or contrastive contexts,
with an explicit referential function. They are usually placed pre-verbally but their position
is in fact rather free. On the other hand, clitics always occur in post-verbal position in all
pragmatically unmarked contexts. (Dal Negro 2004: 164).
io. Dal Negro (2004: 178f.) speculates that the change in question is triggered by the fact that it
serves to repair an otherwise defective agreement paradigm (see Chapter 6 for some discussion).
i:. Interestingly, however, the verb-clitic sequences are very frequent in embedded contexts
where they apparently replace the German pattern in which clitics normally attach to the
complementizer in embedded clauses, cf. the following examples in which the conjunction is
followed by V1-orders (inversion):
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.63 (4052-4117)
(i) a. ds
the.gen
morgen-tsch
morning-gen
wen
when
si=wer
are=we
fertig
nished
gsi
been
im
in.the
gad
stall
In the morning, when we had nished working in the stall
b. der
the
manut
month
wa
that
si=wer
are=we
di
over.there
gsi
been
During the month when we were over there (Dal Negro 2004: 168)
Indirect questions play a somewhat special role in that they exhibit this type of inversion with
all kinds of subjects (clitics and full nominals) in the speech of all informants consulted by Dal
Negro:
(ii) gang
go
lge
look
wa
where
ischt
is
der
the
nonno
grandfather
Go looking where the grandfather is. (Dal Negro 2004: 168)
. Cf. the following statement taken from Dal Negro (2004: 166f.): Also in Pomattertitsch ich
and du are the most frequently omitted pronouns, although other pronouns can be omitted.
[...] As a matter of fact, zero subjects in Walser dialects can be interpreted as one of the possible
outcomes of unstressed subject pronouns. In the case of zero subject sentences both the phonic
form and the referential strength of the pronoun are null. Besides, the grammaticality of zero
subject sentences provides the necessary background that allows the interpretation of verb-subject
clitic sequences as units of verb followed by its ending in which no explicit subject is present and
the information on personal agreement is carried by the new agglutinative endings. [my emphasis,
E.F.]
. On the assumption that agreement morphemes usually exhibit narrow selectional restric-
tions concerning their host, this outcome actually seems to be more likely than the development
of complementizer agreement observed in Bavarian.
. The fact that the development of 2sg /-st/ is not conned to Bavarian, but rather was al-
ready well underway in early Old High German, should lead us to expect that pro-drop and
complementizer agreement (i.e., Agr-on-C) can be found in other German varieties as well in
the context of 2sg. While standard German does not allow referential pro-drop, there are other
non-standard varieties apart from Bavarian in which pro-drop is available with 2sg subjects,
cf. the following examples from Swabian (Carola Trips, personal communication) and Zurich
German (Weber 1987: 174):
(i) Hascht
have-2sg
pro de
the
Peter
Peter
gesehe?
seen
Swabian
Have you seen Peter?
(ii) Hscht
have-2sg
pro e
him
gsee?
seen?
Zurich German
Have you seen him?
In other words, it appears that referential pro-drop is actually quite widespread in (non-
standard) varieties of German in the context of 2sg. Concerning the presence of Agr-on-C, note
that there is at least one example in the OHG Tatian in which a full pronoun co-occurs with an
apparent clitic. Since clitic doubling is otherwise not attested in OHG, the clitic is presumably
to be analyzed as the realization of Agr-on-C:
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.64 (4117-4200)
(iii) Eno
qprt
thu
you
bistu
are-2sg
mera
more
unsaremo
our
fater
father
Iacobe...
Jacob
Are you more powerful than our father Jacob...? (Tatian, 87.3; Sievers 1961: 119)
Accordingly, one might speculate that complementizer agreement had originally been available
in other varieties of German as well (albeit restricted to 2sg) and that it disappeared when the
exponent of Agr-on-C was reanalyzed as the canonical realization of 2sg agreement (the loss
of this dialectal trait was presumably furthered by sociolinguistic factors, that is, the develop-
ment of a standard language). In contrast, further reanalyses of enclitic subjects leading to new
formatives for 2pl (/-ts/) and 1pl (/-ma/) reinforced this grammatical option in Bavarian.
. Linder (1987: 35) attributes this to the fact that some authors/speakers consider inversion to
be an un-RR Germanic trait that should be avoided in pure RR. Interestingly, most examples
cited by Oetzel (1992) involve a conguration where a scene-setting adverbial precedes the sub-
ject. This is reminiscent of the fact that in many older Germanic languages such as Old English
and Old High German, we can observe similar violations of the V2 constraint in connection
with scene-setting adverbs (cf. Fu 2003 for some discussion), cf. the following examples from
Old English:
(i) a. [fter
after
eossum
these
wordum]
words
[se
the
Hlend]
Savior
cw
spoke
to
to
his
his
leornerum...
disciples
(Blickling 135; Swan 1994: 241)
b. [Her]
in-this-year
[Oswald
Oswald
se
the
eadiga
blessed
arceb]
archbishop
forlet
forsook
is
this
lif.
life
(ASC, Laud (992); Kroch & Taylor 1997: 304)
. Other differences between the Swiss RRdialects and the Germanic V2 languages concern the
placement of negation and the behavior of clitic clusters. In Puter and Vallader, the sentential
negation nu(n) precedes the nite verb in inversion contexts, cf.
(i) A
to
Gustav
Gustave
nu
not
pudaivla
could-she
cuntradir...
contradict
She could not oppose Gustav... (Puter; Oetzel 1992: 28)
(ii) Ma
but
il
the
min
my
brn
brown
nun
not
han=a
have=they
pud
can
verer.
see
But they could not see my brown (cat). (Vallader; Oetzel 1992: 62)
This resembles the behavior of negation in many other Romance languages. Note that similar
facts can be observed in Old English and Old High German, where sentential negation is ex-
pressed by a clitic which accompanies verb movement to C, cf. the following example from the
Old High German Isidor translation:
(iii) endi
and
mina
my
milthissa
compassion
ni
not
nimu
take
ih
I
ab
from
imu.
him
And I will not take my compassion from him. (Isidor, 627; Robinson 1997: 16)
Another deviation from the kind of V2 observed in Germanic can be found in Puter, where the
proclitic subject pronoun a can amalgamate with an object clitic, resulting in a clitic cluster in
clause-initial position (Linder 1987: 104):
(iv) A-t
clit.1sg.nom-clit.2sg.dat
racumand
recommend
mia
my
Nanigna!
Nanigna
I recommend my Nanigna to you!
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.65 (4200-4268)
Chapter 5. The reanalysis of C-oriented clitics ii:
(v) A-l
clit.1sg.nom-clit.3sg.acc
faregia
make-fut
sager
try
qualchosa
something
dbun.
good
I will make him try something good.
Note that (v) is particularly interesting, since it represents an example of clitic climbing where
the subject of the embedded innitival fuses with the subject of the causative.
i. Recall that in Puter and Vallader, the sentence negation nu(n) is a clitic that attaches to
the left of the verb and accompanies verb movement to higher head positions (cf. note 26).
Therefore, the relative position of the nite verb and negation cannot be used as a diagnostic of
the position of the nite verb in these dialects.
i8. Another series of agreement endings (the so-called secondary desinences, cf. Haiman and
Beninc 1992 for details) is found in the imperfect indicative, the present subjunctive, the im-
perfect subjunctive, and the future(s). The secondary desinences are generally unstressed, which
gave way to phonological reduction and a higher degree of syncretism.
i. At least in the Engadin dialects Puter and Vallader, the accusative precedes the dative in the
unmarked order, in contrast to German. In the other Swiss RR varieties, the picture is less clear,
possibly due to the inuence of German.
o. Note that (60b) also exhibits a violation of the V2 constraint, since the object pronoun
intervenes between the wh-phrase and the nite verb.
:. According to Linder (1987: 13f.), in Puter, the reduced form -a can be used as a C-oriented
enclitic for 1sg and 3pl as well (attaching to the nite verb or a complementizer). In the latter
usage, it competes with the (older) variants 1sg -i and 3pl -e. The enclitic -a is also found as
3sg.neut when its host ends in a consonant, and as 1pl if the resulting word complies with certain
stress rules (penultimate stress target, Haiman 1971; see below for details). -a is not found as a
2pl enclitic. In contrast, the corresponding vocalic clitic i of Surmeiran is apparently restricted
to proclitic contexts (at least in the written language).
i. The distinction between 3pl.masc and 3pl.femfull forms is found in all ve Swiss RR dialects
under investigation. In addition, Puter and Vallader exhibit feminine variants of 1pl and 2pl
(nussas and vussas, respectively). The individual dialects differ with respect to the shape and the
availability of the clitic forms.
. The dialects under investigation differ with respect to the frequency of enclitic pro-drop. For
example, Linder (1987: 34) notes that enclitic pro-drop is much more frequent in Puter than in
Vallader.
|. Haiman and Beninc (1992: 93) suggest that this development was caused by a pressure for
paradigmatic coherence: in the rst conjugation, 2sg and 3sg had two syllables, while 1sg had
only one. This difference was then leveled by sufxing another vowel.
,. Haimanand Beninc (1992: 95) argue that the old ending for 1pl -(a)in is also an innovation,
which derived historically from a former 3sg pronoun that was reinterpreted as an agreement
sufx for 1pl (within the pronominal paradigm, a similar development is exhibited by French
on): First, the 1st plural was expressed by homo/unus + 3sg. [...] Second, this PRO form ap-
peared postverbally in inverted order as a clitic. Finally, -VN was reinterpreted as a bound sufx
on the verb stem, obligatory in both direct and inverted word order.
o. Widmer (1959: 95) notes that in the translation by Bifrun there are still also examples which
reect the older usage. In the following example (Bifrun, Cor. I, 9.4), the verb exhibits the old,
non-enlarged ending in an inversion context, with the 1pl pronoun following the verb:
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.66 (4268-4341)
(i) Nun
not
haivan
have-1pl
nus
we
forsa
power
da
to
mangir
eat
&
and
da
to
baiver?
drink
Have we not power to eat and to drink?
. Other Swiss varieties of RR such as Surselvan and Sutselvan (and the majority of the RR
dialects spoken in Northern Italian) still show the original agreement ending 2sg -s.
. Note that this correlation holds only for Puter, Vallader and Surmeiran in an obvious form.
Surselvan and Sutselvan showpro-drop in 2nd person contexts although they do not exhibit the
innovation -s+t. Here we might speculate that the latter dialects underwent a similar historical
process as Puter, Vallader, and Surmeiran (i.e., a reanalysis of enclitic -t(i)), but lost nal -t in
a later development (note that loss of nal -t can also be observed in spoken present-day Puter
and Vallader, cf. Linder 1987: 56). Alternatively, it is possible to assume that pro-drop with 2nd
persons is the retention of an older systemof pro-drop (similar to Old French, cf. Roberts 1993a;
Vance 1997), as suggested by Linder (1987: 56).
. Haiman and Beninc (1992: 97) note a similar development in the Gorizian dialect of Friu-
lian, where the 2pl atonic pronoun subject -o has developed into an invariable sufx on the
verb:
(i) o
you.pl
fevel-ez-o
talk-2pl
You talk.
. In the 2pl present indicative, Vallader exhibits the ending -aivat, which is unique among
the Swiss RR dialects (e.g., chant-aivat sing-2pl). Traditionally, it is assumed that this formalso
involved the reanalysis of a clitic form and proceeded in two steps (cf. Gartner 1883; Widmer
1959; Linder 1987). First, a reduced form of the 2pl pronoun vos (clitic va without nal -s; a
similar formvo still exists in present-day Puter) was reanalyzed as an enlargement of the existing
2pl ending -ai, cf. the following example from early Vallader (Chiampel, Ps. 58) cited in Widmer
(1959: 99):
(i) Pud-aiw
can-2pl
wuo
you
foarsilg
perhaps=the
uaira
truth
dyr?
say
Can you perhaps tell the truth?
In a second step, the ending /-t/ was added in analogy to the other tenses where 2pl is signaled
by the agreement sufx /-t/. A somewhat simpler account is proposed by Haiman and Ben-
inc (1992), who argue that Vallader borrowed the unusual ending from the almost identical
imperfect paradigm.
. As noted above in Section 5.2.2, this analysis raises some questions concerning the status of
the EPP. These are probably even more apparent in the present case since it seems that in the
RR languages, the subject moves to SpecTP without exception, in contrast to German/Bavarian.
However, note that even under the more traditional assumption that pro occupies SpecTP in
the RR dialects under investigation, the requirement of structural simplication would still be
satised, due to the loss of clitic movement to C.
. As already mentioned, similar facts can be observed in Non-Standard French and a number
of Northern Italian dialects (cf. Chapter 6 for more detailed discussion).
. Note that different clitics exist for 3sg.masc and 3sg.fem forms. With 3rd person plural sub-
jects, no gender distinctions are marked by the clitic. Doubling with 3rd person neuter forms is
only found in Surmeiran, where it is apparently restricted to wh-questions:
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.67 (4341-4410)
Chapter 5. The reanalysis of C-oriented clitics ii
(i) Ma
but
tge
what
-gl
is-clit.3sg.neut
chegl?
that
But what is that? (Surmeiran; Linder 1987: 156)
||. Note that (77b) is a very early example of clitic doubling, taken from Bifruns bible transla-
tion (1560), where -s presumably is still a clitic form. This might be taken to suggest that dou-
bling structures were also involved in the early development of 1pl nal -s, in contrast to what
has been claimed above. However, all other instances cited by Linder (1987) and Haiman and
Beninc (1992) lack a postverbal full pronoun. In addition, the assumption that the new agree-
ment ending -ains developed from doubling structures would leave us without an explanation
for the fact that 2nd person contexts usually license pro-drop in RR.
|,. Linder (1987) does not give any examples for clitic doubling with 1st person subjects in
Vallader.
|o. Note that there are no examples of doubling with 2nd person subjects, since there are no
enclitic pronouns for 2nd person available.
|. A related idea is proposed in Simpson and Wu (2002), who suggest that in general, AgrPs
develop out of FocPs that are originally selected by some higher functional head (cf. Chapter 1
for discussion).
|8. See Grewendorf (2002) for an analysis of left dislocation in German which is based on the
assumption that the dislocated phrase and the resumptive pronoun start out together as a big
DP.
|. As pointed out to me by Ede Zimmermann, one might wonder whether the loss of stylistic
force should not be analyzed as an earlier, separate structural change, which manifests itself in
a different LF, for example. However, it seems to me that it is not necessary to posit such an
intermediate stage. More specically, I maintain that a direct reanalysis of a formerly marked
construction is possible if the resulting structure is compatible with the original information-
structural function in the target grammar. For example, the reanalysis of a former topic as a
subject does not necessarily lead to a change in informationstructure, since (denite/referential)
subjects act as default topics in the unmarked case (cf. e.g. Givn 1976). Accordingly, a reanalysis
as neutral syntax is possible as soon as the corresponding structure is no longer unambiguously
marked (e.g., by a special intonation) as a topic construction.
,o. The reanalysis in question was presumably promoted by another trait of Sutselvan that fur-
ther weakened the evidence for a pronominal status of the clitics. As noted above, in all Swiss
RR dialects, the structure of nite verbs is governed by a stress rule which forbids that the main
word stress falls on the antepenultimate syllable (the so-called penultimate stress target Haiman
1971; cf. Linder 1987; Haiman & Beninc 1992 for discussion). In some contexts, attachment of
the clitic forces elision of the original verbal agreement ending (3sg.fem. -a in the example at
hand) to ensure that the antepenultimate syllable does not receive stress:
(i) ella
she
chant-a
sing-3sg
*chant-a=la
sing-3sg=clit.3sg.fem
chant=la
sing=clit.3sg.fem
In other words, adherence to the penultimate stress target leads to surface forms where a clitic
apparently replaces the original verbal agreement morphology. Given that the necessary con-
ditions for a reanalysis are fullled in Sutselvan (see below), it is reasonable to suggest that
examples such as (i) tipped the scales in favor of a reinterpretation of the former clitics as
agreement markers (see Chapter 6 for discussion).
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.68 (4410-4494)
. I will not attempt to provide a detailed analysis of these word order facts here apart from
suggesting that variable word order is associated with the optional projection of TopPs/FocPs
within the middle eld/IP domain (cf. Grewendorf 2004).
. The following example fromSutselvan (Linder 1987: 240) seems to suggest that doubling/Agr-
on-C is actually conned to main clauses (recall that doubling/Agr-on-C is otherwise almost
obligatory in Sutselvan):
(i) Tscheart
certain
ear
was
igl
it
[cigls
that=the
pocs
very-few
da
of
nus
us
earan
were
egnea
once
stos
been
a
to
Casti...]
Casti
It was certain that very few of us have ever been to Casti.
However, doubling is possible if the embedded clause exhibits inversion/V2. The following
example from Puter exhibits doubling in a V2 clause introduced by the conjunction per che
since:
(ii) Sullet
Only
uain
is
el
he
dit,
mentioned
[per che
since
da
by
natra
nature
e-l
is-clit.3sg.masc
el
he
sullet
only
lg
son
da
of
Dieu...]
God
Only he is mentioned, since by nature, he is the only son of God... (Puter, Linder
1987: 151)
. The termasymmetric V2 is sometimes also used to refer to the Germanic V2 languages that
showa main/embedded asymmetry with respect to the position of the nite verb (in contrast to
symmetric V2 languages such as Icelandic and Yiddish).
. Note that the order (98b) is presumably ruled out for independent reasons: after movement
to SpecTP, the full nominal must occupy a preverbal position. Furthermore, if both the clitic
and the full DP are merged together in a big DP headed by the clitic, movement of the clitic
to SpecTP presumably cannot strand the subject in a lower position. This would require that
the double is scrambled out of the big DP to a position below T (simply labeled FP here). In
a subsequent operation, the big DP-remnant (containing the clitic) moves to SpecTP, thereby
crossing the double:
(i) *[
CP
[
TP
[
DP
t
i
[
D
clitic ...]]
j
[
T
V+T [
FP
double
i
... [
P
t
j
...
Following Grewendorf (2003), such a derivation can be ruled out as an instance of Improper
Remnant Movement, due to the fact that the movement that creates the remnant (in the case
under consideration, scrambling of the subject) is of a higher type than the movement operation
that affects the remnant subsequently (A-movement).
. Note that Classical Aztec is another example of a language that lacks 3rd person agreement
markers (cf. Chapter 1 and Chapter 6 for discussion).
. Thus, similar to other grammaticalizationprocesses discussed so far, the change in question
affects certain pronominal elements before it affects others. Generally speaking, it seems that 1st
and 2nd person elements are leading the charge (see the discussion and references in Chapter 1;
see Chapter 6 for an explanation).
. In many Uto-Aztecan languages, C-oriented clitics do not show up in second position if
certain inectional markers occupy this position. In these cases, the pronominal clitic usually
attaches to the right of these markers, cf. the following quote from Steele (1977: 549) and the
examples from Luiseo in (i): In fact, elements which correspond roughly to what has been
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.69 (4494-4563)
analyzed as the AUX in English tense, aspect, and modality commonly occur in sentential
second position, as do conjunctions, negatives, and question-markers.
(i) a. nanatmalum
girls
xu=m=po
modal=clit.3pl=asp
hengemalumi
boys
ari.
kick
Girls should kick boys.
b. nawitmal
girl
qay
neg
hengeemali
boy
ariq.
is.kicking
The girl isnt kicking the boy.
c. nawitmal=su
girl=qmarker
hengeemali
boy
ariq?
is.kicking
Is the girl kicking the boy?
. Further support for an analysis in terms of late repositioning comes from the fact that
clitic placement does not respect (syntactic) phrasal boundaries. In the following example from
Tarahumara, the clitic attaches to the rst word of the fronted DP, separating the noun and the
adjective:
(i) [semati=ne
nice=clit.1sg
napaha]
shirt
rarimea
gonna.buy
ar.
probably
I am probably going to buy a nice shirt. (Steele 1977: 554)
. Further evidence for this assumption is provided by an 18th century grammar of Yaqui
(Velasco 1737), where it is explicitly stated that subject pronouns must occur in second position.
In other words, it can be concluded that the prexal person/number markers which are similar
in shape to the second position clitics have developed subsequently.
. Still, there are indications that this grammaticalization process has not yet been completed
in Tarahumara and Yaqui. Steele (1977: 553f.) observes that the prexal person/number markers
are not obligatory in these languages. In Tarahumara, the presence of the verbal prex is bound
to the presence of a second position clitic, that is, second position clitics can occur without the
verbal prex but not vice versa, cf. (i). In contrast, all three logical possibilities can be observed
in Yaqui: second position clitics can co-occur with person/number prexes (as in (113)), and
both elements can occur alone, cf. (ii) and (iii) (all examples from Steele 1977: 554):
(i) semati=ne
nice=clit.1sg
napaha
shirt
rarimea
going.to.buy
ar.
probably
(Tarahumara)
I am probably going to buy a nice shirt.
(ii) tuk=ne
yesterday=clit.1sg
antnyta
Antonio
bicak.
saw
(Yaqui)
Yesterday, I saw Antonio.
(iii) inepo
I
ne-a-meak.
1sg-it-threw
(Yaqui)
I threw it.
These differences can be analyzed as an instance of grammar competition (Kroch 1989 and Chap-
ter 2 above), where speakers have command over more than one internalized grammar: an older
grammar A, producing the sentences without the person/number prex on the verb and a newer
grammar B where former second position clitics have been reanalyzed as (obligatorily present)
verbal agreement markers. The fact that in Tarahumara, the agreement marker cannot occur
without a second position clitic suggests that the reanalysis in question has not given rise to a
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.70 (4563-4636)
pro-drop grammar: in grammar B, an overt pronominal element (e.g., a second position clitic)
must be present in addition to the verbal agreement marker, similar to Standard French, for ex-
ample. See Note 64 below for an alternative explanation that links the distribution of agreement
markers to the syntactic environments (word order) where the change in question was licensed.
. Steele (1977: 554ff.) argues that at least in Tepecano, the verbal prexes developed via a di-
rect reanalysis of second position clitics, which possibly does not require an intermediate stage
of clitic doubling. Evidence for this claim comes from Tepecano examples where the element in
question can apparently act as both a prex/proclitic on the verb and a second position clitic:
(i) ndedos
my.ngers
n=an=ahohoinda.
introducer=1sg=will.shake.them
I will shake my ngers. (Tepecano; Steele 1977: 543)
Note, however, that the verbal person/number markers of Tepecano otherwise behave differently
than second position clitics. For example, they can occur in clause-initial position, as in (ii),
where the clitic attaches to an auxiliary-like tense/aspect marker.
(ii) an=ti
1sg=t/asp
nauw
napal
kGGs.
transplanted
I transplanted napal. (Tepecano; Steele 1977: 555)
. As noted above, the grammaticalization process in question is not completed in languages
such as Tarahumara and Yaqui which still show competition between old, non-prexed verbs
and new prexed forms (presumably an instance of grammar competition). Steele (1977: 557)
assumes that the change from second position clitics to verbal prexes (i.e., stage 4 in Table
9) has been promoted by the fact that the verb attracts grammatical elements such as clitics
(and modals, auxiliaries etc.): It is the attraction of the verb for grammatical elements that
encourages the development of clitic pronouns that are proclitic (or prexed) to the verb from
second position clitic pronouns, whether by copying them there or by reanalysis. This intuition
can be rendered into more formal terms by saying that the verb occupies a functional head such
as T, which by assumption can host an agreement morpheme. This conguration facilitates the
reanalysis of non-stressed clitics that are adjacent to the verb as markers of verbal agreement,
especially in instances of doubling.
. For the sake of concreteness, I suppose that (weak) subject pronouns still move to SpecTP
at this stage. Following ideas by Roberts (1998), I assume that the set of [interpretable] fea-
tures located in T can trigger overt movement of pronominal elements since the latter consist
of nothing but a bundle of formal features (cf. Fu 2003 for a similar analysis of V3 orders with
pronominal subjects in Old English). Further reduction of the pronoun presumably leads to a
reanalysis as X
0
-element which undergoes head movement into the inectional domain (with
the second position effects still resulting from repositioning at MS, see above). In contrast, full
nominal subjects reside in their -position in the languages under consideration, presumably
due to the absence of an EPP feature in T. Accordingly, the double in languages exhibiting stage
3 must remain in a lower position as well. Alternatively, it may undergo A-movement into the
left periphery, which seems more or less in line with the facts observed above.
. In Classical Aztec and Pochutla, SVO or VOS orders are most common; Huichol exhibits
regular OSVorder (presumably derivative froman older SVOstage, Steele 1977: 556); Cora ex-
hibits VSOorder. Steele (1977) gives no informationconcerning the basic word order of Tarahu-
mara, Tepecano and Yaqui. According to the Ethnologue database (http://www.ethnologue.com/)
Tarahumara and Yaqui show basic SOV order, while Southeastern Tepehuan, the closest relative
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.71 (4636-4704)
to Tepecano (Tepecano is by nowextinct), shows both VSO and VOS orders. The SOVcharacter
of Tarahumara and Yaqui seems to pose a problem for the present analysis, since it is obvious
that a second position clitic is not adjacent to the verb in this word order. Interestingly, however,
it seems that these languages showa set of alternative word orders which apparently licensed the
development of new verbal person/number prexes, while the prexes are still absent in SOV
sentences:
(i) semati=ne
nice=clit.1sg
napaha
shirt
rarimea
going.to.buy
ar.
probably
(Tarahumara)
I am probably going to buy a nice shirt.
(ii) tuk=ne
yesterday=clit.1sg
antnyta
Antonio
bicak.
saw
(Yaqui)
Yesterday, I saw Antonio.
Thus, it seems that the distribution of verbal agreement prexes still reects the environments
where their grammaticalization was licensed (absent in SOV orders, present in other orders).
Note that this idea possibly can be used as a diachronic explanation of the fact that the verbal
prexes are sometimes absent in Tarahumara and Yaqui (cf. Note 60 above).
. In VSO/VOS languages, adjacency between the verb and the C-oriented clitic is guaranteed
if the verb moves to T to create V1-order (that is, the subject stays behind in a lower position,
as assumed above). In SVO languages, the clitic moves to the left of the nite verb, while the
double stays behind in a lower position or moves up into the left periphery (see Note 63 above).
. As numerous other languages, Buryat lacks pronouns for 3rd person subjects. Instead,
demonstratives are used to refer to 3rd person entities (Poppe 1960: 53).
. It can be shown that the development of subject agreement sufxes predates the develop-
ment of the possessive sufxes. Apart fromBuryat, the new verbal agreement markers are found
in Oirat, Kalmyk, Dagur, and Moghol. In Dagur and Moghol, the relevant sufxes are very sim-
ilar to the respective nominative pronouns, showing no signs of phonetic reduction (cf. Martin
1961; Comrie 1980).
. See Comrie (1980: 90f.) for a similar conclusion. A similar word order option can still be
found in colloquial Khalkha (Standard Mongolian):
(i) a. bi
I
med-ne.
know-pres
I know.
b. med-ne bi.
I know.
However, in contrast to Classical Mongolian, colloquial Khalkha does not allow doubling in
these contexts (Comrie 1980: 91), cf.
(ii) *bi
I
med-ne
know-pres
bi
I
I know.
. Furthermore, Comrie assumes that the possibility of a reanalysis is sensitive to the ma-
jor afxation pattern in a given language. Thus, in an overwhelmingly sufxing language such
as Buryat, a clitic is more readily reanalyzed as an inectional afx if it follows the verb, in
accordance with the general direction of afxation in the language in question.
JB[v.20020404] Prn:2/08/2005; 15:48 F: LA8105.tex / p.72 (4704-4719)
. To be sure, more data and discussion are needed to evaluate the proposal in (118). For
example, if all SOV orders of Classical Mongolian are taken to follow from this structure, we
should expect that there is no wh-ex-situ, since SpecCP is already occupied by the fronted TP.
Interestingly, this expectation seems to be borne out by the facts, lending further support to an
analysis of SOV order in terms of TP-fronting (examples taken from Poppe 1954: 173):
(i) a. ci
you
qami-a-aca
where-from
irebei?
come-past
From where did you come?
b. ci
you
yaun
what
abumui?
take
What will you take?
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.1 (45-143)
Chapter 6
Morphological blocking and the rise
of agreement
. Introduction
This chapter examines the way the grammaticalization of verbal agreement markers
is shaped by properties of the morphological component of grammar. More speci-
cally, it is claimed that the reanalysis of subject clitics is triggered by a morphological
principle which scans the input for the most specic realization of a given agreement
morpheme. This claim is based on the observation that the change in question does
not replace existing agreement markers in a random fashion. Rather, it can be shown
that the creation of new forms affects only those cells of the paradigm where the ex-
isting verbal agreement morphology is not distinctive. Thus, we can often observe
that clitics turn into obligatorily present markers of verbal agreement only in some
person/number combinations, whereas they continue to be merely optional in other
contexts. The distribution of the obligatory clitics interacts in interesting ways with
properties of the existing verbal agreement morphology. For example, in the North-
ern Italian dialect Piattino (Gerlach 2001, 2002) clitics are obligatory only in contexts
where the existing sufxal agreement morphology fails to signal person and number of
the subject in an unambiguous way (see Section6.5.1 on the special status of 3pl, where
the existing ending seems to be distinctive). This is illustrated in Table 1 (boldface
signals that the clitic is obligatory).
From Table 1 it becomes clear that the presence of subject clitics is merely op-
tional for 1sg, 2sg and 2pl, but obligatory in all other contexts (i.e., 3sg, 1pl, and 3pl).
Hence, it appears that the presence of the clitic is required only in those environments
where the existing verb form is underspecied for agreement features. Thus, it seems
Table 1. Subject agreement and clitics in Piattino.
Clitic + verb Presence of the clitic
1sg (a) guardi optional
2sg (te) guardesc optional
3sg.masc al guarda obligatory
3sg.fem la guarda obligatory
1pl an guarda obligatory
2pl (ve) guard optional
3pl li guarden obligatory
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.2 (143-199)
that the grammaticalization of new means of agreement marking serves to repair a
previously defective agreement paradigm. Below, it is shown that similar phenom-
ena can be observed in various languages, including German varieties, Non-Standard
French, Northern Italian and Rhaeto-Romance dialects. Of course, this chapter can-
not do justice to the vast amount of descriptive literature on the grammaticalization
of inectional morphology. Still, the discussion of selected examples suggests that (1)
represents a valid generalization on the grammaticalization of agreement morphology
across languages.
(1) New verbal agreement morphology arises only for those slots of the agree-
ment paradigm where the existing verbal inection is non-distinctive.
It is argued that (1) follows from innate principles which govern the acquisition of
inectional morphology. More specically, I claim that the reanalysis of pronominal
elements as functional agreement heads is triggered if the change leads to the elim-
ination of syncretism in an agreement paradigm. This is modeled as the outcome
of blocking effects which operate during language acquisition and block the acqui-
sition of a less specied form if a more specied candidate is attested in the Primary
Linguistic Data.
The chapter is organized as follows. Section 6.2 introduces the main theoretical
claim put forward in this chapter, namely that the grammaticalization of inectional
morphology is shaped by a principle (dubbed the Blocking Principle) which oper-
ates during language acquisition and requires new inectional material to be more
specied/distinctive than the relevant inectional formatives already present in the
language. This idea is motivated by taking a closer look at morphological aspects of
the development of new verbal agreement markers in the history of Bavarian in Sec-
tion 6.2.1. In the sections 6.2.2 and 6.2.3, I discuss a set of theoretical issues which
arise in connection with the Blocking Principle (concerning the status of morpholog-
ical doublets and competing changes which give rise to less specied forms), thereby
sharpening and partially reshaping the account in terms of morphological blocking
developed previously. In Section 6.3, it is shown that the Blocking Principle offers a
new explanation for the observation that 1st and 2nd person play a pioneering role
in the grammaticalization of agreement markers across languages (cf. Chapter 1).
Sections 6.46.8 illustrate the workings of the Blocking Principle in the grammati-
calization of agreement markers with examples from Non-Standard French (Section
6.4), Northern Italian dialects (Section 6.5), Rhaeto-Romance (Section 6.6), American
Russian and Walser German (Section 6.7), and Skou, a Papuan language spoken in
New Guinea (Section 6.8). Some of these languages have already been commented on
in Chapter 5, which focused on syntactic aspects of the change in question. Note that
Uto-Aztecan and Mongolian are not included in the discussion, since the Blocking
Principle is trivially fullled in languages which completely lacked verbal agreement
prior to the grammaticalization of relevant markers. Section 6.9 provides a summary
of this chapter.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.3 (199-256)
Chapter 6. Morphological blocking and the rise of agreement
. The Blocking Principle
It is a long-standing idea in morphological theory that inectional paradigms do not
tolerate doublets for a givenparadigmatic slot. In recent approaches, this insight is usu-
ally modeled as a blocking effect which ensures that the availability of a more specied
morphological form blocks the use/insertion of a less specied form, cf. the follow-
ing quote from Sauerland (1996: 4) (for discussion of blocking effects cf. Kiparsky
1973, 1983; Aronoff 1976; Anderson 1986, 1992; Halle & Marantz 1993; Kroch 1994;
Sauerland 1996; Halle 1997, Noyer 1997):
1
If one form possesses all the properties of another form and some more specic
properties, the more specied form must be used wherever possible.
In a realizational model of grammar such as Distributed Morphology, this idea can be
modeled in terms of conditions on the process of Vocabulary Insertion (cf. Chapter 2
above), cf. the italicized passage of Halles (1997) Subset Principle in (2):
(2) The Subset Principle (Halle 1997: 428)
The phonological exponent of a Vocabulary item is inserted into a morpheme
in the terminal string if the item matches all of a subset of the grammatical
features specied in the terminal morpheme. Insertion does not take place if
the Vocabulary item contains features not present in the morpheme. Where
several Vocabulary items meet the conditions for insertion, the item matching the
greatest number of features specied in the terminal morpheme must be chosen.
In other words, if more than one Vocabulary item is compatible with the feature con-
tent of a certain terminal node, the most specied Vocabulary item must be inserted.
This trait of the insertion procedure implies that phonological exponents may be un-
derspecied with respect to the featural content of the morpheme they realize. That
is, for insertion to take place, a phonological exponent need not be specied for all
features present in a terminal node.
This section serves to introduce the central theoretical claim of the present chap-
ter, namely the idea that the rise of new verbal agreement morphology is guided by
morphological blocking effects as well. More specically, it is claimed that morpho-
logical blocking guides the acquisition of Vocabulary items to the effect that a new
verbal agreement marker can only be coined if it is specied for a greater subset of
agreement (i.e., -) features than a competing agreement formative that occupies the
relevant paradigmatic slot in the target grammar. The relevant principle of grammar,
the so-called Blocking Principle, is motivated by inspecting morphological details of
the grammaticalization of new verbal agreement markers in the history of Bavarian
(cf. Chapter 5 above).
In addition, it is argued that these assumptions offer a new (diachronic) explana-
tion of the person/number restrictions on complementizer agreement and pro-drop
observed for Bavarian in Chapter 5.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.4 (256-357)
.. The rise of new verbal agreement endings in Bavarian
Acloser look at morphological aspects of the rise of newagreement sufxes in Bavarian
reveals that the individual diachronic developments share an interesting characteristic:
the grammaticalization of 2pl /-ts/ and 1pl /-ma/ resolved existing homophony in the
verbal agreement paradigm.
The grammaticalization of 2pl /-ts/ (< clit. -(e)s) began in the 13th century (in
Northern and Middle Bavarian, cf. Wiesinger 1989: 72f.). Prior to this development,
the verbal agreement paradigm exhibited homophonous forms for 3sg and 2pl. Thus,
the development of a new distinctive ending for 2pl can be said to have repaired this
defect of the paradigm. The relevant facts are shown in Table 2 (the 3sg, 2pl forms are
marked by boldface):
Table 2. Verbal agreement paradigms (pres. indic.), 13th
century Bavarian.
Old paradigm New paradigm
1sg - -
2sg -st -st
3sg -t -t
1pl -an -an
2pl -t -ts
3pl -ant -ant
Interestingly, it can be shown that a similar situation existed in those varieties that
developeda newagreement ending for 1pl, /-ma/. Due to phonological reduction, nal
/-t/ was lost in the 3pl, leading to homophony of 3pl and 1pl forms in most Bavarian
dialects in the 18th century. In some dialects, this form of syncretism was resolved by
the development of 1pl /-ma/ (from the relevant subject clitic) as a new agreement
ending (initially conned to C, see Section 5.2.1):
Table 3. Verbal agreement paradigms (pres. indic.), late 18th
century Bavarian.
1sg - -
2sg -st -st
3sg -t -t
1pl -an -ma
2pl -ts -ts
3pl -an -an
We can thus conclude that both changes apparently repaired a previously defective
agreement paradigm.
2
Interestingly, it seems that this observation is a general char-
acteristic of the grammaticalization of agreement markers across languages. Thus,
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.5 (357-418)
new forms do not develop randomly, affecting any arbitrary slot of the paradigm.
Instead, the development of new inectional formatives is conned to those slots
of the paradigm where the existing verbal inection is not distinctive (e.g., due to
phonological erosion). This has already been noted in the introduction to this chapter
with respect to the development of new prexal agreement markers in the Northern
Italian dialect Piattino (Gerlach 2002). Similar observations can be made in various
other Italian and Rhaeto-Romance dialects (cf. e.g. Spiess 1956; Kuen 1957; Renzi
& Vanelli 1983; Linder 1987; Haiman & Beninc 1992; Poletto 1997), Non-Standard
French (Wartburg 1970; Ashby 1977; Harris 1978; Auger 1993; Gerlach 2002), the
Walser German dialects discussed in Chapter 5 above, and American Russian (Polinsky
2000). Further examples from somewhat more exotic areas include Khinalug (North-
East Caucasian, Corbett 1991: 123), the Austronesian language Kisar (Blood 1992;
Siewierska 1999, 2004), the Yuman language Maricopa (Siewierska 1999, 2004), and
the Papuan language Skou (Donohue 2002, see Section 6.8).
3
Lets therefore assume
that (3) represents a valid descriptive generalization on the grammaticalization of new
verbal agreement morphology:
(3) New verbal agreement morphology arises only for those slots of the agree-
ment paradigm where the existing verbal inection is non-distinctive.
In other words, the reanalysis of clitics as agreement markers is licensed if the change
leads to the elimination of syncretisms in an agreement paradigm. Of course, this re-
quirement is trivially fullled if new agreement formatives develop in a language that
previously showed no agreement morphology at all, as for example in the Uto-Aztecan
and Mongolian languages discussed in Chapter 5 above. At the same time, (3) rules
out the possibility that the grammaticalization of new agreement formatives leads to
the replacement of inectional markers which are fully distinctive. However, note that
(3) is merely a description, but no explanation. Therefore, it would be desirable if (3)
could be shown to follow from deeper principles of grammar. In this chapter, it is
argued that the generalization in (3) is the outcome of blocking effects that operate
during language acquisition and block the acquisition of a less specied form if a more
specied form is attested in the Primary Linguistic Data. In a Late Insertion model
such as Distributed Morphology, this general idea can be formalized as in (4):
4
(4) Blocking Principle
If several appropriate PF-realizations of a given morpheme are attested in the
Primary Linguistic Data, the form matching the greatest subset of the mor-
phosyntactic features included in the morpheme must be chosen for storage
in the lexicon.
The Blocking Principle is to be understood as a cognitive economy principle which
applies during language acquisition and guarantees an optimal and non-redundant
lexicon (and paradigm) structure. Note that this kind of morphological (or lexical)
economy differs signicantly from syntactic/structural economy principles that are
part of the computational system and shape the course of syntactic derivations (cf.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.6 (418-457)
e.g. Chomsky 1995; Collins 1997; Kitahara 1997). The latter exhibit usually a least
effort character, in the sense that they favor less complex, that is, less marked syn-
tactic derivations/representations (cf. Clark & Roberts 1993; Roberts 1993b; Roberts
& Roussou 2003 on the role of least effort principles in language change). In con-
trast, the Blocking Principle prefers more specic, that is, more marked lexical entries
over less marked ones. Still, this can be seen as a form of economy as well, as noted
above.
5
Similar to structural economy principles, the Blocking Principle is called into
service only if the cues provided by the input data are for some reason ambiguous
and not sufcient for the acquisition of certain properties of the grammar. Thus, in
cases where the continued phonological erosion of a clitic leads to a situation where
the trigger experience contains more than one potential exponent for a given agree-
ment head/morpheme,
6
the BP is invoked to decide which of the candidates is stored
in the lexicon. The ultimate decision is then taken by comparing the feature speci-
cations of the competing potential phonological exponents, favoring the candidate
which realizes the greatest subset of the feature content of the underlying agreement
head/morpheme. If prior to the reanalysis, the language lacked an agreement marker
for the paradigmatic slot in question, the BP is trivially fullled, of course.
7
In this
way, the BP ensures that the development of new inectional formatives can affect
only weak/underspecied slots of the paradigm, replacing Vocabulary items that are
not distinctive.
Coming back to the historical developments observed in Bavarian, it can be shown
that the new agreement sufxes 2pl /-ts/ and 1pl /-ma/ satisfy the Blocking Principle
due to the fact that they are more specied than their respective predecessors. Turn-
ing rst to the change that took place in 13th century Bavarian, a closer look at Table 2
shows that the formative /-t/ occurs in 3sg and 2pl contexts, that is, the relevant Vocab-
ulary item must be underspecied for [person] as well as [number].
8
In other words,
/-t/ represents the elsewhere case and is inserted as the default agreement ending:
(5) elsewhere /-t/
Hence, the introduction of 2pl /-ts/ was licensed by the Blocking Principle, since the
new form is specied for [person] and [number], resolving the existing homophony
between 3sg and 2pl:
9
(6) [2, pl] /-ts/
Similar facts can be observed in 18th century Bavarian. There are only two different
plural forms in the agreement paradigm: /-ts/ is inserted in the context [2, pl], whereas
/-an/ realizes the feature combinations [1, pl] and [3, pl]. Thus, the lexical entry for
/-an/ must be underspecied for the feature [person]. In other words, /-an/ is simply
the elsewhere case among the plural forms, cf. the following insertion rules:
(7) a. [2, pl] /-ts/
b. [pl] /-an/
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.7 (457-536)
Again, the potential new realization of 1pl (/-ma/) is more specied than the existing
Vocabulary item /-an/, since it is in addition specied for [person]. This state of affairs
facilitates the grammaticalization process in question, leading to a fully distinctive set
of plural agreement markers:
10
(8) a. [1, pl] /-ma/
b. [2, pl] /-ts/
c. [pl] /-an/
It now becomes clear that the person/number restrictions on complementizer agree-
ment (and pro-drop) which we observed in Chapter 5 above reect the workings of the
Blocking Principle in the grammaticalization of new verbal agreement morphology in
the history of Bavarian. As noted above, the Blocking Principle licensed the develop-
ment of new agreement formatives only for those slots of the agreement paradigm
where the existing phonological exponents were less specied than their newly emerg-
ing competitors, that is, in 2nd person and 1pl contexts. Accordingly, the reexes of
this historical process (which leads to the presence of a dissociated Agr-morpheme on
C, see Chapter 5 above) complementizer agreement and pro-drop are found only
in these particular slots of the paradigm in the present-day language.
So far, I have shown that an account in terms of the Blocking Principle provides
a new explanation for the question of why new agreement sufxes developed only
for 2pl and 1pl forms in the history of Bavarian (still reected by the restrictions on
complementizer agreement and pro-drop in the present-day language). In the next
section it is argued that this approach can also handle the earlier grammaticalization
process which created the 2sg sufx /-st/ during the Old High German period.
... The diachrony of 2sg /-st/
As already noted above in Chapter 5, it is commonly assumed that the form /-st/ re-
sulted from a reanalysis of -s+thu sequences, in which the onset of the pronoun (or the
enclitic -t(u)) was mistakenly analyzed by the learner as part of the verbal agreement
ending (in analogy with the 2sg of the preterite-presents). As noted in Chapter 5, this
process could only take place in inversion contexts, where the nite verb preceded the
subject clitic. However, it seems that the development of 2sg /-st/ presents a problem
for an account in terms of the BP. Consider the forms listed in Table 4:
11
Table 4. Agreement paradigms (pres. indic.) for nmen
take, early OHG.
1sg nim-u nim-u
2sg nim-is nim-ist
3sg nim-it nim-it
1pl nm-ems (-m, -n) nm-ems (-m, -n)
2pl nm-t nm-t
3pl nm-ant nm-ant
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.8 (536-642)
A look at Table 4 suggests that the change from 2sg /-s/ to /-st/ apparently did not
involve the creation of a Vocabulary item that is more specic than its predecessor.
Both items seem to realize the same set of morphosyntactic features, cf.
(9) a. [2, sg, pres.] /-s/
b. [2, sg, pres.] /-st/
Thus, it appears that the creation of the new ending /-st/ conicts with the BP, since
it apparently does not lead to a more specied form. In what follows, however, I will
argue that the change in question proceeded in accordance with the BP, since it helped
to create an inectional formative for 2sg that was additionally specied for verbal
mood, in contrast to its predecessor. Thus, the feature specications of the relevant
Vocabulary items are actually those in (10).
(10) a. [2, sg, pres.] /-s/
b. [2, sg, pres., indic.] /-st/
In early OHG, the 2sg endings of many verbs were identical in the present indica-
tive and the present subjunctive, while verbal mood was clearly distinguished in other
person/number combinations (apart from 2pl). This is illustrated in Table 5 and Ta-
ble 6 for the weak verbs salbn anoint (conjugation class 2) and the very frequent
habn have (conjugation class 3), which exhibit the characteristic inections of their
respective verb classes.
12
Table 5. Conjugation of salbn anoint (class 2, present
tense), early OHG.
Present indicative Present subjunctive
1sg salbm salbo
2sg salbs salbs
3sg salbt salbo
1pl salbms salbm
2pl salbt salbt
3pl salbnt salbn
Table 6. Conjugation of habn have (class 3, present tense),
early OHG.
Present indicative Present subjunctive
1sg habm habe
2sg habs habs
3sg habt habe
1pl habms habm
2pl habt habt
3pl habnt habn
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.9 (642-686)
A closer look at the Tables 5 and 6 reveals that identical verb forms show up in the
slots for 2sg present indicative and 2sg present subjunctive. We can therefore conclude
that the forms for 2sg are underspecied with respect to verbal mood. Interestingly, the
development of the new formative /-st/ began in the present indicative (cf. Brinkmann
1931). This suggests that the development in question was licensed by the fact that the
new ending was unambiguously specied for verbal mood (i.e., indicative), in contrast
to the earlier formative /-s/. Thus, despite appearances, the original change leading to
2sg /-st/ does not represent a counterexample to the BP. Rather, it proceeded in accor-
dance with the requirement that new inectional formatives realize a greater subset
of morphosyntactic features than their predecessors. In a later development, the new
ending spread via analogical extension to all verb classes, tenses and verbal moods
including the present subjunctive. This subsequent development blurred the original
motivation for the change in question and eliminated the mood distinction in the 2sg.
It is important to note, however, that the original grammar change leading to the cre-
ation of 2sg /-st/ can be shown to comply with the Blocking Principle if a larger portion
of the inectional paradigm (including the distinction of verbal mood) is taken into
account, thereby providing further support for the central claim of this chapter.
To sum up, this section has introduced the central theoretical claim of this chap-
ter, namely that the grammaticalization of agreement markers is guided by blocking
effects which shape the acquisition of inectional morphology. The Blocking Principle
has been motivated by looking at the history of Bavarian, arguing that an approach
in terms of morphological blocking offers an explanation for the observation that
the development of new agreement formatives affected only underspecied slots of
the agreement paradigm. In addition, it has been suggested that the peculiar per-
son/number restrictions on complementizer agreement and pro-drop that can be
observed in present-day Bavarian should receive a diachronic explanation, in the sense
that they reect the workings of the Blocking Principle in the grammaticalization
process under consideration. So far, I have focused on the basic mechanisms of the
Blocking Principle, glossing over several theoretical and empirical issues which arise
in connection with this approach to language acquisition and change. Some of these
are discussed in some more detail in the following two sections.
.. Morphological blocking versus analogical leveling
In this section, I discuss the interaction of grammaticalization processes governed by
the Blocking Principle with other forces in language change which have opposite effects
and lead to the loss of distinctions and the creation of less marked, more regular forms.
It is well-known that in many languages, the historical development of verbal
agreement paradigms (and inectional morphology in general) is characterized by
the loss of distinctive morphology (the Germanic languages, in particular English and
Mainland Scandinavian are often-studied examples).
13
It is immediately clear that this
observation appears to constitute a serious problemfor the assumption that the acqui-
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.10 (686-755)
sition of inectional morphology is guided by something like the Blocking Principle,
repeated here for convenience as (11).
in the lexicon.
If a principle such as (11) is assumed to be operative during the course of language
acquisition, we should expect that the loss of inectional distinctions is considerably
hindered and that the learner should always go for the most specied form attested
in the input. Still, we can observe that in most languages, the effects of phonological
erosion and analogical leveling win out over the processes that create new inectional
morphology. While phonological erosion presumably does not constitute a real issue
in connection to the Blocking Principle (when phonological processes lead to the ero-
sion of inectional morphology, the relevant forms simply disappear from the input
data and fail to be acquired), the phenomenon of analogical leveling appears to be
more problematic, since it leads to the replacement of more specied, irregular forms
by less specied and more regular forms that are absent in the target grammar (cf.
e.g. Bybee 1985; Hock 1991; Albright 2002 for discussion). In other words, the learner
does not acquire a non-distinctive form due to phonological erosion. Rather, he/she
chooses to expand the domain of a more regular inection, substituting less distinctive
forms for more distinctive forms. A typical example of analogical change with respect
to agreement marking is the loss of stem vowel alternations in the development of the
English past tense forms, as shown in Table 7 for the verb to drink.
In Old English, agreement distinctions were not only marked by different agree-
ment sufxes, but also by alternations of the verbs stem vowel (in the past tense).
However, in the historical development of the relevant forms, the vowel changes were
lost and drank surfaced as the only past tense form of the verb drink with the stem
vowel // generalized to all person/number combinations. This represents a typical
example of analogical change, leading to a set of forms that are more regular and
show less distinctions than the inventory found in the target grammar (note that this
Table 7. Analogical leveling of agreement distinctions in the historical development of to
drink.
Old English past tense forms Expected Modern English Actual Modern English
past tense forms past tense forms
1sg dranc drank drank
2sg drunce drunk drank
3sg dranc drank drank
1pl druncon drunk drank
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.11 (755-838)
Table 8. Analogical leveling of agreement distinctions in Yiddish.
Present tense of grfbn to dig Present tense of visn to know
1sg grfb veys
2sg grebst grfbst veyst
3sg grebt grfbt veys(t)
1pl grfbn visn veysn
2pl grfbt vist veyst
3pl grfbn visn veysn
change cannot be attributed to phonological erosion, since the latter affected only
non-stressed nal syllables, but crucially not the verb stem in the history of English).
Similar changes affected the verbs stem vowel in the history of Yiddish present
tense forms, where the 2nd and 3rd singular forms of to dig and the plural forms of
to know have been changed to match the relevant verb stem found in the 1sg, cf. Table
8 (see Albright 2002 for detailed discussion).
14
From Table 8 it becomes clear that the process of analogical leveling led to the
loss of agreement distinctions in the paradigm. In the case of grfbn to dig, the change
conated the forms for 3sg and 2pl, which were formerly distinct due to the existence
of stem vowel alternation. Similarly, analogical leveling led to homophonous forms
for 2sg and 2pl of visn to know. These changes raise a problem for an approach in
terms of the Blocking Principle: the elimination of alternations resulting from analog-
ical leveling is not expected if it is assumed that the acquisition procedure prefers more
specied over less specied forms.
In order to address this problem, some more has to be said on the nature and
possible analysis of analogical leveling. Traditionally, it is often assumed that the drift
towards more uniformity within a paradigm is a natural development towards a situa-
tion where, ideally, a single form corresponds to a single meaning/function (in the case
at hand, a single non-alternating form of the verb stem). In other words, it is assumed
that learners/speakers tend to accept a change eliminating stem alternations in a given
paradigm.
15
Still, this kind of explanation runs counter the idea that the acquisition
of inectional paradigms is governed by a principle that singles out the most specied
candidate (robustly) attested in the input.
In an interesting new approach, Albright (2002) analyzes analogical leveling as
an effect of the way learners acquire the base form of the verb in a given paradigm.
Albright argues that learners scan the input for a base form of the verb which allows to
generate unknown forms as accurately as possible:
The way that they do this [...] is by seeking a base form within the paradigm that
is maximally informative that is, that suffers the least serious phonological
and morphological neutralizations and then deriving the remaining forms in
the paradigm from the base form. (p. 7)
Albright uses the term neutralization to refer to the disappearance of (morpho-)
phonemic contrasts in a certain stem form due to the afxation of inectional ma-
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.12 (838-916)
Table 9. Neutralization due to voicing assimilation in Yid-
dish (present tense).
lib6n to love zip6n to sift
1sg lib zip
2sg lipst zipst
3sg lipt zipt
1pl lib6n zip6n
2pl lipt zipt
3pl lib6n zip6n
terial. He illustrates the effects of neutralization with an example from Yiddish, where
similar to German or English the afxation of a sufx consisting of voiceless ob-
struents requires that a root-nal obstruent become voiceless as well, cf. Table 9. This
phonological process leads to the neutralization of morphophonemic contrasts in the
2sg, 3sg and 2pl, where it is no longer clear whether the base form ends in a voiced (as
in lib6n) or voiceless (as in zip6n) obstruent.
Albright then goes on to show that of all base forms of the present tense paradigm,
the 1sg has a unique status in Yiddish, since it is much less affected by phonological
neutralization than all other base forms. For this reason, the 1sg form is more infor-
mative than any other base form as far as phonemic distinctions are concerned and can
therefore be considered the maximally informative base form. For example, in the case
of lib6n to love, only the 1sg base form would serve to unambiguously differentiate the
verb lib6n from another possible (but non-existing) verb *lip6n ending in a voiceless
obstruent.
How does this relate to the phenomenon of analogical leveling? Albright claims
that the learner has to select a single surface form as the base form of the verb when
he acquires a given inectional paradigm (the so-called single surface base hypothesis).
By assumption, this form is always the maximally informative candidate attested in
the input (in the case at hand the 1sg form). In other words, children acquiring an
inectional paradigm seek out the base form which manifests the distinctive phono-
logical properties of the verb in question (number of phonemes, voicing etc.) in the
most unambiguous and informative way. After the learner has detected the most suit-
able candidate, he/she goes on to derive the remainder of the paradigm from the base
form. Of course, the learner may be later compelled to store separate forms with stem
alternations if they are robustly attested in the relevant input data. However, if for
some reason he/she fails to pay attention to the relevant data, he/she will go on to use
the over-regularized forms, which will eventually replace the old irregular forms if the
change gains a wider distribution in the speaker community.
16
In Yiddish, this kind
of analogical leveling affected the present tense paradigm, in which the 1sg developed
into the only base form for all slots in the paradigm (see Table 8 above).
Interestingly, under these assumptions, it appears that even analogical leveling can
be attributed to the workings of morphological blocking, in the sense that the learner
selects the stem alternant which is most informative/specied with respect to phone-
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.13 (916-985)
mic distinctions as the base form of a given lexical root. This intuition is phrased in
somewhat more formal terms in (12).
(12) Blocking Principle phonemic distinctions
If several realizations of a given lexical root are attested in the Primary Lin-
guistic Data, the form manifesting unambiguously the greatest number of
phonemic distinctions is chosen for storage in the lexicon.
The fact that the effects of analogical leveling seem to win out over the effects of the
Blocking Principle (as dened above in (4) for the acquisition of inectional dis-
tinctions) can then be explained if it is assumed that the acquisition of phonemic
distinctions (and base forms) is more fundamental than the acquisition of inectional
morphology and therefore has priority over the latter, thereby bleeding the effects of
the inectional Blocking Principle in the historical development of languages.
A related but somewhat different scenario arises in cases where paradigm level-
ing does not affect stem allomorphy, but rather inectional afxes, in the sense that a
certain afx gains a wider distribution in a paradigm. As a rule, this kind of change
involves the expansion of underspecied, elsewhere formatives to other slots of the
paradigm. This can be illustrated with a change that affected the plural forms in the
verbal agreement paradigm of many Alemannic dialects. The majority of these dialects
spoken in Switzerland and Southwest Germany exhibit only a single plural ending for
all persons, -ed (so-called Einheitsplural). Interestingly, it can be shown that this end-
ing was originally conned to 3pl (-end prior to elision of /n/) and spread later to the
other persons, affecting rst 2pl forms (cf. e.g. Paul 1952: 194; Weber 1987: 173; Paul
1998: 240). This change is illustrated in Table 10.
In other words, the plural agreement morpheme originally conned to 3rd per-
son evolves into the general plural morpheme, replacing more specied morphemes
that formerly signaled person distinctions. Following Noyer (1997: lxxxlxxxi), I sug-
gest that this kind of change is to be analyzed as the extension (or development) of
Impoverishment rules which delete a subset of the morphosyntactic features of a given
inectional head in a certain context (cf. Chapter 2 above).
17
Recall that Impoverish-
ment generally leads to the insertion of a less marked exponent, due to the fact that
the impoverished inectional head is no longer compatible with the feature matrix
of a more specied Vocabulary item, leading to an expansion of the domain of the
unmarked exponent (Halle 1997: 431). In the change at hand, the relevant Impover-
ishment rule would lead to the deletion of person features in plural contexts, thereby
Table 10. The development of Einheitsplural in Alemannic (pres. indic.)
Original paradigm Intermediate Stage New paradigm
1pl -en -en -ed
2pl -ed -end -ed
3pl -end -end -ed
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.14 (985-1047)
forcing the insertion of the default plural ending -ed (which is specied only for the
feature [+pl]):
18
(13) [person] /[+pl]
Here, it is important to note that Impoverishment and morphological blocking op-
erate in different parts of the grammar and give rise to different kinds of historical
effects: while Impoverishment rules are part of MS and affect the output of the syntac-
tic derivation, that is, the feature content of terminal nodes, the Blocking Principle
governs the acquisition of Vocabulary items realizing inectional heads (and, if a
principle such as (12) is on the right track, the acquisition of lexical roots as well),
favoring more specied over less specied Vocabulary items. In this way, the effects
of the (inectional) Blocking Principle can be thwarted by the expansion of Impov-
erishment rules: after deletion of morphosyntactic features in the relevant underlying
inectional heads, forms which realize a greater set of features simply cannot be in-
serted (or created) and give way to less specied, unmarked inectional formatives. It
is therefore perhaps not accidental that in most languages, the effects of paradigm lev-
eling and phonological erosion win out over the processes that create new inectional
morphology.
19
In the next section, I will deal with another problem in relation to morphological
blocking, namely the observation that the grammaticalization processes discussed so
far do not lead to an instant replacement of the old forms by the newforms. Rather, the
change in question introduces a more specied inectional formative which replaces
the older, less specied form rst in a certain context before it gradually gains a wider
distribution, apparently leading to the existence of morphological doublets.
.. On the status of morphological doublets
In Chapter 5 above, I have noted that in the history of Bavarian the development of new
agreement sufxes proceeded via an intermediate stage where the new form appears
on verbs that occupy C (in inversion contexts), while the original (less specied) agree-
ment sufx is still found on verbs in clause-nal position (i.e., in embedded clauses).
This is illustrated for the rise of 1pl /-ma/ (replacing the less specied form /-a(n)/) by
the following pair of examples from Lower Bavarian:
20
(14) Mia
we
gem-ma/*geng-an
go-1pl/go-1pl
hoam.
home
We are going home.
(15) wa-ma
because-1pl
hoam
home
laff-a/*laff-ma.
go-1pl
Because we are going home
At rst sight, one would perhaps expect the Blocking Principle (and the Subset Princi-
ple) to rule out this kind of variation. First, if the insertion of phonological exponents
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.15 (1047-1110)
is governedby the Subset Principle, the availability of a more specied form(1pl /-ma/)
should always block the use of a less specied competing form (1pl /-a(n)/). Second, if
the Blocking Principle is assumed to operate during language acquisition, one would
perhaps expect that the presence of a more specied form in the Primary Linguistic
Data prevents the acquisition of the less specied form in the rst place.
However, the fact that most modern theories of morphology assume some version
of morphological blocking does not mean that languages never show morphological
doublets. On the contrary, competing forms of a certain word or grammatical forma-
tive can be shown to be quite common in the worlds languages. Well-known examples
include co-existing (new) regular and (old) irregular past tense and participle forms
in German (taken from Bittner 2004), cf.
(16) a. backte (regular) vs. buk (irreg.) I/he/she/it baked
b. erbleichte (regular) vs. erblich (irreg.) I/he/she/it paled
c. glimmte (regular) vs. glomm (irreg.) I/he/she/it glowed
(17) a. gegrt (regular) vs. gegoren (irreg.) fermented (participle)
b. geschnaubt (regular) vs. geschnoben (irreg.) snorted (part.)
c. gebleicht (regular) vs. geblichen (irreg.) bleached (part.)
Note that in contrast to the competition between the forms in (14) and (15), which
arose via the grammaticalization of a new, more specic agreement ending for 1pl, the
variation illustrated in (16) and (17) is again the result of analogical leveling which
creates more regular (i.e., in a certain sense less specic) alternatives to the existing
inventory of forms. After some period of coexistence, the regular forms usually drive
the old forms out of the grammar. The latter type of change leading to morphological
doublets is discussed at length in Kroch (1994), who argues that the development of
doublets is the result of sociolinguistic factors. Krochs argument is based on work by
Taylor (1994), who discusses the occurrence of doublets in the history of English past
tense morphology. Taylor argues that the rise of morphological doublets in the Middle
English (ME) period resulted from language contact of northern ME dialects with
Scandinavian and subsequent dialect mixture with southern dialects.
21
Interestingly,
the life span of such doublets is apparently rather limited: there are no cases where a
doublet that was already present in ME continues to exist in the present-day language
(apart from cases where the two forms have developed different meanings). This can
be taken to indicate that doublets can come into existence due to sociolinguistic factors
such as language contact. However, in the course of time, blocking effects ensure that
one form wins out over the other:
Doublets arise through dialect and language contact and compete in usage until
one or the other formwins out. Due to their sociolinguistic origins, the two forms
often appear in different registers, styles, or social dialects: but they can only coex-
ist stably in the speech community if they differentiate in meaning, thereby ceasing
to be doublets. Speakers learn either one or the other form in the course of basic
language acquisition, but not both. Later in life, on exposure to a wider range of
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.16 (1110-1161)
language, they may hear and come to recognize the competing form, which for
them has the status of a foreign element. They may borrow this foreign form into
their own speech and writing for its sociolinguistic value or even just because it is
frequent in their language environment. Over time, however, as dialects and reg-
isters level out through prolonged contact, the doublets tend to disappear.
(Kroch 1994: 6)
Does this explanation carry over to the kind of morphological doublets created by the
rise of new agreement formatives (cf. e.g. the Bavarian data in (14) and (15) above)?
For the following two reasons, this is not very likely. First, it is fairly clear that lan-
guage/dialect contact usually does not play an important role in the reanalysis of clitics
as agreement markers (however, see Siewierska 2004: 274ff. for a set of examples where
person markers are apparently borrowed from other languages). Rather, the change in
question is most probably a language-internal process which leads to the rise of new,
more specied inectional afxes. In addition, we have seen that at least in Bavarian,
the old and new forms do not vary freely. Instead, they are distributed complementar-
ily, with the newforms initially restricted to inversion contexts, while the old forms are
conned to (embedded) OV clauses.
22
Thus, we have to look for an alternative expla-
nation of the existence and development of morphological doublets which come into
existence as the result of grammaticalization processes similar to the Bavarian case at
hand. Note that this explanation must satisfy two apparently conicting demands. On
the one hand, in order to not violate the Blocking Principle or the Subset Principle,
the connection between the competing doublets must be loose enough to allow the
coexistence of different agreement formatives for a given slot of the paradigm. On the
other hand, the connection must be close enough in order to warrant that the gram-
maticalization process affects only underspecied slots of the paradigm, in line with
the BP. The basic idea I want to pursue in the following makes use of the assumption
that the BP operates in a local fashion, that is, it ensures replacement of a less specied
form only in a certain insertion context (or, with respect to a certain morpheme), but
not in a wholesale fashion. In (18), the relevant passage is marked by italics.
in the lexicon.
Coming back to the Bavarian data under consideration, it is important to note that
according to the analysis proposed in Chapter 5 above, the new and the old agree-
ment ending are different phonological realizations of two different instances of one
and the same underlying agreement head/morpheme. While the old ending (2sg /-s/,
2pl /-t/, 1pl /-a(n)/) is conned to the syntactic Agr-on-T head, the new ending (2sg
/-st/, 2pl /-ts/, 1pl /-ma/) is inserted as the realization of a dissociated agreement mor-
pheme which attaches to the C-head as a copy of the syntactically licensed agreement
morpheme adjoined to T:
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.17 (1161-1211)
(19)
C
T Agr
{/-st/, /-ts/, /-ma/}
TP C
(subj.) C
P T
Agr
{/-s/, /-t/, /-a(n)/}
T
CP
Recall that due to the SOV+V2 syntax of Bavarian, new agreement sufxes could only
develop in V2 contexts via an intermediate stage where the new agreement marker
was introduced as the realization of a dissociated Agr-morpheme attached to C. Thus,
initially the new ending replaced the old form only in a certain context (inversion),
gradually gaining a wider distribution.
23
However, the reanalysis did not only intro-
duce a new phonological exponent of verbal agreement. In addition, it created another
agreement head/morpheme which is post-syntactically adjoined to C.
24
When the
reanalysis in question became possible (e.g., due to phonological erosion of the pro-
noun), it is only in this insertion context (i.e., Agr-on-C) that the BP ensured that a
former pronoun (or, rather, a sufx consisting of the old ending and the onset of the
clitic) was stored in the lexicon as the exponent of an Agr-morpheme, leading to the
obsolescence of the old, less specied verbal agreement endings in inversion contexts.
In other words, initially the BP forced the replacement of the less specied form only
in a certain insertion context (Agr-on-C), while verbs in clause-nal position contin-
ued to exhibit the old agreement ending, triggering the acquisition of two different
exponents for a certain slot of the verbal agreement paradigm.
25
Hence, the BP (and
similarly the Subset Principle) does not prohibit the existence of different agreement
formatives for a given slot of the paradigm as long as these formatives are realizations
of different instances of an Agr-morpheme which occur in different syntactic contexts.
What the BP does rule out, however, is the possibility that the learner acquires compet-
ing phonological realizations for one and the same Agr-morpheme. Interestingly, this
prediction seems to be borne out by the empirical facts. To the best of my knowledge,
there are no instances where complementizer agreement (i.e., Agr-on-C) is realized by
the old agreement formatives, that is, forms such as *ob-s whether-2sg *ob-t whether-
2pl or *ob-an whether-1pl are not attested in any stage or variety of Bavarian. This
can be taken as further support for the analysis proposed above according to which the
BP required the use of the new, more specied exponents when this insertion context
was created via the reanalysis of former clitic pronouns.
26
In other words, the BP (and
the Subset Principle) allows for the existence of morphological doublets if the relevant
agreement formatives are realizations of different agreement morphemes with identi-
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.18 (1211-1266)
cal feature content (or, rather, different instances of a single agreement morpheme), in
the case at hand Agr-on-C (inserted post-syntactically) and Agr-on-T (present in the
syntax). Still, this account is restrictive enough to guarantee that the new exponents
develop only for previously underspecied slots in the verbal agreement paradigm.
This is achieved by the assumption that the BP forces the acquisition of the new, more
specied exponents when the latter compete with the old, less specied formatives for
the realization of the newly created agreement morpheme on C.
.. Section summary
In this section, I have established the claim that the grammaticalization of agree-
ment markers is shaped by morphological blocking effects. The central theoretical
device, the so-called Blocking Principle is understood as an economy principle which
shapes the acquisition of inectional morphology, warranting an optimal and non-
redundant Lexicon (and paradigm) structure. It was demonstrated that this approach
offers an explanation for the observation that the grammaticalization of new agree-
ment markers in the history of Bavarian affected only underspecied slots of the
agreement paradigm. In addition, I proposed a diachronic explanation of the peculiar
person/number restrictions on complementizer agreement and pro-drop exhibited by
present-day Bavarian (cf. Chapter 5 above), arguing that they reect the workings of
the Blocking Principle, in the sense that only in these contexts (2nd person, 1pl) the
Blocking Principle licensed the reanalysis of clitic pronouns as agreement morphemes
attached to C (Agr-on-C). Subsequently, I discussed the status of morphological dou-
blets, which are often a by-product of this kind of change, and the phenomenon of
paradigm leveling, which expands the domain of less marked, more regular forms.
It was argued that analogical leveling affecting stem allomorphy can also be at-
tributed to morphological blocking if it is assumed that the learner prefers to acquire
the stem alternant which manifests the characteristic phonemic distinctions of a given
lexical root in the most unambiguous way. In addition, I considered cases where an
underspecied agreement exponent gains a wider distribution in a paradigm, arguing
that this kind of paradigm leveling is to be attributed to the expansion of the domain
of Impoverishment rules which bleed the effects of the Blocking Principle.
Furthermore, I argued that the Blocking Principle allows for the existence of
morphological doublets only if the relevant agreement formatives are realizations of
different agreement morphemes with identical feature content (in the case at hand
Agr-on-C and Agr-on-T). In other words, the Blocking Principle operates in a local
fashion, restricted to a certain insertion context. In this way, it is possible to main-
tain the hypothesis that the Blocking Principle forces the acquisition of more specied
exponents: in the insertion context where the new forms came to compete with the
less specied old forms (Agr-on-C), only the former, more specied exponents were
acquired, leading rst to a complementary distribution of old and new agreement
sufxes and eventually to the loss of the old forms.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.19 (1266-1329)
In the next section, it is shown that the theoretical hypotheses developed in the
present section receive further support from a typological tendency already noted in
the introductory Chapter 1, namely the observation that 1st and 2nd person play a
pioneering role in the development of new agreement formatives across languages.
. The pioneering role of 1st and 2nd person
In this section I want to add another piece of evidence from language typology which
can be taken to indicate that something like the Blocking Principle is in fact on the
right track. More specically, it is argued that the pioneering role of 1st and 2nd person
forms in the rise of agreement paradigms across the worlds languages (cf. Chapter 1
above) can directly be attributed to the workings of blocking effects during language
acquisition if we adopt certain assumptions about the feature inventory that underlies
person distinctions.
As noted in Chapter 1 above, we can observe that across the worlds languages,
verbal agreement markers are much more common for 1st and 2nd person subjects
than for 3rd person subjects (cf. Bybee 1985; Ariel 2000; Cysouw 2003a). For exam-
ple, Bybee (1985) shows that 54% of the languages (in her sample) which manifest
agreement do not mark 3rd person on the verb (cf. Chapter 1 for examples). Similar
person restrictions can also be observed in languages that are currently developing new
agreement markers from clitic pronouns such as Non-Standard French, where new
agreement formatives are found for 1st/2nd person, but not in 3rd person contexts
(see Section 6.4 below).
27
The special role of 1st and 2nd person in grammaticaliza-
tion processes has inspired numerous functionalist explanations which mostly make
use of the fact that speaker and hearer are the most salient participants in a speech
event (cf. Mithun 1991; Ariel 2000), that is, 1st and 2nd person forms exhibit a high
degree of givenness, discourse accessibility or discourse prominence etc. Since the
reference of pronominal forms for 1st and 2nd persons can be easily deduced from the
immediate discourse situation, so the argument goes, they are more readily reduced to
unstressable clitics, afxes, and eventually zero than 3rd person forms, which denote a
more remote agent.
In this section it is argued that these facts can be explained in purely formal terms
if we combine the Blocking Principle with certain assumptions concerning the featural
basis of person distinctions in natural languages. These assumptions are laid out in the
following, which requires a brief digression on the structure of person features before
we can turnto explaining the typological tendency in questionin terms of the workings
of the Blocking Principle.
The basic assumption I want to adopt is that the feature system behind person
distinctions is not organized around a tripartite division ([1st], [2nd], [3rd]) as tra-
ditionally assumed. Instead, I take up the idea that [3rd person] actually does not
constitute a separate person feature at all. Instead, 3rd person is analyzed as the result
of the absence of (positive values for) the features 1st and 2nd person (cf. Benveniste
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.20 (1329-1419)
Table 11. Person features according to Halle (1997).
Feature combination Interpretation
[+Auth, +PSE] 1st person inclusive
[+Auth, PSE] 1st person exclusive
[Auth, +PSE] 2nd person
[Auth, PSE] 3rd person
1972; Bayer 1984; Halle 1997; Grimshaw 1997; Noyer 1997; Poletto 1999; Ariel 2000;
Harley & Ritter 2002; Baker & Bobaljik 2002; Cysouw 2003a; Frampton 2002, among
others). In a model of person specication that does not include a separate feature
for 3rd person, the traditional three-way system is replaced by a binary feature sys-
tem which only refers to the immediate participants in a speech event, for example
[speaker] and [hearer]. Within this approach, the interpretation 3rd person re-
sults from the absence of positive values for the features [speaker] and [hearer]. The
set of person distinctions listed in Table 11 makes use of the feature inventory intro-
duced in Halle (1997), namely Participant in Speech Event (PSE) and Author in Speech
Event (Auth).
This system gives rise to a four-way person distinction for non-singular forms.
In addition to the traditional notions 1st person, 2nd person and 3rd person, it also
serves to distinguish between 1st person inclusive (non-singular 1st person forms that
include the addressee) and 1st person exlclusive (non-singular 1st person forms that
exclude the addressee).
28
In languages which lack the distinction between inclusive and exclusive we (as
many Indo-European languages such as English, German, Italian etc.), the feature
[PSE] (or, [hearer]) is not relevant for the phonological realization of 1st person
plural pronouns. Likewise, the feature [Auth] does not matter for the realization
of 2pl. As a consequence, the system of plural pronouns is most economically de-
scribed by assuming Vocabulary items which are underspecied for the relevant person
features, cf. the paradigm of plural nominative pronouns in English:
29
(20) a. [+Auth, +pl] we
b. [+PSE] you
c. [+pl] they
With respect to 3pl, the logic of underspecication requires that 3pl be analyzed as
the elsewhere case in the plural paradigm: due to the absence of positive values for
person features (i.e., [Auth], [PSE]), 3rd person is most economically captured
by the absence of any person specication in the relevant Vocabulary item. The same
goes for 3sg forms, which lack person specications, but are specied for gender in
English (if singular is treated as the unmarked default value, they lack number features
as well). In other words, the 3rd person forms represent the maximally underspecied
(elsewhere) cases in the paradigm. This of course carries over to 3rd person forms
across languages, including agreement formatives.
30
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.21 (1419-1458)
Let us now turn to the question of how this binary system of person specications
helps us to explain the pioneering role of 1st and 2nd person in the historical develop-
ment of verbal agreement marking. Here, the important assumption is that 3rd person
forms are generally underspecied for person features, that is, the relevant Vocabulary
items lack specications for [Auth] and [PSE]. The fact that cross-linguistically,
3rd person agreement formatives arise later (if at all) than markers for 1st and 2nd
person can then be attributed to the workings of the Blocking Principle in the follow-
ing way: if UG requires that newly grammaticalized markers realize a greater subset
of agreement features than existing markers, the development of 3rd person forms
is considerably hindered, due to the inherent underspecication of 3rd person forms
with respect to the set of person features. As a consequence, even in a grammar that
previously lacked agreement markers completely, a true 3rd person marker can only
develop if it is specied for some other inectional feature like [gender], [number]
etc. We can therefore conclude that the workings of the Blocking Principle ensure that
the grammaticalization of new 3rd person forms is less likely to be triggered than the
development of forms specied for a separate [person] feature, that is, 1st and 2nd
person markers. This explains the typological tendency in question, lending further
support to the proposal that the acquisition of inectional morphology is guided by
blocking effects that favor more specied over less specied forms.
31
In the remainder of this chapter, the workings of the Blocking Principle are il-
lustrated with examples from a variety of different languages, beginning with Non-
Standard French, where the grammaticalization of new agreement formatives is also
characterized by the dichotomy between 1st/2nd person forms and 3rd person forms
discussed in the present section.
. French
In the history of French, we can observe a cluster of changes involving pronouns, ver-
bal agreement, and the pro-drop property which is cyclic in nature (Guiraud 1968;
Wartburg 1970; Ashby 1977; Harris 1978; Roberts 1993a; Vance 1997):
(21) Distinctive verbal Agr/pro-drop >>> loss of Agr/loss of pro-drop (overt sub-
ject pronouns become obligatory) >>> subject pronouns become clitics >>>
clitics are reanalyzed as verbal Agr/rise of pro-drop
Thus, subject pronouns were merely optional in Old French. Overt pronouns were
used either for reasons of emphasis (similar to the present-day Romance pro-drop lan-
guages like Italian) or inserted to avoid V1 orders, presumably due to the V2 character
of Old French (cf. e.g. Kuen 1957; Haiman 1991; Siewierska 1999).
32
It is generally as-
sumed that in later stages, the subject pronouns gradually lost their emphatic value.
In the 14th century then, the phonological erosion of the inherited verbal agreement
sufxes reached a point where most endings had lost their distinctive character. In
turn, overt subject pronouns became more and more obligatory, in line with the stan-
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.22 (1458-1518)
dard hypothesis that rich verbal agreement is a precondition for pro-drop. The fact
that the original personal pronouns could no longer be used to convey emphasis led
to the use of object pronouns in emphatic contexts (which are doubled by the weak
subject pronouns), as in the following topic left dislocation example from Modern
Standard French (cf. Guiraud 1968; Wartburg 1970; Ashby 1977 on the connection
between the rise of this construction and the weakening/grammaticalization of the
subject pronouns):
(22) Moi,
me
je
I
sors.
leave
As for me, I leave.
In other words, even in the traditional descriptive literature it is often assumed that
the subject pronouns of present-day French undergo a development from referential
elements to some form of agreement marker:
It is, then, possible to see in the history of the minimum nite verb group that a
semantic restructuring has taken place, whereby the once referentially important
subject pronoun has become a morpheme whose only importance is now struc-
tural. In this sense, the subject pronoun has become a mere satellite of the verb.
(Ashby 1977: 55)
This development is even further advanced in present-day Non-Standard French,
where subject clitics are obligatory (see below for details) and exhibit characteristics
typical of prexal inectional morphology (Ashby 1977; Harris 1978; Lambrecht 1981;
Auger 1993, 1994; Gerlach 2002). Accordingly, Ashby (1977: 59) predicts that Modern
Non-Standard French is currently developing into a pro-drop language (with prexal
person/number markers), completing the cycle depicted in (21).
Several arguments have been put forward in the literature on Non-Standard
French that examples such as (22) should not be analyzed as instances of (clitic) left
dislocation, in contrast to present-day Standard French, where je is usually treated as a
resumptive clitic. In what follows, I will review only a selection of the arguments sug-
gesting that the subject clitics in Non-Standard French constitute some formof prexal
agreement marking on the verb (cf. Lambrecht 1981; Auger 1993, 1994; Haiman 1991).
First, the relevant subject clitics are obligatory and cannot be replaced by the full
form (historically an oblique form, see above):
Non-Standard French
(23) a. (Moi)
me
je
clit.1sg
porte
carry
la
the
table.
table
I carry the table.
b. Moi
me
*(je)
clit.1sg
porte
carry
la
the
table.
table
I carry the table. (Gerlach 2002: 224)
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.23 (1518-1612)
The obligatory presence of these clitic elements is also reected by their peculiar be-
havior in conjoined clauses. In most languages, subject pronouns are usually elided in
second conjuncts if they are coreferent with the subject of the rst conjunct. How-
ever, no such ellipsis takes place in Non-Standard French (cf. Lambrecht 1981: 24;
Haiman 1991):
Standard French
(24) Il
he
mange
eats
et
and
boit
drinks
comme
like
un
a
cochon.
pig
Non-Standard French
(25) I
he
mange
eats
et
and
i
he
boit
drinks
comme
like
un
a
cochon.
pig
Again, this behavior points toward an analysis of the preverbal pronouns as bound
agreement markers which are obligatory regardless of the syntactic context.
Second, sentences with apparent clitic doubling generally favor a basic, non-
dislocated interpretation. Thus, sentences such as (23a) receive an interpretation
like (26a) instead of (26b). This indicates that they do not constitute instances of
topic left dislocation, again suggesting an analysis of the apparent clitics as prexal
agreement markers.
(26) a. I carry the table.
b. *As for me, I carry the table.
Third, in contrast to the pronominal clitics of Standard French, cf. (27), the pre-
verbal clitics of colloquial French fail to undergo subject-verb inversion in matrix
interrogatives, as can be seen in the example (28) (taken from Lambrecht 1981: 6):
Standard French
(27) O
where
vas=tu?
go=you
Non-Standard French
(28) O
where
tu-vas?
you-go
In other words, the obligatory preverbal clitics of Non-Standard French seem to ac-
company verb movement to C, in contrast to the subject clitics of the standard lan-
guage. Again, this can be taken to indicate that they should be analyzed in terms of
prexal agreement markers. To sum up, it seems that Non-Standard French exempli-
es another instance of a grammaticalization process in which clitic pronouns develop
into some form of verbal inection.
Interestingly, the distribution of the newly created agreement markers of Non-
Standard French exhibits some parallels to the distribution of complementizer agree-
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.24 (1612-1663)
Table 12. Distribution of subject clitics in
Non-Standard French.
Presence of the clitic
1sg obligatory
2sg obligatory
3sg optional
1pl obligat. (on)/option. (nous)
2pl optional
3pl optional
ment in the Bavarian varieties discussed above. Table 12 shows that the clitic is obliga-
tory in some person/number combinations while it is merely optional in others.
The contrast between 1st person and 2sg clitics, which are always obligatorily
present, and 3rd person and 2pl forms, which can be replaced by full DPs, is illustrated
by the following examples (taken from Gerlach 2002: 224):
(29) a. (Moi)
me
je
1sg
porte
carry
la
the
table.
table
I carry the table.
b. Moi
me
*(je)
1sg
porte
carry
la
the
table.
table
I carry the table.
(30) a. (Jules)
Jules
il
clit.3sg
porte
carry
la
the
table.
table
Jules/he carries the table.
b. Jules
Jules
(il)
clit.3sg
porte
carry
la
the
table.
table
Jules carries the table.
(31) a. (Alexandra et moi)
Alexandra and I
on
1pl
porte
carry
la
the
table.
table
Alexandra and I carry the table.
b. Alexandra et moi
Alexandra and I
*(on)
1pl
porte
carry
la
the
table.
table
(32) a. (Alexandra et moi)
Alexandra and I
nous
clit.1pl
portons
carry
la
the
table.
table
We/Alexandra and I carry the table.
b. Alexandra et moi
Alexandra and I
(nous)
clit.1pl
portons
carry
la
the
table.
table
Thus, it seems that the grammaticalization of new agreement markers has affected
only a subset of the relevant clitic forms. A further indication that 3sg forms have
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.25 (1663-1746)
Table 13. Subject agreement in written and spoken French.
Written language Phonetic form
1sg porte [pfrt]
2sg portes [pfrt]
3sg porte [pfrt]
1pl (on) porte [pfrt]
(nous) portons [pfrtf]
2pl portez [pfrte:]
3pl portent [pfrt]
retained their pronominal character comes from the fact that clitic doubling with 3rd
person forms still exhibits properties of topic left dislocation such as incompatibility
with (negative) quantiers (Friedemann 1995; Laenzlinger 1998: 339):
(33) *Personne
nobody
il
he
rigole
laughs
jamais.
never
Nobody never laughs.
So the question arises of why the reanalysis in question has affected only a subset of the
subject clitics. Interestingly, Gerlach (2002: 225f.) observes that the distribution of the
obligatory subject clitics correlates with properties of the verbal agreement paradigm.
More specically, she argues that clitics are obligatory only when the verbal inection
is underspecied for subject agreement features. According to the written standard
language, French apparently distinguishes a number of different agreement endings
on the verb. However, the spoken language actually makes less distinctions than the
written forms may suggest. This is shown in Table 13, where the phonetic forms of the
present indicative of the verb porter to carry are compared with the relevant forms of
the written language.
In Table 13, we can see that the written language distinguishes ve different agree-
ment endings; only the forms for 1sg and 3sg (and the agreement endings for the
colloquial 1pl form on) are homophonous. In contrast, the spoken language distin-
guishes merely three forms, -(1sg, 2sg, 3sg, 1pl (on), 3pl), -f (1pl, nous) and -e: (2pl).
One of these forms, the 1pl -f, shows up only in connection with the pronoun nous,
which is considered rather formal and is rarely used in the spoken language. Thus, we
actually deal with a highly syncretic agreement paradigm, in which only the 2pl end-
ing signals unambiguously person and number of the subject. In all other contexts, we
nd the ending -, which is underspecied for person and number and accordingly
represents the elsewhere case:
(34) a. [Auth, +PSE, +pl] /-e:/
b. elsewhere -
If we combine our observations concerning the distribution of the obligatory clitics
with the ndings concerning the existing sufxal agreement paradigm in the spoken
language, we arrive at the following generalization:
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.26 (1746-1828)
(35) Correlation between the distribution of subject clitics and properties of sufxal
agreement
In Non-Standard French, subject clitics are obligatory only
a. in non-3rd person contexts, and
b. if they serve to express -features not marked by the existing sufxal
agreement morphology.
This is illustrated in Table 14 where agreement-like elements which reect the -
features of the subject are in boldface (obligatory clitics and agreement sufxes) while
optional clitics (which are presumably still pronominal elements) are set in paren-
theses.
In the following, it is argued that the distribution of obligatory subject clitics in
Non-Standard French can be attributed to the workings of the Blocking Principle if it is
assumed that 3rd person forms are underspecied for person features (see Section 6.3
above). Similar to Bavarian (cf. Section 6.2.1 and Note 31), new agreement formatives
developed only in 1st and 2nd person contexts, where the new agreement formatives
are specied for a greater subset of the relevant -features than the existing verbal
agreement morphology (the elsewhere form -):
(36) a. [+Auth, pl] /6/ (1sg)
b. [+PSE, pl] /ty / (2sg)
c. [+Auth, +pl] /f/ (1pl)
Hence, it is quite obvious that the grammaticalization of the new prexal forms is li-
censed by the Blocking Principle. In contrast, the potential new 3rd person agreement
prexes il/elle and ils/elles carry no person specication. That is, at least with respect
to person features they fail to be stronger specied than the existing forms. However,
it is fairly clear that the pronouns carry gender and (probably) number specications
which at rst sight should qualify them as more specic candidates if compared with
the relevant existing agreement sufxes. Still, the 3rd person clitics are merely op-
tional in Non-Standard French, which suggests that they have not yet developed into
agreement prexes. This fact is at rst sight not expected, given the way the Blocking
Principle is formulated above.
Table 14. Sufxal agreement and clitics in
Clitic + verb
1sg je porte
2sg tu portes
3sg (il/elle) porte
1pl on porte
2pl (vous) portez
3pl (ils/elles) portent
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.27 (1828-1887)
First, I will address the question of whether gender specication alone is suf-
cient to trigger the grammaticalization of new verbal agreement markers. Note that
Greenbergs (1966a) Universal 32 states that verbal agreement for gender entails agree-
ment in number,
33
that is, it seems that verbal agreement in gender becomes available
only if a language has developed a full paradigm of number distinctions. Moreover,
it appears that at least on verbs, gender agreement is a highly marked grammatical
trait that is found only in a couple of languages. This generalization on the distribu-
tion of morphosyntactic features can be explained if it is assumed that -features are
not unordered sets, but rather are organized hierarchally in a feature geometry where
number features dominate gender features (cf. e.g. Noyer 1997; Harley & Ritter 2002),
as illustrated by the following tree structure:
34
(37) Agr-morpheme/pronoun
cvxuvv PSE
xuxnvv vvvsox
sc Auth
Animate Inanim./Neuter
vi
It is generally assumed that in a feature structure such as (37), the individual nodes
(and the features/nodes they dominate) are present only if the grammar in question
exhibits the relevant morphological distinctions. On these assumptions, the observa-
tion that the possibility of gender distinctions seems to depend on the existence of
number distinctions (Greenbergs Universal 32) can be accounted for in the following
way: since the Gender node is a dependent of the Number node, gender agreement is
only possible if the Number node is present as well, but not vice versa.
Furthermore, if the feature geometry in (37) is taken to represent a universal
property of grammar we should expect that it has some bearing on the historical devel-
opment of inectional morphology as well. More specically, the grammaticalization
of morphosyntactic feature distinctions is predicted to proceed in a top-down fash-
ion, in the sense that a language can develop verbal agreement in gender only if it
has previously grammaticalized a set of number distinctions. Moreover, I assume that
this restriction operates over paradigms: languages do not develop gender agreement
in a certain person/number combination (e.g., 3sg.fem) when there are still forms in
the paradigm that fail to show number distinctions. In other words, the development
of verbal gender agreement is bound to the existence of a full paradigm of number
specications for all persons.
35
These assumptions suggest the following explanation for the fact that the gender
specication of the 3rd person pronouns is not sufcient for triggering a reanaly-
sis as agreement markers in Non-Standard French: it is fairly clear that the existing
agreement paradigm does not distinguish number for all persons. Accordingly, gender
distinctions are presumably not visible to the workings of the Blocking Principle and
can therefore not motivate the grammaticalization of new agreement markers.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.28 (1887-1937)
Still, it appears that il/elle and ils/elles are apparently specied for number (via
plural -s), which should nonetheless sufce to mark them as more specied than the
existing elsewhere agreement sufx -. As is well known, however, the number mark-
ing of the plural forms is only perceivable if the verb following the pronoun begins
with a vowel. In all other contexts, nal -s is not pronounced and the plural forms
are homophonous with the relevant 3sg forms. Hence, the number marking of the 3rd
person forms is actually less salient in the PLD than it appears at rst sight. Tentatively,
I suggest that the restricted distribution of unambiguous number marking (plural is
perceivable only in front of verbs beginning with a vowel; in all other contexts the
3sg/pl forms are identical) fails to be robust enough to activate the Blocking Princi-
ple (cf. Note 7 above on the inuence of frequency on the workings of the Blocking
Principle). Note, however, that the existence of clearly different 3sg and 3pl clitics is
expected to trigger the grammaticalization of new agreement formatives if all other
necessary conditions are met. A relevant example comes from the Northern Italian
dialect Piattino (Lombardy) which is discussed in the next section.
. Northern Italian dialects
Many Italian dialects exhibit similar developments as Non-Standard French. In partic-
ular, it has often been observed that Northern Italian dialects exhibit obligatory subject
clitics which may be doubled by an additional full subject. The peculiar properties
of these clitics (including e.g. obligatory presence in conjoined clauses and relative
clauses, obligatory adjacency to the verb, compatibility with a quantied full subject
etc.) have led many scholars to the conclusion that these elements developed into some
form of verbal inection in the history of these dialects. Even more interesting in the
present context is the observation that the change in question correlates with prop-
erties of the existing verbal agreement paradigm (cf. e.g. Meyer-Lbke 1894; Rohlfs
1949; Spiess 1956; Kuen 1957; Renzi & Vanelli 1983; Poletto 1995, 1997, 1999, 2000). It
appears that the subject clitics are obligatory only in those instances where the existing
verbal agreement morphology fails to unambiguously reect the subjects -features.
If it is assumed that the obligatory clitics have in fact developed into (prexal) verbal
agreement markers, their distribution can be explained by assuming that the newly
grammaticalized agreement formatives served to compensate for the loss of distinctive
agreement endings, in line with the Blocking Principle.
Of course, this section cannot serve as a comprehensive overview of the phe-
nomenon in question in the multitude of Northern Italian dialects. Instead, I chose
to focus on a pair of dialects which represent two of the major dialectal areas of North-
ern Italy: Piattino, a variant of Lombardo spoken in and around the village of Piatta
in the Italian Alps and the dialect of Vicenza (Vicentino) which belongs to the dialect
group usually referred to as Veneto. The Piattino data are taken from Gerlach (2001,
2002) while the section on Vicentino reports on my own research on this dialect.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.29 (1937-2009)
.. Piattino
First, let us take a look at Piattino, a Lombardian dialect discussed in Gerlach (2001,
2002). According to Gerlach (2002: 225f.), the Piattino subject clitics are obligatory
whenever the verbal forms lack subject person agreement features. From this, it ap-
pears that Piattino behaves similar to Non-Standard French, exhibiting clitics which
take up the function of agreement markers in contexts where the existing verbal mor-
phology is not distinctive anymore. However, at a closer look it becomes clear that the
Piattino facts differ in important ways from the situation observed in Non-Standard
French. Consider Table 15, where obligatory clitics and distinctive verbal agreement
morphology are marked by boldface (relevant example sentences are shown in (38)
(40)).
36
(38) a. Mi
I
(a)
clit.1sg
guardi.
watch-1sg
I am watching.
b. (Mi)
I
a
clit.1sg
guardi.
watch-1sg
I am watching. (Gerlach 2002: 224)
(39) a. (Nigun/un omen/lomen/Alessio/lu)
nobody/a man/the man/Alessio/he
al
clit.3sg.masc
guarda.
watch.3sg
Nobody/a man/the man/Alessio/he is watching.
b. Nigun/un omen/lomen/Alessio/lu
nobody/a man/the man/Alessio/he
*(al)
clit.3sg.masc
guarda.
watch.3sg
Nobody/a man/the man/Alessio/he is watching. (Gerlach 2002: 223)
(40) a. (Noaltri)
we
an
clit.1pl
guarda.
watch.1pl
We are watching.
b. Noaltri
we
*(an)
clit.1pl
guarda.
watch.1pl
We are watching. (Gerlach 2002: 223)
Table 15. Subject agreement and clitics in Piattino, present indicative.
1sg (a) guardi optional
2sg (te) guardesc optional
3sg.masc al guarda obligatory
3sg.fem la guarda obligatory
1pl an guarda obligatory
2pl (ve) guard optional
3pl.masc i guarden obligatory
3pl.fem li guarden obligatory
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.30 (2009-2072)
Table 15 and the examples in (38)(40) illustrate that the presence of subject clitics is
merely optional in 1sg and 2nd person contexts, but obligatory in all other contexts
(i.e., 3sg, 1pl and 3pl), giving rise to clitic doubling if a full pronoun/subject is added
for reasons of emphasis. The obligatory presence of the 3rd person and 1pl clitics and
the fact that the 3rd person clitics can be doubled by quantied expressions like nigun
nobody suggest that the relevant clitics have changed into agreement markers.
As already mentioned above, the obligatory presence of the subject clitic seems
to correlate with properties of the sufxal agreement morphology. That is, the clitic
is obligatory in the context of 1pl and 3sg, where the existing agreement endings are
homophonous.
37
This is also the conclusion reached by Gerlach, who attributes the
obligatory presence of the clitics to the lack of (distinctive) person agreement features
on the verb.
38
In other words, subject clitics became obligatory in contexts where the
nite verb is underspecied for agreement features, presumably to recomplete a de-
fective agreement paradigm. In contrast, the clitic is merely optional in 1sg and 2nd
person contexts, where the verbal agreement morphology is still fully distinctive, re-
ecting unambiguously the subjects set of -features. This sensitivity to properties
of the sufxal agreement morphology can be taken as a further indication that the
clitics are in fact better analyzed as some form of agreement marking. Thus, it seems
that the behavior of subject clitics in Piattino resembles the situation observed in
However, note that there are also important differences between Piattino and Non-
Standard French. First, there is a context (3pl) where a clitic is obligatory although
the sufxal agreement morphology is not homophonous with any other agreement
ending (and thus seems to be distinctive). Second, in contrast to Non-Standard French,
Piattino exhibits obligatory clitics for all verb forms that are homophonous (in Non-
Standard French, all singular forms and 1pl, 3pl are identical, but the clitic is obligatory
only in non-3rd person contexts), cf. the following quote from Gerlach (2002: 225):
A particular feature of the Piattino subject paradigm is that the 3rd person may
not remain unspecied. Consequently, the clitic is obligatory for 3rd person and
1st person plural.
That is, while the optional presence of clitics in 1sg and 2sg contexts is in line with
the Blocking Principle (the reanalysis of clitics as agreement markers is not called for
as long as the existing agreement morphology is distinctive), the behavior of the 3rd
person forms is unexpected and demands some further consideration, in particular if
it assumed that 3rd person forms are inherently underspecied with respect to person
features (cf. Section 6.3 above). Let us therefore take a closer look at the agreement
distinctions resulting from the development of obligatory clitics in Piattino.
The change affecting the 1pl is rather straightforward from the viewpoint of
the Blocking Principle: while the existing agreement ending consists merely of the
theme vowel and represents the elsewhere case in the present indicative paradigm,
the new agreement formative is clearly specied for both person and number (i.e.,
[+Auth, +pl]).
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.31 (2072-2140)
Table 16. Development of gender agreement in Piattino.
New agreement formatives Old agreement formatives
3sg.masc al- -a
3sg.fem la- -a
3pl.masc i- -en
3pl.fem li- -en
What about the development of obligatory 3rd person clitics for both singular
and plural? Table 16 compares the new 3rd person agreement formatives with the
existing sufxal agreement morphology. Fromthis, it becomes apparent that the gram-
maticalization process led to the development of gender agreement in the context of
3rd person.
Due to their specication for gender, the new 3rd person agreement formatives
count as stronger agreement exponents, even if they do not carry any person speci-
cation. Of course, this raises the question of why gender information is accessible in
Piattino, while it has been argued above that gender distinctions are invisible to the
grammaticalization of agreement formatives in Non-Standard French. Tentatively, I
suggest that the access to gender specications is made available by the existence of
a compared to French much more elaborate inventory of number distinctions
in the existing agreement paradigm of Piattino. Recall that in Section 6.4 above, I
have proposed a feature geometry for agreement distinctions where gender features
are dominated by the node encoding number information. By assumption, the robust
morphological representation of number in Piattino paved the way for the develop-
ment of gender distinctions on the verb by making the gender node accessible for
grammaticalization processes.
Summing up, it appears that after a closer look, it is possible to maintain the
idea that the grammaticalization of new verbal agreement morphology is governed by
morphological blocking effects which require new agreement formatives to be more
distinctive than existing forms. More specically, it has been argued that the obliga-
tory presence of subject clitics in Piattino can be attributed to the fact that the emerging
prexal agreement formatives realize distinctions such as person/number (1sg) or gen-
der agreement (3sg/3pl) which are not marked by the respective sufxal agreement
morphology. In the next section, it is shown that similar correlations can be detected
in Vicentino, another Northern Italian dialect which shows an even richer inventory
of existing sufxal agreement morphology.
.. Vicentino
This section reports on my research on Vicentino, a variant of Veneto spoken in and
around Vicenza (northern Italy).
39
Similar to Piattino, the obligatory presence of sub-
ject clitics seems to be linked to properties of the existing verbal agreement paradigm
in Vicentino. However, as we will see shortly, Vicentino differs in interesting ways from
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.32 (2140-2217)
Table 17. Subject agreement and clitics in Vicentino, present indicative of
mangiare to eat.
1sg (a) magn-o optional
2sg te magn-i obligatory
(enclitic: magn-i=to)
3sg el/a magn-a obligatory
(enclitic: magn-e=o/a)
1pl (a) magn-emo optional
2pl (a) magn- proclitic optional
(enclitic: magn-=o) enclitic obligatory
3pl i/e magn-a obligatory
(enclitic: magn-e=i/e)
Piattino. Table 17 illustrates the agreement paradigm and the distribution of subject
clitics in Vicentino. Distinctive person/number marking (either via sufxes or clitics)
is marked by boldface while optional clitics are set in parentheses.
40
Even a rst glance at Table 17 reveals that the correlation between defective agree-
ment endings and the obligatory presence of subject clitics is less straightforward than
in Piattino. The only homophonous endings are 3sg and 3pl /-a/ (which is traditionally
analyzed as a theme vowel). In all other contexts, the agreement morphology seems to
be distinctive. Still, subject clitics are obligatory for all person/number combinations
apart from 1sg/1pl (2pl forms show a peculiar behavior: while the enclitic o is obliga-
tory, the corresponding proclitic form a is merely optional; see below for discussion).
The distribution of the various clitics is illustrated in the following, beginning with the
observation that clitics are merely optional in 1st person contexts:
(41) (A)
clit.1sg
vegno
come-1sg
da
from
Vicensa.
Vicenza
I come from Vicenza.
(42) (A)
clit.1pl
vegnemo
come-1pl
da
from
Vicensa.
Vicenza
We come from Vicenza.
Thus, in examples such as (41) and (42), the verbal morphology apparently sufces to
identify subject pro. As a result, the overt pronoun a is merely optionally present. In all
other contexts (apart from 2pl, see below), the absence of (distinctive) subject clitics
leads to ungrammaticality (note that the shape of the proclitics differs from that of the
enclitics used in inversion contexts):
(43) a. Da nd
where-from
vien=*(to)
come=clit.2sg
ti?
you
Where do you come from?
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.33 (2217-2281)
b. Ti
you
te
clit.2sg
vien
come
da
from
Montecio.
Montecio
You come from Montecio.
c. *Ti vien da Montecio.
As already noted above, 2pl forms exhibit an asymmetry between inversion and non-
inversion forms: while the enclitic o used in inversion contexts is obligatory, the
corresponding proclitic form a (used in all other contexts) can be omitted:
41
(44) a. Da nd
where-from
vegni=*(o)
come=clit.2pl
(voaltri)?
you.pl
Where do you come from?
b. (A)
clit.2pl
vegn
come-2pl
da
from
Durlo.
Durlo
You come from Durlo.
Similar to Piattino, all 3rd person clitics (in both inversion and non-inversion contexts)
are obligatory. Note that the obligatory 3pl clitics serve to signal gender distinctions in
addition to person and number (e unambiguously marks 3pl.fem, while ()i may be
used to refer to either 3pl.masc or general 3pl without gender specication). The (a)
examples illustrate the use of subject enclitics while the (b) examples depict the series
of 3rd person proclitics:
(45) 3sg.masc
a. Da nd
where-from
vien=*(o)
come=clit.3sg.masc
sto
this
omo?
man
Where does this man come from?
b. Elo/u
he
*(el)
clit.3sg.masc
vien
come
da
from
Durlo.
Durlo
He comes from Durlo.
(46) 3sg.fem
a. Da nd
where-from
vien=*(a)
come=clit.3sg.fem
sta
this
dona?
woman
Where does this woman come from?
b. Ela
she
*(a)
clit.3sg.fem
vien
come
da
from
Arzegnan.
Arzignano
She comes from Arzignano.
(47) 3pl.masc
a. Da nd
where-from
vien=*(i)
come=clit.3pl.masc
lori?
they.masc
Where do they come from?
b. Lori
they.masc
*(i)
clit.3pl.masc
vien
come
da
from
Malo.
Malo
They come from Malo.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.34 (2281-2336)
(48) 3pl.fem
a. Da nd
where-from
vien=*(e)
come=clit.3pl.fem
lore?
they.fem
Where do they come from?
b. Lore
they.fem
*(e)
clit.3pl.fem
vien
come
da
from
Ciampo.
Ciampo
They come from Ciampo.
Before turning to the issues raised by these data for the Blocking Principle, let me take
a closer look at the inventory of subject clitics available in Vicentino. In general, Vi-
centino exhibits a remarkably rich inventory of different types of subject clitics. There
are three different series of clitic pronouns: a set of enclitics which attach to the right
of the verb in inversion contexts and two sets of proclitics. The latter subdivide into
a proclitic a, which can be used for various person/number combinations (1sg, 1pl,
2sg, and 2pl), and a set of distinctive forms, which are unambiguously specied for
person and number. Interestingly, the two types of proclitics also differ in their syn-
tactic properties. First, only the distinctive subject clitics are obligatory in Vicentino
while the non-distinctive pronominal clitic a, which is underspecied for both per-
son and number features, is merely optional (see below for examples). Second, the
distinctive clitics follow the negation no, while the non-distinctive forms precede the
negation (similar phenomena can be observed in various other Northern Italian and
Rhaeto-Romance dialects, cf. Meyer-Lbke 1894: 101f.; Rohlfs 1949: 168ff.; Renzi &
Vanelli 1983; Linder 1987; Poletto 1997, 1999, 2000; Vanelli 1997: 109):
42
(49) a. (A)
clit.1sg
no
not
vegno
come-1sg
da
from
Vicensa.
Vicenza
I do not come from Vicensa.
b. (A)
clit.1pl
no
not
vegnemo
come-1pl
da
from
Vicensa.
Vicenza
We do not come from Vicenza.
(50) a. No
not
te
clit.2sg
vien
come
da
from
Vicensa.
Vicenza
You do not come from Vicenza.
b. No
not
a
clit.3sg.fem
vien
come
da
from
Vicensa.
43
Vicenza
She does not come from Vicenza.
The different distribution of the two kinds of clitics becomes particularly clear in 2sg
contexts, where both types of clitics are available. Thus, in examples such as (51), up
to three different types of pronouns may co-occur: a full pronoun, which usually bears
stress (ti), the non-distinctive clitic a (unspecied for person and number), which pre-
cedes the negation, and the obligatory clitic te, which follows the negation and is always
adjacent to the nite verb.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.35 (2336-2439)
(51) Ti
you
a
clit.2sg
no
not
te
clit.2sg
vien
come
da
from
Vicensa.
Vicenza
You do not come from Vicenza.
The different subject clitics available in Vicentino and their diverging properties are
summarized in Table 18 and Table 19.
Similar to Non-Standard French and Piattino, the special properties of the enclitics
and distinctive proclitics suggests that these elements are presumably better analyzed
as agreement marking elements. In contrast, the fact that the non-distinctive clitics are
merely optional can be taken to indicate that these are still pronominal elements. The
observation that the non-distinctive clitics are not affected by the grammaticalization
process in question can be directly attributed to the workings of the Blocking Principle:
due to the underspecication of the clitic a for both person and number features, it fails
to meet the requirement that newagreement exponents must realize a greater subset of
agreement features than existing agreement formatives. As a consequence, the clitic a
cannot be reanalyzed as an agreement marker and maintains its status as a pronominal
element. The 2sg forms provide a nice minimal pair: while the non-distinctive clitic a is
merely optional and occurs to the left of negation, the distinctive form te is obligatorily
present and always adjacent to the verb. In this way, the distributional asymmetries
between distinctive and non-distinctive forms reect the workings of morphological
blocking in the grammaticalization of new agreement formatives in Vicentino: while
the distinctive clitics were reanalyzed as agreement markers and became obligatory,
the morphosyntactic defectiveness of the non-distinctive forms prevented a categorial
reanalysis, preserving the pronominal properties of the elements in question.
45
Table 18. Three series of subject clitics in Vicentino.
Enclitics Proclitics
distinctive non-distinctive
1sg a
2sg to te a
3sg.masc o (e)l
3sg.fem a a
1pl a
2pl o a
3pl.masc i i
3pl.fem e e
Table 19. Diverging properties of subject clitics in Vicentino.
-feature specication Presence Placement relative to verb
and negation
Enclitics person/number obligatory neg-V-clitic
44
Distinctive proclitics person/number obligatory neg-clitic-V
Non-distinctive proclitics optional clitic-neg-V
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.36 (2439-2593)
However, while the behavior of the non-distinctive forms is expected from the
viewpoint of the Blocking Principle, it is the change affecting the distinctive subject
clitics which raises a more serious issue, in particular if a strong version of the Blocking
Principle is adopted. On this assumption, the agreement formatives developing from
former subject clitics are required to realize a greater subset of agreement features than
existing inectional markers. Above, however, we have seen that Vicentino exhibits a
rather rich paradigm of sufxal agreement morphology, where all endings apart from
3rd person (3sg/3pl /-a/) are apparently fully distinctive. So, again the questionarises of
how the grammaticalization process in question can satisfy the constraints imposed by
the Blocking Principle. In the following it is argued that the changes affecting Vicentino
met these requirements via a conspiracy of a set of independent factors.
First, if we take a look at the resulting paradigm of prexal agreement markers,
it becomes apparent that similar to Piattino, the change in question introduces verbal
agreement in gender in 3rd person contexts, a trait formerly absent in the grammar of
Vicentino, cf. the comparison of 3rd person forms in Table 20.
Thus, even if it is assumed that the absence of person features hinders the gram-
maticalization of new agreement formatives in 3rd person contexts (see Section 6.3
above), it is fairly clear that the new formatives qualify as stronger specied elements
by unambiguously marking gender (and number) of the subject.
46
While this answers
the question of why the 3rd person clitics were eligible for being grammaticalized
as agreement markers, this account cannot be extended to the change affecting 2nd
person forms, which does not show gender distinctions.
Concerning the historical development of new 2nd person agreement formatives
in Vicentino, consider the paradigm of the frequent irregular verb vegnere to come
in Table 21.
Table 20. Development of gender agreement in Vicentino.
New agreement formatives Old agreement formatives
3sg.masc el- (inversion: -o) -a
3sg.fem a (inversion: -a) -a
3pl.masc i (inversion: -i) -a
3pl.fem e (inversion: -e) -a
Table 21. Present indicative of vegnere to come (Vicentino).
vegnere to come
1sg vegno
2sg vien
3sg vien
1pl vegnemo
2pl vegn
3pl vien
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.37 (2593-2594)
Table 22. 2sg, 3sg and 3pl of irregular verbs (present indicative) in Vicentino.
dare to give fare to do nare to go savere to know
2sg de fe ve se
3sg da fa va sa
3pl da fa va sa
In Table 21, we can observe that the 2sg form vien is homophonous with the re-
spective 3rd person forms in the present indicative (tegnere to hold behaves similar).
In other words, it seems that at least in the context of irregular verbs such as vegnere
and tegnere, the reanalysis of the 2sg clitics proceeded in accordance with the Blocking
Principle, creating a new agreement formative which is unambiguously specied for
2sg (i.e., [+PSE, pl]). Subsequently, the newly created agreement formative spread to
other verb classes by analogical extension, becoming obligatory in all contexts. How-
ever, we might ask why the presence of distinctive endings in the other verb classes
did not prevent the reanalysis of 2sg clitics. For this reason, it would be desirable if
we could detect another factor which helped to promote the reanalysis of 2sg clitics.
In the following, it is claimed that the change in question was promoted by the fact
that a number of highly frequent irregular verbs signal subject agreement only via
stem vowel alternations (and not via afxal agreement morphology). Consider Ta-
ble 22 which lists the 2sg, 3sg and 3pl present indicative forms of four very frequent
irregular verbs. Note that these forms are characterized by the absence of sufxal agree-
ment morphology and the fact that they consist merely of the (monosyllabic) verb
root, lacking even the theme vowel found with other short verbs. Interestingly, this
verb class (sometimes referred to as Kurzverben short verbs, cf. Nbling 1995; Dal
Negro 2004) apparently plays a pioneering role in the grammaticalization of new ver-
bal agreement morphology (at least in Germanic, cf. the development of 1pl -ma in
Lower Bavarian and the grammaticalization of new agreement formatives in Walser
German discussed in Chapter 5 above). Why? Tentatively, I suggest that this verb class
is affected by the grammaticalization process in question more readily because the
learner scans the input for the most salient realization of a given inectional marker.
As pointed out by Nbling (1995: 148), agreement distinctions via stem vowel alterna-
tion are much more prone to accidental homophony than other means of agreement
marking (i.e., sufxes). In other words, in a situation where the learner has to decide
whether the subject clitic or the stem vowel alternation is the primary exponent of 2sg
verbal agreement, it is quite likely that he/she will go for the clitic if the latter meets all
other necessary requirements for being regarded as a verbal inection. This hypothe-
sis is also supported by research on the acquisition of inectional morphology, where
it has been observed that children generally prefer to use afxes over other morpho-
logical means (such as stem vowel alternations) for marking additional (inectional)
distinctions (cf. Clark 1998: 384). Now, if this results in a scenario where the reanalysis
of 2sg clitics is not only triggered in the context of vegnere and tegnere (where 2sg is
defective), but also with the set of highly frequent verbs such as to go, to do to give
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.38 (2594-2699)
and to know, it is quite probable that the change in question catches on and develops
into a model of agreement marking which spreads to other verb classes via analogy.
Finally, let me address the development of a new agreement formative for 2pl.
At rst sight, it appears that this change cannot be accounted for in ways similar to
the changes affecting 3rd person forms and 2sg. Apparently, the reanalysis of the clitic
neither leads to the development of an exponent which realizes a greater subset of
-features such as person, number or gender (even the 2pl of short verbs such as veg-
nere is fully distinctive with respect to person and number, see Table 21), nor are there
sufxless short verbs where 2pl is marked solely by stem vowel alternations.
However, at a second look it appears that the rise of 2pl is susceptible to an ex-
planation analogous to the account offered for the development of 2sg /-st/ in the
history of Old High German/Bavarian in Section 6.2.1, where it has been argued that
the grammaticalization of the new agreement ending created a form which was ad-
ditionally specied for verbal mood, distinguishing between present indicative and
present subjunctive.
47
Consider the agreement paradigms for the present indicative
and present subjunctive of a selection of four highly frequent irregular verbs in Table
23 and Table 24.
If we compare the 2pl forms (marked by boldface) in Table 23 and Table 24, it
becomes apparent that the 2pl present indicative and 2pl present subjunctive forms
of these irregular verbs are identical in Vicentino. Thus, we face a situation which is
reminiscent of the explanation offered above for the grammaticalization of 2sg /-st/
in Old High German, which by assumption led to the distinction of verbal mood in
the relevant contexts. More or less the same scenario can be conceived for the reanal-
ysis of the 2pl enclitic -o in Vicentino.
48
Thus, I suggest that the reanalysis in question
Table 23. Present indicative of four frequent irregular verbs in Vicentino.
dare to give fare to do poder can vegner to come
1sg do fo posso vegno
2sg de fe poi vien
3sg da fa pole vien
1pl demo femo podemo vegnemo
2pl dazi f/fazi pod vegn
3pl da fa pole vien
Table 24. Present subjunctive of four frequent irregular verbs in Vicentino.
dare to give fare to do poder can vegner to come
1sg daga fassa possa vegna
2sg dai/daghi fassi possi vegni
3sg daga fassa possa vegna
1pl demo fasemo podemo vegnemo
2pl dazi fazi pod vegn
3pl daga fassa possa vegna
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.39 (2699-2778)
rst affected the present indicative of verbs where the existing 2pl present indicative
and present subjunctive forms were identical. The development of the new agreement
ending 2pl -e+o was licensed due to the fact that the resulting form unambiguously re-
alized verbal mood (i.e., [+indicative]), in contrast to its predecessor. By analogy, the
new agreement formative later extended to other verbs and moods, which blurred the
morphological motivation behind the original change (similar to Old High German,
where 2sg /-st/ spread to the subjunctive as well). Note that in present-day Vicentino
the formative -o is still conned to the environment where the reanalysis became
structurally possible, that is, inversion contexts.
49
Finally, I want to address a general theoretical issue arising in connection with the
kind of analysis presented in this section. If the obligatory subject clitics of dialects like
Piattino or Vicentino are analyzed as agreement morphemes, this seems to give rise to
an instance of multiple exponence, since agreement is realized by both the clitic and
the existing sufxal agreement morphology. However, as is well-known, most current
theories of morphology (with the notable exception of Andersons 1992 A-Morphous
Morphology) rule out the possibility that a given inectional feature/morpheme is re-
alized by more than one primary exponent (cf. e.g. Spencer 1991 or Noyer 1997 for
some discussion). If this view is adopted, it follows that the grammaticalization of new
agreement exponents cannot lead to an outcome where a given underlying agreement
morpheme is realized by more than one primary exponent.
50
However, recall that I
have argued above in Chapters 4 and 5 that the grammaticalization of new agreement
exponents often proceeds via a stage where the clitic is reanalyzed as the realization
of a dissociated Agr-morpheme which is parasitic on another Agr-morpheme valued
in the syntax. Thus, I suggest that in dialects such as Vicentino, the subject clitic is
reanalyzed as a dissociated Agr-morpheme which adjoins post-syntactically to a func-
tional head directly above the head that gives rise to sufxal agreement on the verb.
The newly created dissociated Agr-morpheme is then licensed under structural adja-
cency with Agr-on-T (which hosts the verb). The reanalysis in question is illustrated
by the bracketed structures in (52) (with the host of the dissociated Agr-morpheme
simply labeled F).
(52) [
TP
DP
i
[
T
V+T+Agr ] ... [
P
t
i
...
[
FP
F+Agr [
F
[
TP
[
T
V+T+Agr ] ... [
P ...
Note that the outcome of this reanalysis represents merely a supercial instance of
multiple exponence. In fact, the two agreement exponents correspond to two different
underlying agreement morphemes (with identical feature content), so no real problem
arises here. In other words, the development of multiple agreement is not blocked if
each agreement exponent is associated with a separate agreement morpheme, in the
case at hand Agr-on-T (spelled-out by the existing sufxal agreement morphology)
and a dissociated Agr-morpheme attached to a higher functional head.
51
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.40 (2778-2812)
.. Section summary
In this section I have discussed the development of new agreement markers in Piat-
tino and Vicentino, two Northern Italian dialects where the existing verbal agreement
morphology is still quite rich. At rst sight, this kind of inectional richness seems to
pose a problem for the idea that the grammaticalization of inectional morphology
serves to repair a defective agreement paradigm. However, it has been argued that the
individual historical developments in these dialects can be shown to have in fact led
to the creation of stronger specied forms if more inectional distinctions are taken
into consideration. More specically, I demonstrated that the rise of new 3rd person
agreement markers in the singular as well as the plural led to the grammaticalization
of gender agreement, which was previously absent in Piattino and Vicentino. Further-
more, it was claimed that in Vicentino, the grammaticalization process in question was
also sensitive to the realization of verbal mood, arguing that the reanalysis of the 2pl
clitic was licensed by giving rise to an inectional formative which was unambiguously
specied for [+indicative], in contrast to its predecessor. Based on the observation that
the agreement paradigms of a number of very frequent irregular verbs exhibit certain
defects which do not occur with regular verbs, a more complex scenario was developed
for the change affecting 2sg clitics in Vicentino. First, it was shown that in the present
indicative, the 2sg and 3sg/pl forms of verbs like vegnere to come and tegnere to hold
are homophonous. Second, I demonstrated that in another set of frequent irregular
verbs (including to do, to give, to go, to know), agreement marking for 2sg is sig-
naled only via stem vowel alternations. It was then argued that the grammaticalization
of new agreement markers was triggered rst only in these two contexts, leading to
more distinctive (and salient) inectional formatives. In a later development, the rele-
vant changes gradually expanded to other verbs via analogical extension, giving rise to
the system exhibited by present-day Vicentino.
Thus, even in languages where the existing paradigm of agreement marking is
quite rich, it is possible to maintain the idea that the grammaticalization of new agree-
ment markers must conform to the Blocking Principle. Crucially, as already hinted at
in Section 6.2.1.1 above, to prove that a new form is in fact more specied than its pre-
decessor often requires that a larger portion of the existing inectional morphology is
taken into consideration. Here it is important to note that the additional specication
of a given new form does not necessarily involve agreement features such as person,
number or gender. Instead, the relevant additional specication may involve any kind
of verb-related inectional features like verbal mood, tense etc. In other words, the
Blocking Principle requires (i) that the set of inectional features realized by the new
form includes the set of features realized by the old form and (ii) that the cardinality
of the set of features realized by the new form exceeds that realized by the old form at
least by one.
The next section discusses the development of new agreement formatives in Sut-
selvan, a Swiss Rhaeto-Romance dialect where the existing agreement morphology
appears to be even richer than in the Italian dialects discussed in the present section.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.41 (2812-2888)
. Rhaeto-Romance
This section deals with morphological aspects of the changes affecting the Swiss
Rhaeto-Romance (RR) dialects examined in Chapter 5 above. Note that the follow-
ing discussion focuses on the rather recent development of obligatory clitic doubling
in Sutselvan (and, partially, Surmeiran, cf. the previous chapter, Section 5.3.3), leaving
aside previous developments that led to a reanalysis of C-oriented clitics in the Swiss
RR dialects (cf. Section 5.3.2).
Most Rhaeto-Romance dialects are quite conservative with respect to their inec-
tional morphology, that is, they preserve the system of conjugation classes and agree-
ment endings inherited fromearlier stages of Romance quite faithfully (cf. Haiman and
Beninc 1992 for an overview). This is also true of the Swiss RR dialects. Apart from
Surmeiran and Sutselvan, where 2sg and 2pl are (at least partially) homophonous,
the dialects show a fully distinctive agreement paradigm in the present indicative, as
illustrated in Table 25.
At rst sight, the partial homophony of 2sg and 2pl should lead us to expect that
new agreement formatives develop preferably in this context. However, as noted in
Section 5.3.3, the phenomenon of obligatory clitic doubling is restricted to 1st and 3rd
person contexts, due to the fact that Sutselvan lacks subject enclitics for 2nd person
(similar to the other Swiss RR dialects). As a consequence, there is simply no lexical
material available which may feed the grammaticalization of new 2nd person agree-
ment formatives. In the following, I will therefore focus on the question of whether
the reanalysis of 1st and 3rd person enclitics has taken place in accordance with the
Blocking Principle.
First, note that in many instances, enclitics replace the existing agreement mor-
phology in inversion contexts due to a stress rule sometimes referred to as penultimate
stress target (Haiman 1971; Linder 1987; Haiman & Beninc 1992). Recall that in all
Swiss RR dialects, the syllable structure of nite verbs is governed by a phonological
rule which forbids that the main word stress fall on the antepenultimate syllable. In
instances where the inected verb consists of two syllables and receives stress on the
penultimate syllable, attachment of the clitic would lead to a stress pattern which is not
well-formed. This conict is often resolved by eliding the original verbal agreement
ending (which in the example at hand consists only of the theme vowel -a), ensuring
that the antepenultimate syllable does not receive stress (cf. Haiman 1971: 806):
52
Table 25. Verbal agreement (present indicative) in ve Swiss RR dialects.
Puter Vallader Surmeiran Surselvan Sutselvan
1sg - - - -6l -
2sg -ast -ast -as -as -(a)s
3sg -a -a -a, -e -a -a
1pl -ains -ain(a) -(g)n -in, -ein -(g)n
2pl -ais -aivat -es, -as -is, -eis -(e)s, -(a)s
3pl -an -an -an -an -an
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.42 (2888-2984)
(53) ella
she
chant-a
sing-3sg
*chant-a=la
sing-3sg=clit.3sg.fem
chant=la
sing=clit.3sg.fem
As a consequence, clitics regularly replace the original verbal agreement morphology
in inversion contexts in the Swiss RRdialects, cf. the following example fromSutselvan:
(54) A
at
tut
all
puder
can
curr-la
run-3sg.fem
sper
alongside
laua
the-water
giou.
down
She runs down alongside the water as fast as she can. (Sutselvan, Linder
1987: 63)
Similarly, the clitic appears as the only formative marking the subjects -features in a
number of frequent irregular verbs:
(55) Egn
one
da
of
quels
those
lev-i
wanted-1sg
ear
also
jou.
I
I also wanted one of those. (Sutselvan, Linder 1987: 148)
(56) Cunquegl
since
cigl
it
eara
was
november,
November
vev-la
had-3sg.fem
la
the
scola
school
antschiat.
begun
Since it was November, the school had begun. (Sutselvan; Linder 1987: 155)
(57) Natiral
of course
vev-in
had-3pl
las
the
matns
girls
radetg
brought up
sei mailenders.
Milans
Of course, the girls had brought up some Milans [pastries]. (Sutselvan;
Linder 1987: 161)
Similar to other dialects, 3pl forms are somewhat special in this respect (cf. Note 52).
First, there is a 3pl clitic -i which attaches either to the existing ending, in violation of
the penultimate stress rule, or may cause elision of the vowel of the agreement ending:
(58) A
at
tgea
home
fagev-an=i
made-3pl=clit.3pl
quegl
that
betga
not
savens
often
ascheia.
like-this
At home, they didnt do that often in this way. (Sutselvan, Linder 1987: 86)
(59) Mo
but
gid-n=i
help-3pl=clit.3pl
ear
also
a
to
tgei,
you
scha
when
tei
you
s
have
malavetta?
sorrows
But do they also help you when you are troubled? (Sutselvan, Linder
1987: 87)
Only in Sutselvan, however, there is still another clitic 3pl -in which completely
replaces the original verbal agreement ending in inversion contexts, cf. (60) and
(57) above:
(60) Qua
then
bavev-in
drank-3pl
vinars
brandy
e
and
sa
re
divertetan
entertain-3pl
cun
with
gois.
games
Then they drank brandy and entertained themselves with games. (Sutselvan,
Linder 1987: 88)
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.43 (2984-3037)
Thus, from the perspective of the learner, the clitics are often the only realization of
verbal agreement in inversion contexts. This may give rise to a special verbal inec-
tion conned to inversion contexts, which is sometimes referred to as Inversionsformen
inversion forms in the traditional literature on the Swiss Rhaeto-Romance dialects
(see Linder 1987: Ch. 8 and Section 5.3 above).
53
Thus, if the learner realizes that the
relevant verb form results from eliding the original agreement ending due to a phono-
logical rule, he will acquire the clitic as a pronominal element. In contrast, if for some
reason, he/she misses this fact, he/she will possibly acquire a grammar which differs
from the target grammar in that the clitic is reanalyzed as an agreement formative.
54
Still, this raises the question of whether this reanalysis is in fact licensed by the Block-
ing Principle after all, the present indicative agreement endings are still distinctive in
non-inversion contexts. Furthermore, recall that in Bavarian, the development of new
agreement formatives in inversion contexts (via Agr-on-C) was apparently sensitive to
properties of the existing agreement endings (showing up on clause-nal verbs) and
was licensed only in contexts where the existing agreement morphology was not dis-
tinctive. This seems to suggest that the failure to detect any other potential agreement
formatives in inversion contexts apart from the enclitics is probably not sufcient to
trigger the change in question if the original (distinctive) agreement endings are ro-
bustly attested in other contexts. In other words, we have to look for other factors
which helped to promote the reanalysis in question (in line with the Blocking Princi-
ple). In the following, it is argued that there are at least two settings where the potential
new agreement markers realize more distinctions than the existing forms.
First, at least in the case of 3sg, the new forms are also specied for gender, in con-
trast to the existing agreement formatives (which consist merely of the theme vowel).
This is shown in Table 26.
Accordingly, the new 3sg agreement formatives clearly satisfy the requirements
imposed by the Blocking Principle. However, it is quite obvious that no additional
gender specication is involved in the other contexts where obligatory clitic doubling
is observed in Sutselvan, that is, 1sg/1pl and 3pl. Concerning the latter, I suggest that
the relevant grammaticalization process (leading to Agr-on-C) affected rst verb forms
which exhibit an agreement paradigm which is less rich than that found in the present
indicative.
One characteristic of Rhaeto-Romance languages is the presence of two series of
agreement endings, referred to by Haiman and Beninc (1992) as primary and sec-
ondary desinences. The primary agreement endings are immediately adjacent to the
verb root. They signal a rich array of person/number distinctions and at least par-
tially preserve the four conjugation classes inherited from early Romance, leading to
Table 26. Development of gender agreement in Sutselvan.
Old agreement formative New agreement formative
3sg.masc -a -a+l
3sg.fem -a -la
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.44 (3037-3137)
conjugational allomorphy. Furthermore, at least the 1st and 2nd person endings are
stressed and trigger a stress shift from the verb root to the agreement sufx. In con-
trast, the secondary endings do not attach directly to the verb root, but rather to other
inectional sufxes conveying tense or mood information. The secondary endings are
generally unstressed, lack conjugational allomorphy and show a greater number of
syncretisms in the agreement paradigm (i.e., they signal less person/number distinc-
tions than the primary endings).
55
The differences between primary and secondary
endings are summarized in Table 27 (cf. Haiman & Beninc 1992: 77).
In the dialects under consideration, the primary endings are usually found in the
present indicative and the imperative while the secondary endings show up in the im-
perfect indicative, the present subjunctive, the imperfect subjunctive, and the future.
The following discussion focuses on the imperfect indicative and present subjunctive.
Consider Table 28, which lists the different agreement endings found in the present
indicative (primary) and imperfect indicative (secondary) in Sutselvan.
56
In contrast to the present indicative paradigm, where only the 2sg and 2pl forms
are similar, the secondary endings found in the present subjunctive and imperfect
indicative are much less distinctive. As can be seen from Table 28, 1sg is identical
with 3sg, 2sg is identical with 2pl, and 1pl is identical with 3pl. This state of affairs
is most economically represented by assuming that there are in fact only three separate
Vocabulary items, where /-a/ is simply the elsewhere case.
57
(61) a. [+PSE, Auth] /-as/ (2sg, 2pl)
b. [+pl] /-an/ (1pl, 3pl)
c. elsewhere /-a/ (1sg, 3sg)
Recall that due to the penultimate stress rule, the clitics often replace the old agree-
ment formatives in inversion contexts. Only in 3sg.masc contexts, the resulting form is
Table 27. Primary and secondary agreement endings of Rhaeto-Romance.
Primary agreement endings Secondary agreement endings
Adjacency to verb root yes no
Conjugational allomorphy yes no
Moveable stress yes no
Loss of agreement distinctions no yes
Table 28. Agreement formatives in Sutselvan, present and imperfect (indic.).
Present indicative Imperfect indicative & present subjunctive
1sg - -a
2sg -as -as
3sg -a -a
1pl -(g)n -an
2pl -es, -as -as
3pl -an -an
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.45 (3137-3203)
a combination of the old (secondary) agreement ending -a and the clitic -l. When the
clitics are reanalyzed as the realization of an agreement morpheme on C, this gives rise
to a special verbal inection in inversion contexts. What I suggest is that this change af-
fected rst the impoverished secondary endings, which showup in tenses/moods other
than present indicative and imperative. Subsequently, the agreement endings created
by this grammaticalization process may spread to other tenses/moods via analogical
extension (eventually replacing/enlarging the existing exponents of Agr-on-T in all
contexts). If we compare the existing set of secondary agreement endings with the set
of agreement formatives resulting from the reanalysis of enclitic pronouns, it is imme-
diately clear that the new agreement paradigm includes more agreement distinctions
than the old one (marked by boldface), cf. Table 29.
In particular, it becomes apparent that apart from 2nd person forms, which retain
the ending -as (due to the fact that there are no clitics available), there is a distinct
agreement ending for each slot of the paradigm. Let me now take a closer look at
the set of Vocabulary items resulting from the change in question and check whether
each of the resulting Vocabulary items actually realizes a greater subset of agreement
features than its predecessor.
(62) a. [+Auth, pl] /-i/ (1sg)
b. [+Auth, +pl] /-nsa/ (1pl)
c. [pl, +masc] /-al/ (3sg.masc)
d. [pl, +fem] /-la/ (3sg.fem)
e. [+PSE] /-as/ (2sg, 2pl)
f. elsewhere /-in/ (3pl)
Note that if the set of Vocabulary items in (62) is posited (which presumably repre-
sents the most economical analysis of the paradigm in question),
58
it appears that the
change in question actually led to some forms which are less specied than their pre-
decessors: 2sg, 2pl /-as/ was previously specied for [+PSE, Auth], while it is now
analyzed as the realization of [+PSE]. In addition, 3pl /-in/ is now the elsewhere case,
while its predecessor /-an/ was specied for [+pl]. In other words, while a supercial
inspection of the resulting paradigm shows that the change led to distinct forms for
each slot of the paradigm (apart from 2nd person contexts), a more ne-grained mor-
Table 29. Development new agreement formatives in Sutselvan, secondary endings.
Old agreement formative New agreement formative
1sg -a -i
2sg -as -as (no change)
3sg.masc -a -al
3sg.fem -a -la
1pl -an -nsa
2pl -as -as (no change)
3pl -an -in
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.46 (3203-3275)
phological analysis seems to suggest that some changes violated the Blocking Principle.
This curious discrepancy calls for some further discussion.
First, note that the 2nd person forms in fact do not constitute a problem for the
claim that the reanalysis of clitics is constrained by morphological blocking effects,
since there is no such change involved. Phonologically, the 2sg/2pl exponent /-as/
remained the same. However, due to changes affecting other parts of the paradigm
(namely the creation of stronger specied 1st person forms which block the inser-
tion of /-as/ in these contexts, see Note 57 above), the specication [Auth] is not any
longer necessary to warrant that /-as/ is inserted only in 2nd person contexts. Thus, if
it is assumed that the child acquires the most economical paradigmstructure involving
the least number of feature specications compatible with the input (Halle 1997; see
Note 18 above), the redundant feature specication [Auth] should not be part of the
Vocabulary items realizing 2nd person in (62). In this way, changes affecting one part
of an inectional paradigm (namely, the creation of new 1st person agreement forma-
tives in inversion contexts) may have an effect on other members of the paradigm (in
the case at hand, the feature specication of 2nd person forms).
What about the new agreement ending 3pl /-in/? In (62) above, /-in/ has been
analyzed as the elsewhere form, which is completely underspecied for person and
number. Accordingly, /-in/ is apparently less specied than its predecessor /-an/, which
realizes the feature /+pl/. This is somewhat surprising if the Blocking Principle is
adopted. However, note that the new formative /-in/ is defective only in the context
of the newly established agreement paradigm. The apparent problem disappears if we
become clear about the fact that the Blocking Principle is not concerned with complete
paradigms (i.e., the outcome of the acquisition process), but rather with selecting one
out of a set of individual forms which compete for a particular slot in the paradigm. It
does not look at the paradigm as a whole.
59
If the input contains more than one po-
tential candidate for realizing a specic feature set (say, [+past], [+indicative], [+pl]),
the learner must select one of these forms for storage in the Lexicon. By assumption,
he/she will go for the most explicit realization of the relevant set of features. Thus,
in the case at hand, the learner rst compares the immediate competitors (i.e., 3pl
/-in/ vs. /-an/) with each other before he/she considers further information like the
feature specication of other inectional formatives already acquired up to this point.
With respect to the potential realizations of 3pl, this procedure delivers a clear-cut
result: while the formative /-an/ shows up in 3pl as well as 1pl contexts (present sub-
junctive/imperfect indicative), the immediate competitor /-in/ is conned to 3pl and
therefore represents the most explicit realization of this feature combination. Conse-
quently, /-in/ is stored as the realization of 3pl present subjunctive/imperfect indicative
in inversion contexts. The eventual complete underspecication of 3pl /-in/ presum-
ably results from a later operation which compares the feature matrices of all mem-
bers of the paradigm to determine the most economical paradigm/feature structure
(cf. Note 18).
To summarize, the present section has focused on changes affecting the catego-
rial status of enclitics in the Swiss RR dialect Sutselvan, giving rise to a special form
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.47 (3275-3315)
of verbal inection in inversion contexts (which has not yet spread to verbs in other
positions). Similar to the Northern Italian dialects discussed in the previous section,
the existing agreement paradigm of Sutselvan seems to be quite rich. However, it has
been claimed that the grammaticalization of new agreement formatives did not start
in the present indicative, but rather in those tenses/moods which exhibit a defective
paradigmof secondary agreement endings. Subsequently, the newly created agreement
morpheme on C (and its exponents) spread to other contexts via analogical extension,
leading to a special inection in inversion contexts for all tenses/moods. In addition, I
have argued that the change in question was promoted by the fact that the attachment
of enclitics often affects the existing sufxal agreement morphology in the Swiss RRdi-
alects, due to a stress rule (penultimate stress target) which requires word stress to not
fall on the antepenultimate syllable of the verb. In many instances, this gives rise to the
impression that the enclitic replaces the existing agreement morphology, which abets
a reanalysis of the enclitic as the realization of a special form of inection associated
with inversion contexts.
In the next section, it is argued that language obsolescence may constitute another
setting where the loss of existing agreement morphology (due to incomplete acquisi-
tion of the target grammar) accelerates the development of new, alternative means of
agreement marking.
. Language loss and the grammaticalization of agreement markers
In this section, it is argued that minority languages which are in danger to disap-
pear may instantiate an alternative scenario where the grammaticalization of new
agreement markers correlates with the existence of a defective agreement paradigm.
More specically, it appears that pronouns may be employed as new markers of ver-
bal agreement if a generation of learners fails to fully acquire the original inectional
paradigm in a sociolinguistic setting where the language in question competes with
another language which is dominant and associated with social or cultural prestige.
Relevant examples come from the Walser German dialects discussed in Section 5.2.3.2
and American Russian, that is, the form of Russian spoken by the second (or third)
generation of Russian immigrants in the United States (Polinsky 2000).
Polinsky (2000) describes a set of structural properties of American Russian, fo-
cusing on speakers whose linguistic competence exhibits a higher degree of attrition.
The speakers under consideration are usually born in the US and acquired Russian
as their rst language. Later, but still as a child, they switched to English as their pri-
mary language, failing to acquire many aspects of the grammar of Russian properly.
In particular, the speakers have problems with the complex inectional morphology
of Russian, producing numerous verbal and nominal forms that are not acceptable in
standard Russian. Polinsky (2000: 87) summarizes the relevant structural changes as
follows:
60
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.48 (3315-3385)
[...] loss of case distinctions; loss of verbal agreement; elimination of the condi-
tional; loss or simplication of reexivization rules; development of resumptive
pronouns; loss of null copying under clause linkage, increased redundancy in
discourse.
In this section, I will focus on the use of resumptive pronouns in American Russian and
its correlation with the loss of agreement distinctions. Speakers of American Russian
tend to use only the masculine singular formof the verb in the past tense while they use
the 3rd person singular (or the innitive) in all other contexts, giving arise to apparent
agreement violations (examples taken from Polinsky 2000: 58):
(63) a. American Russian
moi
my
roditeli
parents
oni
they
kupil
bought-masc
drugoj
another
dom.
house
My parents bought another house.
b. Standard Russian
moi
my
roditeli
parents
kupili
bought-pl
ece
more
odin
one
dom.
house
My parents bought another house.
American Russian
(64) v
in
universitet
university.nom
knig-i
book-nom.pl
budet
will-be.3sg
dorogo.
expensively
Books will be costly when you go to the university.
Interestingly, this trait of American Russian correlates with another characteristic of
this language, namely the extensive use of resumptive pronouns in preverbal position,
as shown for example in (63a) above, where the subject my parents is resumed by the
pronoun oni they, giving rise to a doubling conguration. The resumptive pronoun
is obligatory if the subject is not adjacent to the verb or if the subject consists of two
conjoined DPs (the latter condition applies only to 1st and 2nd person contexts):
(65) ty
you
vcera
yesterday
ty
you
pozvonila
called.perf
moja
my.nom
mat
mother.nom
dlja
for
manikjur?
manicure
Did you call my mother for a manicure appointment yesterday? (Polinsky
2000: 59)
(66) [Dima
Dima
i
and
ja]
I
my
we
byli
were
vmeste
together
v
at
kole.
school
Dima and I went to school together. (Polinsky 2000: 59)
There are some indications that the resumptive pronouns in these examples do not
resume a left dislocated topic but should rather be analyzed as agreement marking ele-
ments (Polinsky 2000: 60ff.). First, whereas in Standard Russian, this kind of topic left
dislocation is only used to signal a change of topic in discourse, no such restriction can
be observed in American Russian. Second, in Standard Russian the resumptive pro-
noun is separated from the preceding phrase by an intonational break which may be
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.49 (3385-3453)
absent in American Russian.
61
Finally, in conjoined clauses, the resumptive pronouns
of American Russian exhibit a similar behavior as the subject clitics of Non-Standard
French, that is, they are obligatorily present in the second conjunct clause:
(67) ona
she
togda
then
ona
she
uvidela
saw
moju
my
mamu,
mom
i
and
ona
she
govorila
spoke
s
with
moej
my
mamoj.
mom
Then she met with my mom and nally spoke to her. (Polinsky 2000: 63)
Thus, American Russian differs signicantly from languages such as English, where
subject pronouns are usually elided in second conjuncts if they are coreferent with
the subject of the rst conjunct. Again, this suggests that the resumptive pronouns of
Standard Russian have evolved into agreement markers which are obligatorily present
independent of the syntactic context.
Polinsky (p. 58f.) proposes a functionalist explanation of these facts which links
the development of obligatory resumptive pronouns to the incomplete acquisition of
the verbal inection in American Russian:
As verbal agreement deteriorates, there arises a need for some other grammatical
mechanism marking the relation between the subject and the predicate. This ex-
plains, if only partially, another striking feature of American Russian, namely, the
widespread occurrence of the subject resumptive pronoun before the verb.
In the context of the present chapter, the basic idea of Polinskys can be captured in
more formal terms if it is assumed that the reanalysis of resumptive pronouns (pre-
sumably in topic left dislocation contexts) was guided by the workings of the Blocking
Principle: in environments where the resumptive pronouns met the necessary precon-
ditions on a reanalysis as agreement markers, the change in question was promoted
by the fact that the learner failed to develop full command over the existing agree-
ment paradigm. Instead, he/she acquired the (former) resumptive pronouns, which
carry unambiguous person/number/gender distinctions, as an alternative means of
agreement marking, in line with the Blocking Principle.
It seems that this kind of explanation carries over to the case of Pomatter-
titsch (Walser German) discussed in Chapter 5 above, Section 5.2.3.2. Recall that
in Pomattertisch, enclitic pronouns have evolved into sufxal agreement markers in
inversion contexts:
Pomattertitsch
(68) a. dets
this
bin-i
am-1sg
ich.
I
This is me. (Dal Negro 2004: 165)
b. d
the
mter
mother
set-sch.
says-3sg
Her mother says. (Dal Negro 2004: 171)
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.50 (3453-3515)
Dal Negro (2004) argues that this change is linked to the special socio-linguistic situa-
tion of Pomattertitsch, which is less prestigious than Standard Italian and the sur-
rounding Northern Italian dialects, facing extinction. As a consequence, younger
speakers often fail to fully acquire their dialect and in particular the inectional mor-
phology of the language. Similar to American Russian, this appears to lead to a gram-
maticalization process which serves to recomplete an otherwise defective agreement
paradigm:
62
In fact, original verb endings may have started to seem to be too opaque, at least
in the case of less uent dialect speakers, making it difcult to distinguish sin-
gle forms from one another, while the trend to omit subject pronouns was also
spreading. Hence, old, worn out endings have been replaced or rather reinforced
by new, more salient and agglutinative endings. (Dal Negro 2004: 178)
Again, the development of new agreement formatives appears to be linked to the in-
complete acquisition of the existing agreement paradigm which is a characteristic of
language loss. Similar to the case of American Russian, this observation can be at-
tributed to the workings of the Blocking Principle, which favors the acquisition of
the most specied agreement formative attested in the linguistic input. In the case at
hand, the BP stimulates the reanalysis of former subject enclitics in inversion contexts,
replacing the original agreement endings that are not acquired properly.
63
Let me now summarize the conclusions reached in the present section. It has been
shown that in languages which are threatened by obsolescence, the younger genera-
tion often develops only a restricted competence of the target grammar, in particular
failing to fully acquire the complex inectional morphology of the original system.
More specically, it has been demonstrated that in American Russian and the Walser
German dialect Pomattertitsch, the incomplete acquisition of the verbal agreement
morphology goes hand in hand with the development of obligatory subject clitics, giv-
ing rise to instances of clitic doubling if a full subject is present. It has been argued
that this phenomenon is also to be attributed to the workings of morphological block-
ing effects during language acquisition which scan the input for the most specied
candidate realizing a given inectional morpheme. Thus, the limited linguistic com-
petence characteristic of language loss/death may be accompanied by a set of rapid and
catastrophic changes which consist not only in the loss of structures of the target lan-
guage, but interestingly at times may also create new grammatical structures absent in
the target language, thereby reecting properties of UG such as the Blocking Principle
(similar observations can be made in pidgin and creole languages, cf. e.g. Bickerton
1984, 1998; Lightfoot 1991, 1999 for discussion). Note, however, that the present dis-
cussion covered only a minute fraction of the ever growing number of languages facing
extinction. In particular, the languages under consideration most probably do not rep-
resent typical examples. Usually, the disappearance of inectional morphology is not
compensated by innovative developments in languages facing extinction. Rather, it
appears that the effects of phonological and morphological reduction more often than
not win out over processes creating new morphology, similar to healthy languages
which are not in danger of disappearing (cf. Section 6.2.2 above).
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.51 (3515-3587)
. Grammaticalization and multiple agreement in Skou
So far, this chapter focused primarily on examples from Indo-European languages. Of
course, we expect that similar phenomena can be observed in other language families
as well. In this section, I present one such example coming from the Papuan family,
drawing on data collected by Mark Donohue.
Donohue (2002) discusses the phenomenon of multiple agreement in Skou, an
SOVlanguage spoken in some areas along the north coast of central NewGuinea. Skou
exhibits a set of different, but sometimes co-occurring strategies to mark agreement
with a subject: all verbs obligatorily carry pronominal clitics for this purpose. In addi-
tion, some verbs undergo vowel and consonant alternations, as well as combinations of
these. Donohue shows that the kind of multiple agreement found in Skou most likely
resulted from the repeated reanalysis of pronominal clitics as means of verbal agree-
ment marking in the history of this language. Accordingly, the consonant alternations
found on the verb root are attributed to an earlier process in which subject pronouns
fused with the verb. The subsequent loss of this kind of distinctive agreement mor-
phology (due to phonological reduction) is in turn compensated for by a second wave
of cliticization that led to the peculiar kind of multiple agreement marking that can be
observed in the present-day language. Before we turn to the historical developments
affecting clitic pronouns in some more detail, the following section briey reviews the
different strategies of verbal agreement marking which can be observed in Skou today.
.. Strategies of agreement marking in present-day Skou
With all verbs, subject-verb agreement is obligatorily marked by pronominal clitics
that attach to the left of the verb (or to the left of a so-called adjunct nominal which
is part of the verbal complex). The proclitics, which are listed in Table 30, are more or
less similar in shape to the free pronouns.
For a certain number of verbs (according to Donohue, this class consists of 20
verbs) agreement is solely marked by the use of proclitics. The following examples
show that the proclitic must be present to form a grammatical sentence:
(69) a. *Peangku
girl
fue
that
a wung.
die
That girl died.
b. Peangku
girl
fue
that
a pe=wung.
3sg.fem=die
That girl died.
Table 30. Proclitic agreement markers in Skou.
Singular Plural
1. n ne
2. m e
3. masc/neut ke te
3. fem pe
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.52 (3587-3684)
Table 31. Agreement via stem vowel alternation: l weng to sleep.
Singular Plural
1. l weng l weng
2. l weng l weng
3. masc/neut l weng l weng
3. fem l weng
Table 32. Agreement via stem vowel alternation: fue to see.
Singular Plural
1. fue
2. fue
3. masc/neut fue
3. fem fu
Furthermore, the proclitic apparently has no pronominal function: an independent
overt subject (a pronoun or a full DP) is always required, as illustrated in (70). Ac-
cording to Donohue, the pronominal clitics are therefore better analyzed as simple
agreement markers.
64
(70) a. *Pe=wung
3sg.fem=die
b. Pe
she
pe=wung.
3sg.fem=die
She died.
A second strategy to mark agreement with the subject is realized by stem vowel al-
ternations affecting the verb root. In contrast to agreement marking via a proclitic,
these vowel changes affect only a couple of verbs and mark only a restricted set of dis-
tinctions.
65
More specically, stem vowel alternation is used to mark only the features
[feminine] and (less often) [plural] (mostly 3pl), where rounding and backing signals
gender and fronting signals plurality. This is illustrated in Table 31 and Table 32 for the
verbs l weng to sleep and fue to see, respectively.
With the complex verb l weng to sleep, vowel changes are found in the forms for
3sg.fem and 3pl. Note that in the case of l weng, the vowel alternations affect the so-
called adjunct nominal, a nominal element which cannot be separated from the verb
root, while the verb root itself remains unchanged.
66
In the case of fue to see, the vowel
changes affect the verb root. Moreover, in contrast to l weng to sleep, all plural forms
exhibit the relevant stem vowel alternation. As noted above, the vowel changes alone
are not sufcient for marking agreement with the subject an additional proclitic is
always required:
(71) a. Ke
he
me
sh
ke=fue.
3sg.masc=see
(*Ke me fue.)
He saw a sh.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.53 (3684-3782)
b. Pe
she
me
sh
pe=fu.
3sg.fem=see.fem
(*Pe me fu.)
She saw a sh.
The majority of verbs (68%according to Donohue 2002: 13) exhibits an additional way
of agreement marking via consonant alternations which affect the onset of the verb
root (or the adjunct nominal). This kind of agreement serves to signal more distinc-
tions than the vowel changes discussed previously. Some verbs (such as h to close)
exhibit up to six different onsets dependent on the -features of the subject.
67
Still,
all verbs that signal agreement via consonant changes must appear with an additional
proclitic and an overt subject:
68
(72) a. Pe
she
m
you
pe=w.
3sg.fem=catch.3sg.fem
She caught you.
b. *M w.
c. *M pe=w. (Donohue 2002: 13)
Although there is considerable idiosyncratic variation between different verbs, Dono-
hue identies a set of underlying consonantal prexes, which are listed in the last
column of Table 33 (Donohue 2002: 12).
Another interesting property of the consonant changes comes to light if we com-
pare the shape of the consonantal prexes which replace the onset of the verb root with
the proclitics which are used on all verbs to mark agreement, cf. Table 34.
Table 33. Agreement via onset alternations: selected verbal paradigms.
k h w lng ang o Prex
catch close get chop eat seawards
1sg k h w lng kang o -, k-, n-
2sg b m p png mang mo m-
3sg.masc/neut k k w lng kang ko k-
3sg.fem w w w wng pang po p-
1pl k n w tng nang no n-
2pl k h w lng ang o -
3pl ki y w yng tang to t-, y-
Table 34. Proclitics and consonantal agreement prexes in Skou.
Proclitics Prexes
1sg n -, k-, n-
2sg m m-
3sg.masc/neut ke k-
3sg.fem pe p-
1pl ne n-
2pl e -
3pl te t-, y-
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.54 (3782-3844)
Even a supercial inspection of the paradigms listed in Table 34 reveals the sys-
tematic similarities between the proclitics and the corresponding consonantal prexes.
These similarities clearly indicate that there is a diachronic relation between the pro-
clitics and the consonantal prexes which is briey discussed in the next section.
.. The historical origin of multiple agreement marking in Skou
As already noted above, it is fairly clear that the consonant prexes found in Skou
developed from former clitic pronouns that underwent phonological reduction und
fused onto the verb (Donohue 2002: 28) in the course of time. Furthermore, it has
been noted that these prexes occur only with a highly restricted set of onsets in Skou.
A comparison of the agreement prexes found in Skou with those of closely related
languages reveals a set of interesting differences. In contrast to Skou, languages such as
Dumo, Dusur, and Leitre exhibit a more articulated system of consonantal agreement
prexes. In these languages, the consonantal prexes attach to a wider range of on-
sets and the resulting verbal agreement paradigms are generally more complete, that
is, there are more distinctions available and there is less amount of syncretism. Dono-
hue (2002: 29) summarizes the differences between Skou and the neighbouring eastern
languages as follows:
(73) Agreement marking via consonantal prexes in Skou
(i) phonological mergers have collapsed a number of contrasts;
(ii) there are less syllable onset types available for inecting verb roots (only
four in Skou, as opposed to six or seven in the east);
(iii) phonotactic constraints rule out clusters on onsets;
(iv) there is a large number of uninecting verb roots.
Donohue shows that the reduction of consonant clusters led to a collapse of contrasts
in many agreement paradigms of Skou (due to the simple (C)V syllable structure of
Skou, cf. Donohue 2002: 4, Fn. 4). In some cases, cluster reduction resulted in the com-
plete loss of consonantal agreement marking, giving rise to the class of non-inected
verbs, where agreement is solely marked by the use of a proclitic in the present-day
language.
69
Crucially, all languages that retained a rather rich system of consonant prexes
have not developed a further way of agreement marking via proclitics. Donohue there-
fore concludes rightly in my view that there is a connection between the loss of
agreement distinctions and the second wave of pronominal clitics that became re-
analyzed as agreement markers. However, in contrast to Donohue, who proposes a
functionalist explanation of this development, I suggest that the loss of agreement
distinctions paved the way for the reanalysis of subject clitics as new means of ver-
bal agreement marking in accordance with the Blocking Principle. More specically,
it seems likely that the grammaticalization of new verbal agreement markers began
in those contexts where cluster reduction led to the complete loss of consonantal
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.55 (3844-3914)
agreement distinctions. In a subsequent change, the new agreement markers spread
to other verb classes via analogical extension, similar to the examples from Vicentino
and Sutselvan discussed above. Presumably, the latter change involved the insertion of
a dissociated agreement morpheme to some higher functional head, giving rise to the
form of multiple agreement exhibited by Skou (cf. Chapter 3 above).
. Conclusion
In this chapter, I have focused on the observation that across languages, the cre-
ation of new inectional morphology generally serves to rebuild previously defective
paradigms, exchanging weak old forms for stronger new forms. It has been claimed
that this observation can be explained if it is assumed that the grammaticalization
of new agreement markers from former clitic pronouns is triggered if this change
leads to new agreement markers which are more distinctive than the relevant agree-
ment formatives which already exist in the language. This has been modeled as the
outcome of blocking effects which operate during language acquisition and block the
acquisition of a less specied form if a more specied form is attested in the Primary
Linguistic Data. Thus, if the learner encounters more than one potential realization of
a certain feature combination in the input, he/she will select the candidate which re-
alizes the greatest subset of the relevant agreement features for storage in the Lexicon.
The relevant principle of grammar, dubbed the Blocking Principle, is understood as
an economy principle which shapes the acquisition of inectional morphology, war-
ranting an optimal and non-redundant Lexicon (and paradigm) structure. Similar to
other (structural) economy principles (cf. Clark & Roberts 1993), the Blocking Princi-
ple is called into service only if the cues provided by the input data are for some reason
ambiguous and not sufcient for the acquisition of certain properties of grammar.
The Blocking Principle was introduced in Section 6.2 by looking at the grammat-
icalization of new agreement formatives in the history of Bavarian, arguing that an
approach in terms of morphological blocking explains that this change affected only
underspecied slots of the agreement paradigm. It was shown that this approach ac-
counts not only for the development of new agreement exponents in Bavarian, which
unambiguously mark 2pl (/-ts/) and 1pl (/-ma/), but also for the much earlier devel-
opment of 2sg /-st/, which took place in the Old High German period and led to a 2sg
present form that was unambiguously specied for verbal mood (indicative), in con-
trast to its predecessor, /-s/. Thus, assessing of whether a certain change complies with
the Blocking Principle sometimes requires that inectional distinctions other than per-
son and number are taken into consideration. Moreover, I argued that the diachronic
analysis presented in this section makes available a newexplanation of the peculiar per-
son/number restrictions on complementizer agreement and pro-drop in present-day
Bavarian which have been observed in Chapter 5. More specically, it was claimed that
the limited distribution of these properties reects the morphological contexts where
the reanalysis in question was licensed by the Blocking Principle.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.56 (3914-3962)
Subsequently, I discussed a set of further theoretical issues in connection with this
approach to the rise of inectional morphology, focusing on the status of morphologi-
cal doublets (Section 6.2.3) and the fact that most changes seem to expand the domain
of less marked, more regular forms (Section 6.2.2), which seems to be at odds with the
claimthat stronger inectional formatives are preferred. In Section 6.2.2, it was argued
that analogical changes which lead to the loss of stemallomorphy can also be attributed
to the workings of morphological blocking if it is assumed that the learner prefers to
acquire the stem alternant which manifests the characteristic phonemic distinctions of
a givenlexical root in the most unambiguous way. In addition, I considered cases where
an underspecied exponent gains a wider distribution in a paradigm, arguing that this
kind of paradigm leveling is to be attributed to the expansion of the domain of Im-
poverishment rules which bleed the effects of the Blocking Principle. Concerning the
existence of morphological doublets, that is, apparent multiple competing realizations
of a given inectional head, I argued in Section 6.2.3 that the Blocking Principle allows
the coexistence of morphological doublets only if the relevant agreement formatives
are realizations of different agreement morphemes with identical feature content (e.g.,
Agr-on-C and Agr-on-T).
In Section 6.3, it was demonstrated that the pioneering role of 1st and 2nd per-
son in the historical development of verbal agreement marking (cf. Chapter 1) can be
attributed to the workings of the Blocking Principle if it is assumed that 3rd person
forms are underspecied for person features. As a result, the grammaticalization of
3rd person forms is less likely to be triggered than the development of 1st and 2nd
person markers, which are generally specied for a set of [person] features.
The Sections 6.46.8 presented a set of case studies which further corroborated
the hypothesis that the reanalysis of pronominal clitics is guided by morphological
blocking effects. Section 6.4 dealt with the development of new agreement forma-
tives in Non-Standard French. It was shown that the grammaticalization process in
question proceeded in line with the Blocking Principle, affecting only those slots of
the paradigm where the existing agreement morphology is non-distinctive. Concern-
ing the question of why the number and gender specication of the 3rd person clitics
il(s) and elle(s) is apparently not sufcient to trigger the grammaticalization process in
question, I claimed that verbal agreement in gender cannot evolve unless the language
has not acquired a full paradigm of number distinctions for all persons (assuming
that -features are organized in a feature geometry where number features dominate
gender features). Furthermore, I conjectured that the limited distribution of plural
-s (which is only perceivable if the following word begins with a vowel) reduces its
visibility for the workings of the Blocking Principle.
In Section 6.5, I discussed the development of new agreement markers in Piattino
and Vicentino, two Northern Italian dialects where the existing verbal agreement mor-
phology is quite rich. It was argued that the changes in these dialects can be shown to
have led to the creation of stronger specied forms if more inectional distinctions are
taken into consideration. More specically, I demonstrated that in both dialects, the
grammaticalization process in question led to 3rd person agreement markers which
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.57 (3962-4013)
are additionally specied for gender. Furthermore, it was claimed that in Vicentino, the
grammaticalization process in question was also sensitive to the realization of verbal
mood, giving rise to a 2pl form which was unambiguously specied for [+indicative],
in contrast to its predecessor. Finally, a more complex scenario was devised for the
development of new 2sg agreement markers in Vicentino. It was argued that 2sg cli-
tics were rst reanalyzed in the context of a set of very frequent irregular verbs which
are characterized by a defective agreement paradigm. By assumption, the new agree-
ment formatives expanded to other verbs via analogical extension in a later gradual
development.
A similar scenario, where a change begins in a certain verb class before it gains a
wider distribution, was proposed for the Swiss Rhaeto-Romance dialect Sutselvan in
Section 6.6. I argued that the reanalysis of enclitics was initially conned to contexts
where Sutselvan exhibits a set of impoverished (secondary) agreement endings which
appear in all tenses/moods apart from the present indicative and imperative. Subse-
quently, the newly created agreement exponents spread to other contexts via analogical
extension, leading to a special inection in inversion contexts for all tenses/moods. In
addition, I suggested that the change in question was promoted by the fact that the
existing agreement morphology is often truncated when enclitics attach to the verb in
inversion contexts, due to a stress rule (penultimate stress target) which requires word
stress to not fall on the antepenultimate syllable of the verb.
In Section 6.7, I showed that the Blocking Principle can also be invoked to account
for the observation that new means of agreement marking may evolve in sociolin-
guistic contexts where speakers develop only a restricted competence of the target
grammar. It was argued that the development of obligatory clitic doubling in Ameri-
can Russian and certain Walser German dialects which are threatened by extinction is a
consequence of the incomplete acquisition of the existing verbal agreement morphol-
ogy: if the learner fails to acquire the existing agreement distinctions properly, he/she
might scan the input for a more salient (and more specied) candidate realizing a
certain inectional morpheme, which might lead to rapid and catastrophic changes
before a language nally vanishes.
In Section 6.8, I discussed data from the Papuan language Skou, in which subject
clitics were repeatedly reanalyzed as agreement markers. It was argued that the gram-
maticalization of these newproclitic agreement markers balanced the loss of distinctive
verbal agreement markers, which resulted from the phonologically triggered reduction
of consonant clusters. Although the brief discussion could not really do justice to the
complex agreement patterns exhibited by Skou, the evidence presented was neverthe-
less suggestive, indicating that the reanalysis of subject clitics as new means of verbal
agreement marking proceeded in line with the Blocking Principle.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.58 (4013-4076)
Notes
. Often the intuition behind the blocking effect is that the availability of an irregular form
blocks the use of the regular form. Note that a formulation of the blocking effect that refers to
the irregular/regular distinction covers a wider array of cases and is not conned to inectional
morphology (e.g., it includes cases such as the famous thief vs. *stealer contrast).
. Note that this use of the term defective actually deviates from the traditional understanding
of defectiveness in morphological theory. As pointed out by Albert Ortmann to me, the notion
defective paradigm is commonly used to refer to a situation where a given lexical item fails to
exhibit certain inected forms (such as some of the English modals which cannot showup as in-
nitives: *Peter wants to can play the guitar.). However, for expository reasons, I will continue to
use the notion defective paradigm to refer to an agreement paradigm which exhibits a number
of syncretisms.
. The correlation between a defective agreement paradigm and the obligatory presence of
pronominal clitics has already been noted in traditional work on the history of the Romance lan-
guages (including French and various Northern Italian dialects), see Kuen (1957) or Wartburg
(1970), for example. Note that the idea (going back to Gabelentz 1891) that grammaticalization
processes are motivated by the need to compensate for the loss of distinctions due to phonolog-
ical reduction is widely held in the literature on grammaticalization, cf. Ldtke (1980), Hopper
and Traugott (1993), Haspelmath (1995), Siewierska (1999, 2004), Ariel (2000), and Lehmann
(2002), among others.
. Possibly, the Blocking Principle can be reduced to general principles governing the proce-
dure of Vocabulary Insertion, that is, the Subset Principle (cf. Chapter 2). Recall that the Subset
Principle requires that Vocabulary items realizing functional morphemes (f -morphemes) en-
ter into a competition which warrants that only the most specied candidate is inserted in a
given insertion context. In contrast, exponents of lexical morphemes (l-morphemes) are freely
inserted, that is, there is no competition. If it is further assumed that grammaticalizationis to be
analyzed as a change in which an exponent of an l-morpheme changes into an exponent of an
f -morpheme, one might posit that this transition is only possible if the resulting inectional for-
mative fullls the Subset Principle, that is, qualies as the most specied candidate with respect
to the insertion context at hand. Note that the enhanced paradigmaticity of grammaticalized
elements (cf. Chapter 2 above) is presumably also to be attributed to this change concerning
the insertion procedure, in the sense that the Vocabulary items which compete for realizing a
given f -morpheme share a certain feature specication which qualies them for entering into
this competition and gives rise to the impression that these Vocabulary items constitute a certain
paradigm.
. Another piece of evidence which helps us to gain some insights into the workings of blocking
effects during language acquisition comes from the chronological sequence of the acquisition
of inectional morphology (cf. Clark 1998). It is a well-known fact that children typically be-
gin to use a certain inection sporadically until they have mastered the relevant morphological
rule. From this point on, they use the given inection consistently and produce over-generalized
forms such as foot-s and man-s (in the case of the English plural inection). In a subsequent
development, they come to recognize the limits of the rule in question and replace the over-
regular forms with the correct irregular ones (e.g., feet and men). If we apply these observations
to the acquisition of (irregular) inectional formatives in general, this suggests the following
two things: (i) presumably, the Blocking Principle becomes relevant only after the acquisition
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.59 (4076-4157)
of basic morphological rules (perhaps due to maturation processes); (ii) at this stage of lan-
guage acquisition, the storage of inectional formatives for a given paradigm is not yet xed,
so irregular forms can replace over-regular forms (irregular forms are presumably learned in a
one-by-one fashion, cf. Bybee & Slobin 1982; Kiparsky 1983; Prasada & Pinker 1993).
. As noted in Chapter 4 above, the pronoun must fulll a whole set of independent necessary
conditions before it can be reanalyzed as the exponent of an agreement morpheme (e.g., it must
be a phonologically reduced clitic that cannot receive stress, the reanalysis has to satisfy a set of
syntactic conditions such as adjacency to the host, etc.).
. Presumably, the BP ensures acquisition of the more specied form only if the relevant can-
didate form is already robustly attested in the PLD. In other words, frequency is another factor
that determines the acquisition of morphology if a less specied form is much more frequent
than the more specied form, then the learner will probably acquire the less specied form,
despite the workings of the BP. Furthermore, regularization is a factor that might work against
the effects of the BP: In the absence of robust evidence for a irregular/more specied form, the
learner may acquire a regular/underspecied form for a given verb as a default. Evidence from
language acquisition (Prasada & Pinker 1993; Clark 1998) and language change (Taylor 1994)
shows that this process affects primarily verbs that are less frequent in the PLD, a fact which is
in line with the properties of the BP suggested above (i.e., the BP selects between forms that are
robustly attested in the input).
. As Sauerland (1996: 24) puts it: If two feature complexes in the same position and in the
same syntactic environment get spelled out by the same phonological string, the lexical entry of
this phonological string has to be unspecied with respect to all contrasting features of the two
feature complexes.
. For 13th century Bavarian, the sets of Vocabulary items that competed for insertion into the
Agr-morpheme (present tense indicative) are thus as follows. (i) lists the set of items prior to the
development of 2pl /-ts/, (ii) shows the situation after the change in question has taken place.
(i) [1, +pl] /-an/
[1, pl]
[2, pl] /-st/
[pl] /-ant/
elsewhere /-t/
(ii) [1, +pl] /-an/
[2, +pl] /-ts/
[1, pl]
[2, pl] /-st/
[pl] /-ant/
elsewhere /-t/
. The relevant sets of Vocabulary items for 18th century Bavarian are listed below. (i) lists the
items prior to the rise of 1pl /-ma/, while (ii) shows the resulting paradigm.
(i) [2, +pl] /-ts/
[1, pl]
[2, pl] /-st/
[pl] /-an/
elsewhere /-t/
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.60 (4157-4223)
(ii) [1, +pl] /-ma/
[2, +pl] /-ts/
[1, pl]
[2, pl] /-st/
[+pl] /-an/
elsewhere /-t/
. Note that the initial vowel in formatives such as -ems is actually not part of the agreement
sufx, but rather a so-called theme vowel that originally served to derive verb stems fromroots.
. Strong verbs and the weak verbs of conjugation class 1 exhibit -is and -s for 2sg present
indicative and 2sg present subjunctive, respectively. Here, the difference in vowel quality was
perhaps not salient enough to differentiate the forms. Furthermore, the difference was pre-
sumably further weakened by phonological erosion that affected non-stressed nal syllables.
Alternatively, one might assume that the change rst affected the weak verbs of the conjugation
classes 2 and 3 and spread later to other verb classes by analogy.
. In fact, the syntactic consequences of morphological erosion represent one of the best stud-
ied phenomena in diachronic generative syntax (cf. e.g. van Kemenade 1987; Lightfoot 1991,
1999; Roberts 1993a, 1997; Falk 1993; Allen 1995; Haeberli 1999; Pintzuk 1999; Lightfoot 2002).
. The stem vowel alternations found in older Yiddish have different historical origins. With
verbs such as grfbn to dig, the vowel alternations were introduced by a phonological process
(Umlaut) operative in Middle High German (note that Yiddish is assumed to have evolved from
this historical stage of German) which changed an a in the verb stem to an e in the 2sg and
3sg, due to the presence of an /i/ in the sufxes of these forms (strong classes VI and VII). In
contrast, vowel alternations between singular and plural forms (in the present tense) of verbs
such as visn to know are of a more ancient origin. Verbs such as visn are so-called preterite-
presents, the present tense of which derives from Proto-Indo-European perfect forms where
stem vowel alternations distinguished between singular and plural (Ablaut).
. Recently, this intuition has been formalized in Optimality Theoretic terms, positing con-
straints such as UNIFORM EXPONENCE or PARADIGM UNIFORMITY (cf. e.g. Kenstowicz
1997; Steriade 2000), which prefer inectional paradigms to be uniform. Analogical leveling is
then described in terms of constraint re-ranking.
. To be sure, further questions arise concerning the diffusion of analogical leveling in a speaker
community. Concerning the changes affecting Yiddish, Albright (2002: 29) stresses that his ap-
proach is not primarily concerned with the latter problem, but rather deals with the question
which forms may be affected by analogical leveling (my emphasis, E.F.): There are clearly many
factors at play in determining how willing a community is to adopt new forms; I conjecture that
the thoroughness of the change in Yiddish may have been facilitated by the lack of a standard
language or widespread literacy, and perhaps even by a conscious desire to differentiate Yiddish
from German. The model that I am presenting here is simply an attempt to predict which forms
would have been available as potential regularizations for Yiddish speakers, and which would not.
. Noyer (1997) discusses several examples of analogical change which can be observed in the
historical development of a number of Afroasiatic languages and analyzes them in terms of the
gain or loss of Impoverishment rules. Another example which is quite similar to the change
depicted in Table 10 involves the loss of gender distinctions in the plural part of the verbal
agreement paradigm of Egyptian Arabic (Noyer 1997: 58ff.).
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.61 (4223-4293)
. Halle (1997: 430) speculates that the expansion of Impoverishment rules is linked to the
economy principle in (i), which is supposed to minimize the load on memory for the learner
(see Noyer 1997: lxxxi for similar considerations):
(i) The number of features mentioned in the Vocabulary must be minimized.
Due to the fact that Impoverishment expands the domain of underspecied forms, (i) may pro-
mote the development of Impoverishment rules which delete certain morphosyntactic features
of syntactic heads in certain contexts. Thus, at rst sight, an economy principle such as (i) seems
to contradict the workings of the Blocking Principle, since the latter requires new Vocabulary
items to realize more feature distinctions than the existing items. However, note that (i) is in
fact not intended to mean that the learner selects the least marked/specied of several potential
candidates and stores it in the lexicon. Rather, it is supposed to ensure that the child acquires the
most economical lexical inventory compatible with the input he/she is exposed to. For example,
if a feature specication [+speaker] is sufcient to guarantee that a 1sg exponent is inserted in
the contexts where it appears in the input, the child will not acquire a redundant feature speci-
cation [+speaker, hearer] for this exponent. Crucially, however, this function of (i) does not
interfere with the claimthat the learner scans the input for the most marked (and therefore most
salient) realization of a given inectional head. In other words, the Blocking Principle ensures
that the most specied candidate is selected, while (i) warrants that this candidate is assigned a
non-redundant feature specication. Thus, the two principles actually work hand in hand dur-
ing acquisition, thereby warranting that the acquisition process leads to an optimal paradigm
and Lexicon structure.
. It is quite obvious that languages with agreement syncretisms may exist happily over cen-
turies. This seems to support the viewthat language change is a contingent process (cf. Lightfoot
1999). On this view, properties of UG draw up the boundaries of possible changes, but crucially,
they do not trigger changes. Note that this apparently conicts with the workings of the BP as
portrayed above, where it has been argued that the BP actually may trigger a change in the gram-
mar if more than one phonological exponent is available for a given inectional head. Here, I
will simply note this tension without trying to resolve it.
. As noted above in Chapter 5, the development of 2sg -s+t and 2pl -t+s presumably pro-
ceeded in a similar fashion, with the new formative initially conned to inversion contexts.
. In most cases, the observed variation can be traced back to northern innovations that pro-
duced new regular (weak) forms for historically irregular (strong) verbs such as welk vs. walked,
awoke vs. awaked etc. Note that this kind of morphological regularization is a characteristic trait
of imperfect second language acquisition by adults. Taylor therefore concludes that the weak past
tense variants were the outcome of an intense language contact that obtained between Scandi-
navian immigrants and speakers of northern English dialects from the late Old English period
on. Subsequently, the new forms spread to other (southern) ME dialects via dialect mixture. A
remarkable number of past tense doublets can be found in the contemporary language of the
Londonarea, which was a melting pot of different dialects in the ME period. For example, Taylor
shows that in the work of Chaucer, 24 verbs can be found that exhibit varying forms. In contrast,
the Paston letters, which were written in a local Norfolk dialect approximately at the same time,
contain only three pairs of morphological doublets for past tense verbs.
. However, the scenario envisaged by Kroch can probably account for the gradual replacement
of the old ending in clause-nal contexts, which apparently affects one verb after the other,
starting with frequently used short verbs such as have and do (see Chapter 5).
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.62 (4293-4362)
io The Rise of Agreement
i. In addition, we have to acknowledge that there is a difference between the grammar change
as such and its diffusion in the speaker community. Thus, even if a given change has taken place
(in the grammar of some speakers), it does not necessarily prevail, due to socio-linguistic factors.
In the case of 1pl /-ma/, for example, the change is restricted to some Lower Bavarian varieties
which became marginalized due to the inuence of urban (i.e., Munich) Bavarian and are nearly
extinct today (cf. Altmann 1984; Kollmer 1987).
i|. Recall that in Chapter 3, Note 62, it has been assumed that an independent principle of
Morphological Structure requires that in a given head complex, only the hierarchically highest
of two identical agreement morphemes is subject to Vocabulary Insertion, ruling out forms of
the type V+Agr-on-T+Agr-on-C. Presumably, this principle forces the learner to assume the
existence of a separate, higher agreement head when the relevant exponent resulting from the
reanalysis of a clitic differs from the (existing) agreement ending found in other contexts: if
the new exponent (found in inversion contexts) differs from the old agreement ending (found
in clause-nal position), then the learner will assume that the new form is the exponent of an
additional, hierarchically higher agreement morpheme the spell-out of which overwrites the
old exponent in a certain syntactic context. Note that the relevant input did not contain forms
which could be reanalyzed as instances of double agreement (i.e., V+Agr-on-T+Agr-on-C): in
the case of the enlargements 2sg -s+t and 2pl -t+s, the combination of old agreement marker
and clitic is reinterpreted as a single new agreement ending while in the case of 1pl /-ma/, the
reanalysis rst affected bisyllabic verbs, where the old agreement ending has been deleted in
the presence of a clitic, quite similar to the effects of the penultimate stress rule in the Swiss
Rhaeto-Romance dialects (cf. Section 5.3.4).
i,. The condition that the newly created agreement exponent is conned to the C-systemcan be
expressed by specifying the syntactic insertion context in the relevant lexical entry, for example:
(i) [1, +pl/C] /-ma/
[+pl] /-an/
Note that initially, the old and the new forms do not compete for the realization of Agr-on-T.
The gradual expansion of the new form to Agr-on-T can then be modeled as the loss of the
contextual restriction in the lexical entry of the more specied form, which eventually drives the
old form out of the grammar, presumably along the lines envisaged in Kroch (1994).
io. Note that when the reanalysis in question took place, the input presumably still included
examples with full (instead of clitic) pronouns where the old, underspecied agreement ending
showed up on verbs in inversion contexts. Still, the old endings were not acquired as realizations
of Agr-on-C.
i. The special role of 1st and 2nd person is also reected by the historical developments in
Bavarian discussed above. In cases where 1st/2nd person forms and 3rd person forms were
homophonous, the grammaticalization processes always created new, more specied 1st/2nd
person forms, while in the 3rd person, the old forms were preserved.
i8. Further support for a model that makes use of only two instead of three person features
comes from an interesting cross-linguistic restriction on possible pronoun systems (Bobaljik &
Baker 2002). On the basis of the traditional model that includes separate features for 1st, 2nd,
and 3rd person, we should expect to nd languages that distinguishes up to seven different
morphologically simple pronouns in the plural:
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.63 (4362-4425)
(i) a. 1+2 inclusive we
b. 1+3 exclusive we
c. 1+2+3 complete we (refers not only to speaker and hearer, but also to
other persons)
d. 1 groupspeak we (refers to a plurality of speakers, e.g., a crowd
shouting we want more! at the end of a concert)
e. 2 exclusive you
f. 2+3 inclusive you
g. 3 they
However, the attested maximum number of distinct morphologically simple plural pronouns
is apparently four (as exemplied e.g. by the Australian Aboriginal language Kitja, Dixon
2002: 247). In descriptive terms, the fact that only four out of seven possible interpretations
are realized by separate pronominal forms is due to the following pattern of syncretisms, which
seems to hold universally across languages (cf. Bobaljik & Baker 2002):
(ii) a. inclusive we = complete we
b. exclusive we = groupspeak we
c. exclusive you = inclusive you
(iia) asserts that in all languages that show the maximum number of four different plural pro-
nouns the form for inclusive we is also used to express the reading of complete we. Similarly,
according to (iib), there are no morphologically simple pronouns that refer solely to groupspeak
we, whereas (iic) states that there can be only a single form that is used for reference to both
exclusive you and inclusive you. Interestingly, exactly this pattern is predicted by a binary per-
son feature system, in which it is not possible to distinguish between inclusive we and complete
we. Since there is only a single feature combination that corresponds to both inclusive we and
complete we, one and the same pronoun must be used to refer to these different readings of 1st
person plural. The other universal syncretisms receive a similar explanantion, cf.
(iii) a. [+Auth, +PSE] = inclusive we & complete we (cf. iia)
b. [+Auth, PSE] = exclusive we & groupspeak we (cf. iib)
c. [Auth, +PSE] = inclusive you & exclusive you (cf. iic)
. Note that you is also the formused in the context of 2sg. Therefore, you is underspecied for
number features. Since you and they realize the same number of features, the ordering between
(20b) and (20c) is necessary to prevent they from being inserted as the realization of 2pl.
. Note that this generalization is apparently contradicted by English, where 3.sg.pres.indic.
/-z/ appears to be the only verbal agreement ending. This seems to suggest that the relevant
Vocabulary item carries a person specication which is absent otherwise (in all other contexts
we nd the elsewhere case -, which is completely underspecied for person and number).
However, see e.g. Halle (1997) for an analysis where 3.sg.pres.indic. /-z/ is analyzed as an in-
ectional marker that in fact does not realize person distinctions, but rather carries the feature
specication [pl, +pres, +nite]. Ian Roberts pointed out to me an alternative way of analyzing
3.sg.pres.indic. /-z/, in which /-z/ is treated as the non-default ending in a system that makes use
only of a two-way default/non-default distinction in the present tense indicative (see Haeberli
2004 for related considerations).
. This reasoning also provides an answer to the question of why new agreement sufxes devel-
oped for 2pl and 1pl, but not for 3rd person forms in the history of Bavarian. Note that the latter
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.64 (4425-4492)
ii The Rise of Agreement
development would have been equally sufcient to repair a defective paradigm where 2nd (and
1st) person forms are homophonous with 3rd person forms. However, if 3rd person forms are
inherently underspecied for person features, they are less prone to be replaced by new, equally
underspecied exponents, which explains the fact that the new sufxes developed only in 1st
and 2nd person contexts.
i. Guiraud (1968: 98) notes that in Old French, SVOorder requires anovert subject pronoun in
97% of all cases, whereas the ratio of overt subject pronouns is only 15% in OVS orders where
the clause-initial object is a full DP (i.e., where the V2 constraint is satised for independent
reasons).
. Cf. Greenberg (1966a: 94): Universal 32: Whenever the verb agrees with a nominal subject
or nominal object in gender, it also agrees in number.
|. The structure in (37) represents a modied version of proposals by Harley and Ritter (2002),
who assume for example an elaborate structure of number features to account for intricate
number systems which show distinctions such as dual, trial or paucal. For expository reasons,
however, I chose to adopt a somewhat simplied model, which is sufcient for the present
purposes.
,. In addition, we might speculate that, somewhat paradoxically from the viewpoint of the
Blocking Principle, it is exactly the gender distinctions which lead the learner to assume that
il/elle and ils/elles are still pronouns and not agreement markers: recall that gender features are
a well-established aspect of the pronoun system of French, but completely absent in the verbal
agreement paradigm (at least in the case of nite verbs). It is therefore conceivable that it is this
difference which inhibits a categorial reanalysis of il/elle and ils/elles.
o. Gerlach (2001, 2002) proposes an Optimality Theoretic account of the distribution of clitics
in Piattino. Note furthermore that according to Gerlach (2002: 225), the 3sg and 1pl forms actu-
ally do not have an agreement ending at all, but end merely in the theme vowel -a. For expository
reasons, I nevertheless labeled the respective forms 3sg and 1pl in the glosses.
. Historically, the 1pl an developed from impersonal structures of the kind homo (man) +
verb-3sg, quite similar to the use of on + verb-3sg in colloquial French (cf. e.g. Savoia 1997: 78).
8. As noted above in Note 36, -a is usually analyzed as a theme vowel and does not constitute
an agreement sufx. Furthermore, Gerlach (2002: Fn. 24, p. 235) argues that 3pl -en is only the
realization of [number] and not of [person].
. I am grateful to Adriana Castagna and Ermenegildo Bidese for much discussion and native
speaker judgements on Vicentino.
|o. Note that in the 3sg and 3pl, the attachment of the enclitic leads to a change in the quality
of the theme vowel ([a] [e]).
|:. Of course, the full 2pl pronoun voaltri can be added for reasons of emphasis:
(i) Voaltri
you.pl
a
clit.2pl
vegn
come-2pl
da
from
Durlo.
Durlo
You come from Durlo.
|i. Poletto (1999, 2000) links the morphological properties of the different kinds of clitics with
their different syntactic distribution. More precisely, she identies four different types of clitics
in the Northern Italian dialects and claims that these clitics realize different kinds of agreement
heads (some above and some belowNegP) each of which is associated with a different set of mor-
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.65 (4492-4565)
phosyntactic features. Accordingly, AgrS is split into a series of functional heads with different
content, corresponding to the features [deictic], [person], [gender], and [number].
. Interestingly, in the case of 3sg.masc, the proclitic (e)l does not attach to the verb, but
rather to the preceding negation no, presumably for phonological reasons (no+l leads to a well-
formed syllable structure, while the combination of l+vien would create a consonant cluster in
the onset).
(i) u
he
no=l
not=clit.3sg.masc
vien
come
a
to
casa
home
stasera.
tonight
He is not coming home tonight.
. Similar to other Romance languages, the negation no accompanies verb movement in inver-
sion contexts:
(i) No
not
vien=e
come=clit.3pl.fem
mia
prt
a
to
casa
home
stasera?
tonight
Are they really not coming home tonight?
. This leaves open why distinctive and non-distinctive clitics occupy different positions rel-
ative to the negation no (see e.g. (51) above). However, there are some indications that this
difference is in fact another consequence of the grammaticalization process in question. In a
study of the grammaticalization of agreement markers from former subject clitics in the history
of Veneto, Poletto (1995) shows that the linear order of clitic, verb and negation changed as a re-
sult of the reanalysis of clitics as realizations of an AgrS head. Thus, while in Renaissance Veneto
the clitics preceded the negation no, they follow no in the present-day language (according to
Poletto, this change took place in the 17th century):
(i) La
she
no
not
vaga
goes
a
on
mio
my
conto.
count
(Renaissance Veneto, Calmo, 79; Poletto 1995: 301)
(ii) No
not
la
she
vien.
come
She does not come. (Veneto; Poletto 1995: 297)
It is thus likely that a similar change in word order accompanied the grammaticalization of new
agreement markers in the history of Vicentino. More precisely, the reanalysis in question was
presumably only possible in contexts where the clitic was adjacent to the verb (in line with the
Word building constraint, cf. Chapters 4 and 5 above). After the reanalysis, the newly developed
agreement markers could only show up in a position adjacent to the verb, that is, to the right
of negation. It seems likely that the change in question proceeded via an intermediate stage
which exhibited variation between pre-neg and post-neg placement of clitics (the latter being in
fact agreement markers). When the change was completed and the former clitics had been fully
established as agreement formatives, only the post-neg placement survived.
. Similar to Piattino (but in contrast to Non-Standard French), the grammaticalization of
gender agreement was made available by the existence of a much more elaborate inventory of
number distinctions in the existing agreement paradigm of Vicentino, which by assumption
renders the gender node visible for grammaticalization processes (see Section 6.4 above).
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.66 (4565-4636)
i| The Rise of Agreement
|. It is a well-known fact that many Northern Italian dialects exhibit a highly syncretic present
subjunctive paradigm where most forms are identical to the relevant present indicative forms
(cf. e.g. Meyer-Lbke 1890: 225; Meyer-Lbke 1894: 184ff.; Rohlfs 1949: 346f.; Savoia 1997: 84).
|8. Recall that the proclitic a is not distinctive and therefore not eligible for the grammatical-
ization process in question.
|. This observation suggests a possible alternative explanation of the change in question ac-
cording to which the reanalysis of the enclitics satised the Blocking Principle by creating a
verb form which is specied for clause type, that is, [+interrogative]. Note that this idea is also
already present in historical comparative grammars of the Romance languages such as Meyer-
Lbke (1894: 367f.) or Rohlfs (1949: 179f.). See also Haiman and Beninc (1992) and Savoia
(1997: 85f.).
,o. However, note that it is possible that grammaticalization processes lead to instances of sec-
ondary exponence, where the shape of a given exponent which is the primary exponent of
another inectional feature such as Tense is inuenced by the feature content of an agreement
morpheme, giving rise to conditioned allomorphy.
,:. A related question arises in connection with the claim (see Chapter 3, Note 62) that only the
highest Agr-morpheme is spelled out if a certain head complex contains two Agr-morphemes
with identical feature content (Kinyalolo 1991; Carstens 2003):
(i) Morphological Economy
In an adjoined structure, Agr on a lower head is inert iff its features are predictable from
Agr on a higher head.
Accordingly, a reanalysis such as (52) giving rise to a verb form which exhibits a prexal agree-
ment marker in addition to the existing sufxal agreement morphology should be possible only
if the newhigher Agr-morpheme is located in a separate X
0
-complex which does not contain the
old Agr-morpheme. Thus, I assume that prexal agreement morphology may also result from a
syntactic conguration where the relevant agreement head does not form an X
0
-complex with
the verb, but is merely linearly adjacent to it (cf. e.g. Embick & Noyer 2001; Julien 2002).
,i. The Swiss RR dialects actually make use of a variety of different morphological means to
satisfy the penultimate stress target. For example, apart from causing deletion of the existing
agreement sufx/theme vowel of the verb, the attachment of a clitic may also affect the clitic
itself, leading to elision of the clitics vowel (e.g., 3sg.masc chanta + el chantal). In other
contexts (often 3sg.neut or 1pl) a clitic simply cannot be used if this would result in a verb form
which violates the penultimate stress target. In Vallader, a semantically empty, stressed sufx
-sch is sometimes added to the verb in addition to the clitic. As a consequence, the word stress
shifts from the stem to -sch, leading to acceptable penultimate stress if a clitic is added (cf.
Haiman 1971: 805). In addition, for reasons which are unclear, some dialects such as Puter ap-
parently tolerate violations of the penultimate stress rule in 3pl contexts (cf. Haiman 1971: 807;
Linder 1987: 82ff.). The individual dialects differ with respect to the use of these mechanisms.
Moreover, the distribution of these mechanisms is inuenced by the stress pattern of the verb
(which is partially dependent on tense and mood), verb class, shape of clitics and a set of other
factors such as interference with other dialects. As a consequence, more often than not, different
strategies appear to exist side by side for a single form in a single dialect (see Haiman 1971 and
in particular Linder 1987: Ch. 8 for details).
,. Here, we might speculate that these forms also serve to realize morphosyntactic features
associated with the C-head (thereby fullling the Blocking Principle). Recall that the new
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.67 (4636-4700)
Chapter 6. Morphological blocking and the rise of agreement i,
agreement formatives are conned to inversion contexts and cannot show up in embedded
clauses. This might be taken to indicate that these inections also serve to mark the distinc-
tion [subordinate], which is associated with C (the fact that the new endings do not show up
in subject-initial clauses follows from the assumption that the verb does not move to C in these
contexts, cf. Chapter 5, Section 5.3.4.1).
,|. In the resulting grammar, deletion of the original agreement ending (realizing Agr-on-T)
is then the result of a morphological rule which guarantees that of two agreement morphemes
(with identical feature content) in a given head complex, only the higher one (here: Agr-on-C)
is subject to Vocabulary Insertion (see Chapter 3, Note 62).
,,. The loss of conjugational allomorphy and agreement distinctions, which characterizes the
secondary endings, is to be attributed to the fact that these endings do not receive stress and are
therefore more likely to undergo phonological reduction than the primary endings.
,o. Some Sutselvan dialects (e.g., the dialect of Flem, cf. Decurtins 1958) show1sg, 3sg -i instead
of -a. Still, what is important is that 1sg and 3sg forms are identical in these dialects as well.
,. The 2nd person form must carry the specication [Auth] to prevent that /-as/ is inserted
as the realization of the feature matrix [+Auth, +PSE] (we-inclusive). Furthermore, note that
the insertion of /-an/ as the realization of 2pl is blocked by the existence of a stronger specied
form /-as/, which realizes two inectional features.
,8. In the case of the 3sg forms, the specication [pl] is necessary to prevent that the exponents
/-al/, /-la/ are inserted in the context of 3pl, where the relevant agreement morpheme also carries
gender specication (recall that by assumption the complete set of the subjects -features are
copied into the agreement morpheme, cf. Halle 1997).
,. Note that the (complete) underspecication of the formative /-in/ results from inspecting
other parts of the paradigm (e.g., the existence of a form which signals unambiguously 1pl).
However, it is a well-known fact that children do not acquire entire paradigms at once. Rather,
the individual formatives which constitute a paradigmare acquired one after the other (the exact
sequence is subject to cross-linguistic variation, cf. e.g. Clark 1998). Accordingly, determining
the exact feature (under-) specication of a given element is a task which can be accomplished
only after all forms that constitute a given paradigm have been acquired. In contrast, the Block-
ing Principle is concerned with a more basic choice which must be decided on earlier during
language acquisition.
oo. These changes are reminiscent of the characteristic properties of creoles, a fact also noted by
Polinsky (p. 86): The syntactic redundancy demonstrated for American Russian is paralleled
by similar tendencies in extended pidgins and early creoles, which points to a more general
correspondence between language loss and pidginization.
o:. Note, however, that Polinsky (2000: 60) expresses some reservations with respect to the latter
criterion, since American Russian is characterized by aberrant pauses.
oi. The special status of short, monosyllabic verbs where agreement distinctions are marked
solely by stem vowel alternations is presumably another factor which played a role in this devel-
opment (cf. Section 6.5.2 above): In this perspective it is not by chance that Kurzverben have
been the rst context in which subject clitics grammaticalized into verb endings. In fact, this
verbal class presents a very reduced word nal morphology, as most grammatical information is
carried by stem vowel alternation rules (cf. Nbling 1995). Therefore, Walser dialects provided
Kurzverben with longer endings, transferring the functional load from the stem to the end of the
word. (Dal Negro 2004: 178f.)
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.68 (4700-4781)
. Presumably, these considerations carry over to the development of new agreement forma-
tives in Cimbrian discussed in Chapter 5, Section 5.2.3.1.
. However, Donohue (2002: 7) notes that the requirement for subjects to be overtly realized
is dependent on person: With rst or second person clitics this restriction is relaxed somewhat,
and sometimes the clitic pronoun alone is sufcient for a grammatical reading (especially with
the singular n and m).
. According to Donohue, agreement via vowel changes is found only in seven verbs (in com-
bination with a proclitic and sometimes consonant changes): fu to fear, fue to see, lu weng to
sleep, le to shave, leng to tell, re to go, and i ri to fall.
. The factors that condition inection of the adjunct nominal (instead of the verb root)
remain somewhat unclear. Donohue (2002: 11) observes that the adjunct nominal remains un-
changed in cases where the proclitic intervenes between the adjunct nominal and the verb stem
(in contrast to l weng to sleep, where the proclitic attaches to the left of the whole complex
verb), as in png l to shoot at someone:
(i) N
I
pe
3sg.fem
png
arrow
n=l.
1sg=release
(*N pe n=png l.)
I shot her.
(ii) Pe
she
n
me
png
arrow
pe=r.
3sg.fem=release.3sg.fem
(*Pe n pe=png l, *Pe n pe=png r)
She shot me.
. The availability of consonant alternation as a means of agreement marking is apparently
restricted by the onset of the verb root: it is found only with verb roots that begin with the
consonants k, h, w, l or with the vowels a, o, oe.
. It has already been noted that agreement marking via changes affecting the onset or stem
vowel is always accompanied by agreement marking via a proclitic. Some verbs display even
more redundant agreement patterns, where subject-verb agreement is marked by up to three
different morphological means. For example, there are verbs such as le shave and leng tell,
order, promise, persuade, which display a combination of vowel and consonant alternations
(plus the obligatory proclitic, cf. Donohue 2002 for details). This gives rise to sentences such
as (i), where subject-verb agreement is realized in three different ways on the verb (marked by
boldface):
(i) Pe
she
yu-pe-p=pe
brother-dat.3sg.fem-poss.3sg.fem=dat.3sg.fem
ta-k=ke
hair-poss.3sg.masc=dat.3sg.masc
pe=r-e.
3sg.fem=3sg.fem-shave.3sg.fem
She shaved her brothers hair. (Donohue 2002: 14)
Donohue shows that the different agreement patterns (i.e., proclitic, proclitic+consonant prex
etc.) found in Skou cannot be correlated with different verb classes (such as e.g. transitive vs
intransitive verbs) or phonological factors. Apparently, then, there are no systematic synchronic
factors/criteria that determine the way a given verb inects for agreement (apart from the fact
that the presence of consonant prexes is limited to a certain set of onsets, see Note 67 above).
. Cf. Donohue (2002: 30): [...] verbs [in Skou] can begin with a larger range of consonants
than k h w l a o or oe [...] additionally, p j m f h y are found, but these verbs do not inect by
prex. Historically it is likely that these forms showed inection, just as in the Eastern languages.
JB[v.20020404] Prn:2/08/2005; 15:55 F: LA8106.tex / p.69 (4781-4787)
Extreme simplicationin Skou resulted in the complete loss of the inectional systemwith these
verbs; there is no longer any evidence in the paradigm for any inection. Taking these verbs into
account, assuming seven distinct paradigm sets as in the majority of the Eastern languages, we
arrive at a gure of 79% for loss of contrast in verbal paradigms.
JB[v.20020404] Prn:2/08/2005; 15:57 F: LA8107.tex / p.1 (44-121)
Chapter 7
Concluding summary
This study has investigated the historical paths leading from pronominal clitics to
markers of verbal agreement, focusing on the rise of subject-verb agreement. I have
argued that the grammaticalization process in question cannot be reduced to a single
scenario such as the reanalysis of resumptive clitics in NP-detachment construc-
tions which proceeds uniformly across languages (in contrast to claims widely held
in the literature). Instead, I have tried to show that the reanalysis of clitic pronouns
as agreement markers is shaped by a variety of different but interwoven syntactic and
morphological factors which derive from the way syntactic agreement is established
in natural languages and the way language change is anchored in the workings of lan-
guage acquisition. Accordingly, the change in question can be triggered in a multitude
of historical scenarios which still share a set of clearly identiable common properties
(such as structural simplication, adjacency between the nite verb and the pronoun,
and the presence of an alternative carrier of the -role previously assigned to the pro-
noun). These underlying similarities have been shown to result from the universal
nature of the restrictions that the reanalysis of clitics is subject to. A subset of the
relevant syntactic contexts has been explored in more detail, focusing on historical
developments affecting C-oriented clitics which either attach to the nite verb in V2
languages or occupy clausal second position in languages such as Uto-Aztecan or Mon-
golian. A central claim put forward in this book is that the grammaticalization of
agreement markers from pronominal elements is shaped by a certain division of labor
between syntax and morphology in which the workings of the syntactic component
dene a set of necessary conditions which a reanalysis must satisfy, while the ultimate
trigger for the change in question is morphological in nature. More specically, I have
argued that the reanalysis of pronominal clitics (and the acquisition of inectional
morphology in general) is governed by morphological blocking effects which trig-
ger the grammaticalization of new inectional markers in contexts where the existing
morphology is defective, that is, signals fewer distinctions than the potential new for-
matives. In the following, the content of the individual chapters is briey recapitulated,
highlighting the thread of the individual arguments put forward in this study.
In the introductory Chapter 1, I introduced the empirical phenomenon under
consideration and gave an overview of general characteristics of the transition from
pronominal elements to markers of verbal agreement. Based on a critical discussion
of previous analyses, I highlighted a set of descriptive and explanatory issues which an
adequate theoretical account of the grammaticalization of agreement markers should
JB[v.20020404] Prn:2/08/2005; 15:57 F: LA8107.tex / p.2 (121-182)
address, including the pioneering role of 1st and 2nd person in the development of
agreement markers and the fact that cross-linguistically, verbal agreement is preferably
realized by sufxes.
In Chapter 2, I introduced the basic theoretical assumptions advocated in this
book, including an outline of minimalist syntax (i.e., the probe-goal mechanism) and
key properties of Distributed Morphology. In addition, I provided an overview of gen-
erative approaches to language change. With respect to the latter, I focused on the
phenomenon of grammaticalization, claiming that this type of change is to be ana-
lyzed as a historical process in which a phonological exponent formerly realizing a
lexical head (or l-morpheme) turns into the realization of a functional head (or f -
morpheme). It was demonstrated that this model provides a unied explanation for
a number of surface characteristics of grammaticalization phenomena, attributing the
latter to universal properties of functional categories. In this context, however, I already
hinted at the special status of agreement, calling into doubt whether agreement should
be analyzed on a par with categories such as T or D, which are commonly assumed to
belong to a universally present inventory of functional heads.
The peculiar properties of agreement were explored in some more detail in Chap-
ter 3, where I developed a synchronic analysis of predicate-argument agreement. The
central claims of this approach can be summarized as follows. I argued that agree-
ment morphemes do not head their own projection in the syntax. Rather, they are
parasitic on other contentful functional heads such as C, T, or , with which they
can combine in two different ways. First, the agreement morpheme may attach to its
functional host prior to the insertion of that host into the syntactic derivation (canon-
ical predicate-argument agreement). Second, the agreement morpheme may be added
post-syntactically as a dissociated Agr-morpheme at the level of Morphological Struc-
ture. I claimed that Agr-morphemes present in the syntactic derivation initiate an
Agree-operation, accessing the closest interpretable -set to value their feature con-
tent and that this operation may in principle be independent from Case licensing,
contra Chomsky (2000). In contrast, dissociated Agr-morphemes are parasitic on an
agreement relation established in the syntax. That is, they are licensed under structural
adjacency with an Agr-morpheme that has been licensed in the syntax. As a conse-
quence, the insertion of dissociated morphemes typically gives rise to instances of
multiple agreement where the arguments -features are reected in more than one
location across the clause. It has been shown that this approach makes available a new
analysis of complementizer agreement in Germanic which avoids the shortcomings of
previous analyses and successfully captures the PF-avor of this phenomenon (ad-
jacency effects, sensitivity to elision of the nite verb). In addition, I have taken up
the question of how agreement morphemes are attached to the verb stem, defending
the traditional idea that afx order is determined by lexical properties of individual
Vocabulary items (i.e., the prex/sufx distinction), in addition to structural restric-
tions imposed by the hierarchical relations created in the syntax. Here, I argued that
the effects of the Mirror Principle can be derived if it is assumed that head adjunction
structures are processed in a bottom-up fashion during Vocabulary Insertion, creating
JB[v.20020404] Prn:2/08/2005; 15:57 F: LA8107.tex / p.3 (182-250)
Chapter 7. Concluding summary
the linear ordering of morphemes. Furthermore, I pointed out that the workings of
the mapping procedure can also account for the fact that agreement markers exhibit
more ordering possibilities than other inectional afxes. This explanation crucially
relies on the special phrase-structural status of agreement morphemes as adjuncts to
other inectional heads, introducing a point of variation in the linearization of head
complexes. As a consequence, an Agr-morpheme adjoined to T may give rise to either
V+T+Agr or V+Agr+T in individual languages.
The diachronic implications of this approach to predicate-argument agreement
were explored in greater detail in Chapters 46. In Chapter 4, I demonstrated that the
specic analysis advocated in the previous chapter makes certain predictions with re-
spect to the syntactic environments where a reanalysis of clitic pronouns as exponents
of agreement morphemes is possible. In addition, I introduced a set of criteria for dis-
tinguishing between pronominal elements and markers of verbal agreement, arguing
that these criteria can also be conceived of as cues which determine the acquisition of
agreement markers during language acquisition. On these assumptions, I developed a
set of necessary requirements that the grammaticalization of agreement markers must
satisfy. More specically, I argued that the change in question is only possible if (i) it
preserves the predicates argument structure, (ii) provides a means to value the feature
content of the resulting agreement morpheme, (iii) warrants that (bound) agreement
morphemes combine with the verb prior to Vocabulary Insertion, and (iv) satises
economy considerations which require that the resulting structure is (derivationally)
less complex than the target structure. Based on these assumptions, I presented a (pre-
sumably non-exhaustive) set of syntactic scenarios in which the reanalysis of clitic
pronouns as agreement markers is licensed. In addition, I claimed that the sufxing
preference should receive a diachronic explanation, arguing that the grammaticaliza-
tion process in question is more likely to be triggered in syntactic environments which
give rise to sufxal agreement morphology.
In Chapter 5, I took a closer look at a subset of the scenarios outlined in Chap-
ter 4, focusing on syntactic aspects of the reanalysis of C-oriented clitics in languages
with (Bavarian, Cimbrian, Walser German, Rhaeto-Romance) and without the V2
property (Uto-Aztecan and Mongolian). I demonstrated that in the former, the V2
property played a key role in this change by creating adjacency between subject clitics
and the nite verb. In addition, I showed that we still must distinguish between two
different historical scenarios where new agreement markers may arise in V2 languages.
First, in Bavarian, Walser German, Cimbrian and older stages of Rhaeto-Romance,
the conversion of subject clitics into agreement markers led to structures where the
subjects -role is assigned to a referential pro, either exploiting pro-drop properties
already available in the target grammar or creating new instances of pro-drop. Second,
I demonstrated that in some Swiss Rhaeto-Romance dialects (in particular Sutselvan),
the reanalysis targets subject clitics in clitic doubling structures, resulting in a congu-
ration where the relevant -role is assigned to the former double (which was originally
added for reasons of emphasis). I further argued that in both scenarios, the creation
of new agreement markers proceeded via an initial stage where a C-oriented clitic is
JB[v.20020404] Prn:2/08/2005; 15:57 F: LA8107.tex / p.4 (250-290)
reanalyzed as the realization of a dissociated agreement morpheme on C, giving rise
to instances of multiple agreement (still reected by complementizer agreement in
present-day Bavarian). In a subsequent development, the exponents of Agr-on-C may
then replace the older exponents of canonical subject-verb agreement (associated
with Agr-on-T). The theoretical insights gained from the discussion of V2 languages
can be summarized as follows. First, there is a systematic link between the pro-drop
property and the rise of new agreement markers, which can be taken to suggest that
historically, the former is a by-product of the latter. Second, the observation that ex-
ponents of a higher functional morpheme (Agr-on-C) may turn into realizations of
a lower functional morpheme (Agr-on-T) contradicts the claim that grammaticaliza-
tion always leads to new exponents of higher functional heads (cf. Roberts & Roussou
2003). Third, at least in languages which already exhibited verbal agreement mor-
phology prior to the reanalysis in question, the grammaticalization of new agreement
formatives proceeds via an initial stage where a clitic is reanalyzed as a dissociated
Agr-morpheme which is parasitic on another Agr-morpheme that is valued in the syn-
tactic derivation. Accordingly, the grammaticalization path in questioncan be depicted
as in (1).
(1) free pronoun weak pronoun clitic pronoun
(dissociated Agr-morpheme) syntactic Agr-morpheme
In the remainder of Chapter 5, I discussed two examples in which C-oriented clitics
were directly reanalyzed as realizations of a syntactic Agr-morpheme (i.e., Agr-on-
T) in languages which previously lacked agreement markers. First, I showed that a set
of Uto-Aztecan languages exhibits various stages of a historical process where second
position clitics (in doubling congurations) turn into prexal agreement markers in
a set of SVO languages. Second, I discussed an alternative historical path which may
lead to sufxal agreement morphology in true SOVlanguages, arguing that in a set of
Mongolian languages, C-oriented clitics were reanalyzed as sufxal agreement markers
in a structural conguration where the TP is fronted into the specier of CP, creating
the necessary adjacency between the clitic (adjoined to C) and the nite verb which
occurs at the right end of the fronted TP.
In Chapter 6, I focused on morphological factors which shape the grammatical-
ization of agreement markers, starting out from the observation that across languages,
the grammaticalization of new agreement formatives does not proceed in a random
fashion, but rather affects only contexts where the existing verbal morphology is non-
distinctive. I claimed that this observation can be explained if it is assumed that the
acquisition of inectional morphology is guided by blocking effects which prefer
new verbal agreement formatives to be more specic than the existing morphology.
Framed in a realizational model of grammar (i.e., Distributed Morphology), this idea
is expressed by the so-called Blocking Principle:
JB[v.20020404] Prn:2/08/2005; 15:57 F: LA8107.tex / p.5 (290-358)
Chapter 7. Concluding summary
in the lexicon.
By assumption, the Blocking Principle operates during language acquisition as an
economy principle which warrants an optimal and non-redundant Lexicon (and
paradigm) structure. Similar to structural economy principles which guide the ac-
quisition of syntactic properties (cf. Clark & Roberts 1993), the Blocking Principle is
invoked only if the trigger experience is not sufcient for determining which Vocabu-
lary item realizes a certain inectional head, as in cases where the learner is confronted
both with an existing, underspecied exponent of a certain Agr-morpheme and a
pronominal clitic which is stronger specied and fullls all necessary conditions for
being reanalyzed as an agreement formative. In turn, I demonstrated that the work-
ings of the Blocking Principle can be detected in the development of new agreement
formatives in languages such as Bavarian, Non-Standard French, Northern Italian di-
alects, Rhaeto-Romance and the Papuan language Skou. Furthermore, in the case of
Bavarian, I argued that the peculiar person/number restrictions on pro-drop and com-
plementizer agreement observed in Chapter 5 should receive a diachronic explanation,
in the sense that these restrictions reect the contexts where the grammaticalization of
new forms was licensed by the Blocking Principle. In many instances, we observed that
the reanalysis initially affects clitics only in a restricted context in which the relevant
existing verb forms are defective (e.g., a certain verb class, tense or mood) before the
change gains a wider distribution. Here, it was argued that only the initial reanaly-
sis which gives rise to a new agreement formative proceeds in accordance with the
Blocking Principle, whereas the gradual expansion to other contexts was attributed to
the workings of analogical change. In addition, I argued that the analysis presented
in this chapter is further supported by the fact that the Blocking Principle provides
a new explanation for the pioneering role of 1st and 2nd person in the development
of predicate-argument agreement across languages (assuming that 3rd person forms,
including potential new agreement formatives, lack person features and are therefore
inherently underspecied).
As noted repeatedly throughout this work, the present study focused only on a
small portion of the vast array of diachronic phenomena which involve the creation
of new agreement markers. At least in part, the decision to concentrate on two major
issues (syntactic aspects of the reanalysis of C-oriented subject clitics and the work-
ings of morphological blocking in the rise of agreement) was governed simply by
practical considerations. An in-depth treatment of the complete set of grammatical-
ization processes leading to new agreement formatives is a task simply too ambitious
to be accomplished within a single monograph. So I decided to remain silent about
other major topics such as the grammaticalization of object or wh-agreement (cf. e.g.
Chung 1998 on the latter), the rise of DP-internal (head-modier) agreement (cf.
e.g. Lehmann 2002), the development of ergative/absolutive agreement (cf. Ander-
son 1977, 1980; Givn 1984), alternative paths to agreement which do not involve
JB[v.20020404] Prn:2/08/2005; 15:57 F: LA8107.tex / p.6 (358-381)
pronominal elements (such as the reinterpretation of other pieces of inection, cf. e.g.
Chafe 1977 on the 3rd person subject prexes of Iroquoian, see Siewierska 2004 for
an overview), or the connection between the rise of agreement and other syntactic
changes, each of which justies a monograph in itself.
1
Still, I reckon that the analy-
sis developed in the previous chapters is not limited to the data set discussed in this
book and can be extended to at least some of the above mentioned phenomena. The
factors and principles proposed in this work such as the Blocking Principle, Preserva-
tion of argument structure or Structural simplication are not bound to a single surface
phenomenon (such as the reanalysis of subject clitics). Rather, they derive from (and
model) deeper properties of grammar and are therefore expected to carry over to other
changes creating newinectional markers across languages. In this way, I hope that the
insights gained in this work add to our understanding of the processes which underlie
the rise of agreement and the phenomenon of grammaticalization in general.
Note
. The latter includes such intriguing topics as the link between the rise of agreement and the
development of free word order (cf. e.g. Mithun 1991), the development of verb movement
(Alexiadou & Fanselow 2001, 2002) or changes affecting the syntactic alignment (nomina-
tive/accusative vs. absolutive/ergative) of a language. For example, it appears that in a number
of ergative/absolutive languages, the development of person agreement on the verb affects the
overall organization of grammar, giving rise to nominative/accusative patterns (cf. e.g. Harris
1994; Schulze 1998 on the Caucasian family, Cysouw 2003b on a set of Austronesian languages
spoken in Sulawesi). Thus, the new person markers do not follow the ergative/absolutive pat-
tern of the existing agreement markers (such as noun class agreement marking in the Caucasian
languages). Instead, they always signal agreement with the subject DP of both intransitives and
transitives.
JB[v.20020404] Prn:8/09/2005; 13:19 F: LA81RE.tex / p.1 (44-168)
References
Abney, Steven (1987). The English Noun Phrase in its Sentential Aspect. Doctoral dissertation,
MIT.
Abraham, Werner (1995). Deutsche Syntax im Sprachenvergleich. Tbingen: Gunter Narr.
Abraham, Werner (2003). Autonomous and non-autonomous components of grammat-
icalization: Economy criteria in the emergence of German negation. Sprachtypologie und
Universalienforschung, 56 (4), 325365.
Abraham, Werner (2004). The grammaticalization of the innitival preposition toward a
theory of grammaticalizing reanalysis. Journal of Comparative Germanic Linguistics, 7,
111170.
Ackema, Peter, & Ad Neeleman (2003). Context-sensitive Spell-out. Natural Language and
Linguistic Theory, 21, 681735.
Ackema, Peter, & Ad Neeleman (2004). Beyond Morphology. Interface Conditions on Word
Formation. Oxford: Oxford University Press.
Albright, Adam C. (2002). The Identication of Bases in Morphological Paradigms. Doctoral
dissertation, UCLA.
Alexiadou, Artemis, & Elena Anagnostopoulou (1998). Parametrizing AGR: Word order, V-
movement and EPP-checking. Natural Language and Linguistic Theory, 16, 491539.
Alexiadou, Artemis, & Gisbert Fanselow (2001). Laws of diachrony as a source for syntactic
generalisations: The case of V to I. GLOW Newsletter, 46, 5758.
Alexiadou, Artemis, & Gisbert Fanselow (2002). On the correlation between morphology and
syntax: The case of V-to-I. In J.-W. Zwart & W. Abraham (Eds.), Studies in Comparative
Germanic Syntax (pp. 219242). Amsterdam: John Benjamins.
Allen, Cynthia (1995). Case Marking and Reanalysis: Grammatical Relations from Old to Early
Modern English. Oxford: Clarendon.
Allen, Cynthia (1997). The origins of the group genitive in English. Transactions of the
Philological Society, 95 (1), 111131.
Altmann, Hans (1984). Das System der enklitischen Personalpronomina in einer mittel-
bairischen Mundart. Zeitschrift fr Dialektologie und Linguistik, 51 (2), 191211.
Andersen, Torben (1988). Ergativity in Pari, a Nilotic OVS language. Lingua, 75, 289324.
Anderson, Stephen R. (1977). On mechanisms by which languages become ergative. In C. N. Li
(Ed.), Mechanisms of Syntactic Change (pp. 317363). Austin: University of Texas Press.
Anderson, Stephen R. (1980). On the development of morphology from syntax. In J. Fisiak
(Ed.), Historical Morphology (pp. 5169). Berlin: Mouton de Gruyter.
Anderson, Stephen R. (1986). Disjunctive ordering in inectional morphology. Natural
Language and Linguistic Theory, 4, 131.
Anderson, Stephen R. (1992). A-morphous Morphology. Cambridge: Cambridge University
Press.
Ariel, Mira (1990). Accessing NP Antecedents. London: Routledge [Croom Helm Linguistics
Series].
Ariel, Mira (1998). Three grammaticalization paths for the development of person verbal
agreement in Hebrew. In J.-P. Koenig (Ed.), Discourse and Cognition: Bridging the Gap
(pp. 93111). Stanford, CSLI.
Ariel, Mira (2000). The development of person agreement markers: From pronoun to higher
accessibility markers. In S. Kemmer & M. Barlow (Eds.), Usage-based Models of Language
(pp. 197261). Stanford, CSLI.
Aronoff, Mark (1976). Word Formation in Generative Grammar. Cambridge, MA: The MIT
Press.
Arregi, Karlos (1999). Person and number inection in Basque. In K. Arregi, B. Bruening,
C. Krause, & V. Lin (Eds.), MIT Working Papers in Linguistics 33: Papers on Morphology
and Syntax, Cycle One (pp. 229264). Cambridge, MA: Department of Linguistics and
Philosophy, MIT.
Ashby, William (1977). Clitic Inection in French. An Historical Perspective. Amsterdam: Rodopi.
Auger, Julie (1993). More evidence for verbal agreement marking in Colloquial French. In
W. J. Ashby, M. Mithun, G. Perissinotto, & E. Raposo (Eds.), Linguistic Perspectives on the
Romance Languages (pp. 17798). Amsterdam: John Benjamins.
Auger, Julie (1994). Pronominal Clitics in Quebec Colloquial French: A Morphological Analysis.
Doctoral dissertation, University of Pennsylvania.
Baker, Mark C. (1985). The Mirror Principle and morphosyntactic explanation. Linguistic
Inquiry, 16 (3), 373415.
Baker, Mark C. (1988). Incorporation. A Theory of Grammatical Function Changing. Chicago:
Chicago University Press.
Baker, Mark C. (1996). The Polysynthesis Parameter. Oxford: Oxford University Press.
Baker, Mark C. (2002). Building and merging, not checking: The non-existence of (Aux)-S-V-O
languages. Linguistic Inquiry, 33 (2), 321328.
Baker, Mark. C., & Jonathan Bobaljik (2002). Introduction to Morphology. Ms., Rutgers and
McGill.
Barlow, Michael (1992). A Situated Theory of Agreement. New York, NY: Garland.
Battye, Adrian, & Roberts, Ian (Eds.). (1995). Clause Structure and Language Change. Oxford:
Oxford University Press.
Bayer, Josef (1984). COMP in Bavarian syntax. The Linguistic Review, 3, 209274.
Beard, Robert (1988). The separation of derivation and afxation: Toward a Lexeme-Morpheme
Base Morphology. Quarderni di semantica, 5, 277287.
Beard, Robert (1995). Lexeme-Morpheme Base Morphology. A General Theory of Inection and
Word Formation. Albany NY: SUNY Press.
Bech, Gunnar (1955). Studien zum deutschen Verbum innitum. Tbingen: Niemeyer.
Beckman, Jill N. (Ed.). (1995). Proceedings of NELS 25. Vol. 2: Papers from the Workshops on
Language Acquisition and Language Change. University of Pennsylvania.
Bejar, Susana (2003). Phi-syntax: A Theory of Agreement. Doctoral dissertation, University of
Toronto.
Belletti, Adriana (1990). Generalized Verb Movement: Aspects of Verb Syntax. Turin: Rosenberg
and Sellier.
Belletti, Adriana (2001). Agreement projections. In M. Baltin &C. Collins (Eds.), The Handbook
of Contemporary Syntactic Theory (pp. 483510). Oxford: Blackwell.
References
Bennis, Hans, & Liliane Haegeman (1984). On the status of agreement and relative clauses in
West Flemish. In W. de Geest & Y. Putseys (Eds.), Sentential complementation (pp. 3355).
Dordecht: Foris.
Benveniste, Emile (1968). Mutations of Linguistic Categories. In W. Lehmann & Y. Malkiel
(Eds.), Directions for Historical Linguistics: A Symposium (pp. 8394). Austin: University
of Texas Press.
Benveniste, Emile (1972). Problmes de linguistique gnrale. Paris: Editions Gallimard.
Berwick, Robert C. (1985). The Acquisition of Syntactic Knowledge. Cambridge, MA: The MIT
Press.
Besten, Hans den (1982). On the interaction of root transformations and lexical deletive rules.
In W. Abraham(Ed.), On the Formal Syntax of the Westgermania (pp. 47132). Amsterdam:
John Benjamins.
Bickerton, Derek (1984). The language bioprogram hypothesis. Behavioral and Brain Sciences,
7, 173221.
Bickerton, Derek (1998). How to acquire language without positive evidence: What
acquisitionists can learn from creoles. In M. DeGraff (Ed.), Language Creation and
Language Change: Creolization, Diachrony and Development (pp. 4974). Cambridge, MA:
The MIT Press.
Bittner, Andreas (2004). Wenn Spracherwerb Sprachwandel wr. Paper presented at the
University of Frankfurt, January 14th, 2004.
Blood, Cindy (1992). Subject-verb agreement in Kisar. NUSA. Linguistics Studies of Indonesia
and Other Languages in Indonesia, Vol. 34 (pp. 121). Jakarta: Badan Penyelenggara NUSA.
Bobaljik, Jonathan (1994). What does adjacency do? In H. Harley & C. Phillips (Eds.),
MIT Working Papers in Linguistics 22: The Morphology-Syntax Connection (pp. 132).
Cambridge, MA: Department of Linguistics and Philosophy, MIT.
Bobaljik, Jonathan (1995). Morphosyntax: The Syntax of Verbal Inection. Doctoral
dissertation, MIT.
Bobaljik, Jonathan (2002). A-chains at the PF interface. Natural Language and Linguistic Theory,
20, 197267.
Bobaljik, Jonathan (2003). Realizing Germanic inection: Why morphology does not drive
syntax. Journal of Comparative Germanic Linguistics, 6 (2), 129167.
Bobaljik, Jonathan (2005). Wheres ? Agreement as a postsyntactic operation. Ms., University
of Connecticut, Storrs.
Bobaljik, Jonathan, & Samuel Brown (1997). Interarboreal operations: Head movement and the
Extension Requirement. Linguistic Inquiry, 28 (2), 34556.
Bobaljik, Jonathan, &Diane Jonas (1996). Subject positions and the role of TP. Linguistic Inquiry,
27 (2), 195236.
Bobaljik, Jonathan, & H. Thrinsson (1998). Two heads arent always better than one. Syntax, 1,
3771.
Bobaljik, Jonathan, & Susi Wurmbrand (2004). The domain of agreement. To appear in Natural
Language and Linguistic Theory.
Bonet, Eulalia (1991). Morphology after Syntax. Pronominal Clitics in Romance. Doctoral
dissertation, MIT.
Bopp, Franz (1816). ber das Conjugationssystem der Sanskritsprache in Vergleichung mit jenem
der griechischen, lateinischen, persischen und germanischen Sprache. Frankfurt am Main:
Andreische Buchhandlung.
Borer, Hagit (1984). Parametric Syntax. Dordrecht: Foris.
Borer, Hagit (1998). Morphology and Syntax. In A. Spencer & A. Zwicky (Eds.), The Handbook
of Morphology (pp. 151189). Oxford: Blackwell.
Brandi, Luciana, &Patrizia Cordin (1989). Two Italiandialects and the null subject parameter. In
O. Jaeggli & K. Sar (Eds.), The Null Subject Parameter (pp. 111142). Dordrecht, Kluwer.
Branigan, Phil, & Marguerite MacKenzie (2002). Altruism, A-movement and object agreement
in Innu-aimn. Linguistic Inquiry, 33, 385407.
Braune, Wilhelm (1950). Althochdeutsche Grammatik (7th edition). Halle: Niemeyer.
Bresnan, Joan (1973). The syntax of the comparative clause construction in English. Linguistic
Inquiry, 4, 275343.
Bresnan Joan, & Sam A. Mchombo (1987). Topic, pronoun, and agreement in Chichewa.
Language, 63, 741782.
Brinkmann, Hennig (1931). Sprachwandel und Sprachbewegung in althochdeutscher Zeit. Jena.
Brugmann, Karl (1916). Vergleichende Laut-, Stammbildungs- und Flexionslehre der indo-
germanischen Sprachen (2nd edition). Vol. 2/3: Lehre von den Wortformen und ihrem
Gebrauch: Verbum nitum. Strassburg: Trbner.
Bruening, Benjamin (2001). Syntax at the Edge: Cross-clausal Phenomena and the Syntax of
Passamaquoddy. Doctoral dissertation, MIT.
Burenhult, Niclas (2002). A Grammar of Jahai. Lund: Department of Linguistics and Phonetics,
Lund University.
Bybee, Joan L. (1985). Morphology. Amsterdam: John Benjamins.
Bybee, Joan L., WilliamPagliuca, & Revere Perkins (1990). On the asymmetries in the afxation
of grammatical material. In W. Croft, K. Denning & S. Kemmer (Eds.), Studies in Typology
and Diachrony (pp. 142). Amsterdam: John Benjamins.
Bybee, Joan L., & Dan I. Slobin (1982). Rules and schemas in the development and use of the
English past tense. Language, 58, 265289.
Campbell, Lyle (1991). Some grammaticalization changes in Estonian. In E. C. Traugott &
B. Heine (Eds.), Approaches to Grammaticalization, Vol. I: Theoretical and Methodological
Issues (pp. 28599). Amsterdam: John Benjamins.
Cardinaletti, Anna, & Michal Starke (1999). The typology of structural deciency: A case study
of the three classes of pronouns. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe
(pp. 145232). Berlin: Mouton de Gruyter.
Carstens, Vicki (2003). Rethinking complementizer agreement: Agree with a Case-checked goal.
Linguistic Inquiry, 34 (3), 393412.
Cecchetto, Carlo (2000). Doubling structures and reconstruction. Probus, 12, 93126.
Chafe, Wallace L. (1977). The evolution of third person verb agreement in the Iroquoian
Languages. In C. N. Li (Ed.), Mechanisms of Syntactic Change (pp. 493524). Austin:
University of Texas Press.
Chomsky, Noam (1957). Syntactic Structures. The Hague: Mouton.
Chomsky, Noam (1965). Aspects of the Theory of Syntax. Cambridge, MA: The MIT Press.
Chomsky, Noam (1981). Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam (1982). Some Concepts and Consequences of the Theory of Government and
Binding. Cambridge, MA: The MIT Press.
Chomsky, Noam (1986a). Knowledge of Language: Its Nature, Origin and Use. New York, NY:
Praeger.
Chomsky, Noam (1986b). Barriers. Cambridge, MA: The MIT Press.
Chomsky, Noam (1991). Some notes on economy of derivations and representations. In
R. Freidin (Ed.), Principles and Parameters in Comparative Grammar (pp. 417454).
Cambridge, MA: The MIT Press.
References
Chomsky, Noam (1993). A minimalist program for linguistic theory. In K. Hale & S. J. Keyser
(Eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger
(pp. 152). Cambridge, MA: The MIT Press.
Chomsky, Noam (1995). The Minimalist Program. Cambridge, MA: The MIT Press.
Chomsky, Noam (2000). Minimalist inquiries: The framework. In R. Martin, D. Michaels, &
J. Uriagereka (Eds.), Step by Step (pp. 89155). Cambridge, MA: The MIT Press.
Chomsky, Noam (2001a). Derivation by phase. In M. Kenstowicz (Ed.), Ken Hale. A Life in
Language (pp. 152). Cambridge, MA: The MIT Press.
Chomsky, Noam (2001b). Beyond explanatory adequacy. MIT Occasional Papers in Linguistics,
20, 128. Cambridge, MA: Department of Linguistics and Philosophy, MIT.
Chomsky, Noam (2002). On Nature and Language. Cambridge: Cambridge University Press.
Chomsky, Noam (2005). On phases. Ms., MIT.
Chomsky, Noam, & Howard Lasnik (1993). Principles and Parameters theory. In J. Jacobs,
A. von Stechow, W. Sternefeld, & T. Vennemann (Eds.), Syntax: An International Handbook
of Contemporary Research (pp. 496569). Berlin: Walter de Gruyter.
Chung, Sandra (1998). The Design of Agreement. Evidence from Chamorro. Chicago: Chicago
University Press.
Cinque, Guglielmo (1999). Adverbs and Functional Heads: A Cross-linguistic Perspective. Oxford:
Clark, Eve V. (1998). The acquisition of morphology. In Andrew J. Spencer & Arnold M. Zwicky
(Eds.), The Handbook of Morphology (pp. 374389). Oxford: Blackwell.
Clark, Robin, & Ian Roberts (1993). A computational model of language learnability and
language change. Linguistic Inquiry, 24, 299345.
Collins, Chris (1997). Local Economy. Cambridge, MA: The MIT Press.
Comrie, Bernhard (1978). Ergativity. In W. P. Lehmann (Ed.), Syntactic Typology: Studies in the
Phenomenology of Language (pp. 329394). Austin, TX: University of Texas Press.
Comrie, Bernard (1980). Morphology and word order reconstruction: Problems and prospects.
In J. Fisiak (Ed.), Historical Morphology (pp. 8396). The Hague: Mouton.
Comrie, Bernard (2003). When agreement gets trigger-happy. Transactions of the Philological
Society, 101 (2), 313337.
Corbett, Greville (1991). Gender. Cambridge: Cambridge University Press.
Corbett, Greville (1995). Agreement. In J. Jacobs, A. von Stechow, W. Sternefeld, & T.
Vennemann (Eds.), Syntax: An International Handbook of Contemporary Research Vol. 2
(pp. 12351244). Berlin: Walter de Gruyter.
Corbett, Greville (1998). Morphology and agreement. In A. Zwicky & A. Spencer (Eds.), The
Handbook of Morphology (pp. 191205). Oxford: Blackwell.
Corbett, Greville (2000). Number. Cambridge: Cambridge University Press.
Corbett, Greville (2003). Agreement: The range of the phenomenon and the principles of the
Surrey Database of Agreement. Transactions of the Philological Society, 101 (2), 155202.
Craig, Colette Grinevald (1977). The Structure of Jacaltec. Austin, TX: University of Texas Press.
Cutler, Anne, John A. Hawkins, & Gary Gilligan (1985). The sufxing preference: A processing
explanation. Linguistics, 23, 723758.
Cysouw, Michael (2003a). The Paradigmatic Structure of Person Marking. Oxford: Oxford
University Press.
Cysouw, Michael (2003b). From pronouns to agreement (and back again). Paper presented at
the symposium Grammatikalisierung: Wege von der Performanz zur Kompetenz. FU Berlin,
28/29.11.2003.
Dal Negro, Silvia (2004). The Decay of a Language. The Case of a German Dialect in the Italian
Alps. Frankfurt: Peter Lang.
Decurtins, Alexi (1958). Zur Morphologie der unregelmigen Verben im Bndnerromanischen.
Historisch-deskriptive Studie mit besonderer Bercksichtigung des Sur- und Sutselvischen
[Romanica Helvetica 62]. Bern: A. Francke.
Dimmendaal, Gerrit Jan (1983). The Turkana Language. Dordrecht: Foris.
Dixon, R. M. W. (1980). The Languages of Australia. Cambridge: Cambridge University Press.
Dixon, R. M. W. (2002). Australian Languages. Their Nature and Development. Cambridge:
Cambridge University Press.
Donohue, Mark (2002). A most agreeable language. Ms., University of Sydney.
Dresher, Elan (1999). Charting the learning path: Cues to parameter setting. Linguistic Inquiry,
30 (1), 2767.
Dresher, Elan, & Jonathan Kaye (1990). A computational learning model for metrical phono-
logy. Cognition, 34, 137195.
Ebneter, Theodor (1994). Syntax des gesprochenen Rtoromanischen. Tbingen: Niemeyer.
Ellegrd, Alvar (1953). The Auxiliary Do: The Establishment and Regulation of its Use in English.
Stockholm: Almquist and Wiksell.
Embick, David (1997). Voice and the Interfaces of Syntax. Doctoral dissertation, University of
Pennsylvania.
Embick, David, & Rolf Noyer (2001). Movement operations after syntax. Linguistic Inquiry,
32(4), 555595.
Emonds, Joseph (1978). The verbal complex VV in French. Linguistic Inquiry, 21, 4977.
Falk, Cecilia (1993). Non-referential Subjects in the History of Swedish. Doctoral dissertation,
University of Lund.
Fintel, Kai von (1995). The formal semantics of grammaticalization. In J. Beckman (Ed.),
Proceedings of NELS 25: Papers from the Workshops on Language Acquisition & Language
Change (pp. 175189). Amherst, MA: GLSA.
Fodor, Janet (1998). Unambiguous triggers. Linguistic Inquiry, 29 (1), 136.
Frampton, John (2002). Syncretism, impoverishment and the structure of person features. In
M. Andronis, M. Debenport, A. Pycha & K. Yoshimura (Eds.), Proceedings of CLS 38 (pp.
207222). Chicago: Chicago Linguistic Society.
Friedemann, Marc-Ariel (1995). Sujets Syntaxiques: Positions, Inversions et pro. Doctoral
dissertation, Universit de Genve.
Friedmann, Naama, & Yosef Grodzinsky (2000). Split inection in neurolinguistics. In M.-
A. Friedemann & L. Rizzi (Eds.), The Acquisition of Syntax: Studies in Comparative
Developmental Linguistics (pp. 84104). Geneva: Longman.
Fukui, Naoki (1986). A Theory of Category Projection and Its Application. Doctoral
dissertation, MIT.
Fu, Eric (2003). On the historical core of V2 in Germanic. Nordic Journal of Linguistics, 26 (2),
195231.
Fu, Eric, & Carola Trips (2002). Variation and change in Old and Middle English. On the
validity of the Double Base Hypothesis. Journal of Comparative Germanic Linguistics, 4,
171224.
Gabelentz, Georg von der (1891). Die Sprachwissenschaft. Ihre Aufgaben, Methoden, und
bisherigen Ergebnisse. Leipzig: Weigel.
Gartner, Theodor (1883). Rtoromanische Grammatik. Heilbronn: Henninger.
Gartner, Theodor (1910). Handbuch der rtoromanischen Sprache und Literatur. Halle:
Niemeyer.
References
Gelderen, Elly van (1993). The Rise of Functional Categories. Amsterdam: John Benjamins.
Gelderen, Elly van (2001). Modal particles in Germanic. Folia Linguistica Historica, 22 (12),
31033.
Gelderen, Elly van (2004a). Economy, innovation, and prescriptivism: From spec to head and
head to head. Journal of Comparative Germanic Linguistics, 7, 5998.
Gelderen, Elly van (2004b). Grammaticalization as Economy. Amsterdam: John Benjamins.
Gerlach, Birgit (2001). Optimality aspects of clitic doubling in Romance. In P. de Carvalho,
L. Labrune, F. Lambert, CL. Muller, & K. Ploog (Eds.), Clitiques et cliticisation (pp. 343
363). Paris: Champion.
Gerlach, Birgit (2002). Clitics Between Syntax and Lexicon. Amsterdam: John Benjamins.
Gibson, Edward, & Kenneth Wexler (1994). Triggers. Linguistic Inquiry, 25 (3), 407454.
Givn, Talmy (1971). Historical syntax and synchronic morphology. An archeologists eld
trip. In Papers from the Seventh Regional Meeting Chicago Linguistic Society (pp. 394415).
Chicago, IL: Chicago Linguistic Society.
Givn, Talmy (1976). Topic, Pronoun, and Grammatical Agreement. In C. N. Li (Ed.), Subject
and Topic (pp. 149188). New York, NY: Academic Press.
Givn, Talmy (1984). Syntax: A Functional-Typological Introduction, Vol. I. Amsterdam: John
Benjamins.
Greenberg, Joseph (1963). Essays in Linguistics. Chicago: Chicago University Press.
Greenberg, Joseph. (Ed.). (1966a). Universals of Language (2nd edition). Cambridge, MA: The
MIT Press.
Greenberg, Joseph (1966b). Language Universals, with Special Reference to Feature Hierarchies.
The Hague: Mouton.
Greenberg, Joseph (1978). How does a language aquire gender markers? In J. Greenberg, C. A.
Ferguson, & E. Moravcsik (Eds.), Universals of Human Language Vol. II (pp. 4782).
Stanford CA: Stanford University Press.
Grewendorf, Gnther (1987). Kohrenz und Restrukturierung. Zu verbalen Komplexen im
Deutschen. In B. Asbach-Schnitker & J. Roggenhofer (Eds.), Neuere Forschungen zur
Wortbildung und Histographie. Festgabe fr Herbert Brekle zum 50. Geburtstag (pp. 123
144). Tbingen: Gunter Narr.
Grewendorf, Gnther (1989). Ergativity in German. Dordrecht: Foris.
Grewendorf, Gnther (2002). Minimalistische Syntax. Tbingen: A. Francke.
Grewendorf, Gnther (2003). Improper remnant movement. Gengo Kenkyo: The Journal of the
Linguistic Society of Japan, 123, 4794.
Grewendorf, Gnther (2004). The discourse congurationality of scrambling. To appear in
J. Sabel & M. Saito (Eds.). The Free Word Order Phenomenon: Its Syntactic Sources and
Diversity. Berlin: Mouton de Gruyter.
Grimshaw, Jane (1997). The best clitic: Constraint conict in morphosyntax. In L. Haegeman
(Ed.), Elements of grammar (pp. 169196). Dordrecht: Kluwer.
Groat, Erich, &John ONeil (1996). Spell-Out at the LF-Interface. In W. Abraham, S. D. Epstein,
H. Thrinsson, & J.-W. Zwart (Eds.), Minimal Ideas (pp. 113139). Amsterdam: John
Benjamins.
Grnbech, Kaare, & John R. Krueger (1955). An Introduction to Classical Mongolian. Wiesbaden:
Otto Harrassowitz.
Guiraud, Pierre (1968). Lancien franais (3rd edition). Paris: Presses Universitaires de France.
de Haan, Germen, &Fred Weerman (1986). Finiteness and verb fronting in Frisian. In H. Haider
& M. Prinzhorn (Eds.), Verb Second Phenomena in Germanic Languages (pp. 77110).
Dordrecht: Foris.
Haas, Mary (1977). From auxiliary verb phrase to inectional sufx. In C. N. Li (Ed.),
Mechanisms of Syntactic Change (pp. 525537). Austin, TX: University of Texas Press.
Haeberli, Eric (1999). Features, Categories and the Syntax of A-positions. Synchronic and
Diachronic Variation in the Germanic Languages. Doctoral dissertation, University
of Geneva.
Haeberli, Eric (2002). Features, Categories and the Syntax of A-Positions. Cross-Linguistic
Variation in the Germanic Languages. Dordrecht, Kluwer.
Haeberli, Eric (2004). Syntactic effects of inectional morphology and competing grammars.
In E. Fu & C. Trips (Eds.), Diachronic Clues to Synchronic Grammar (pp. 101130).
Amsterdam: John Benjamins.
Haegeman, Liliane (1990). Subject pronouns and subject clitics in West Flemish. The Linguistic
Review, 7, 333363.
Haegeman, Liliane (1992). Theory and Description in Generative Syntax: A Case Study in West-
Flemish. Cambridge: Cambridge University Press.
Haider, Hubert (1986). Fehlende Argumente: Vom Passiv zu kohrenten Innitiven.
Linguistische Berichte, 101, 333.
Haider, Hubert (1993). Deutsche Syntax generativ. Tbingen: Gunter Narr.
Haiman, John (1971). Targets and paradigmatic borrowing in Romantsch. Language, 47 (4),
797809.
Haiman, John (1974). Targets and Syntactic Change. The Hague: Mouton.
Haiman, John(1991). FromV/2 to subject clitics: Evidence fromNorthern Italian. In E. Traugott
& B. Heine (Eds.), Approaches to Grammaticalization, Vol. 2 (pp. 135157). Amsterdam:
John Benjamins.
Haiman, John, & Paola Beninc (1992). The Rhaeto-Romance Languages. London: Routledge.
Hale, Kenneth (1973). Person marking in Warlbiri. In S. Anderson & P. Kiparsky (Eds.), A
Festschrift for Morris Halle (pp. 308344). New York, NY: Holt, Rinehart and Winston.
Hale, Mark (1996). Theory and method in historical linguistics. Ms., Concordia University,
Montreal.
Hale, Mark (1998). Diachronic syntax. Syntax, 1 (1), 118.
Hall, Christopher J. (1988). Integrating diachronic and processing principles in explaining the
sufxing preference. In J. A. Hawkins (Ed.), Explaining Language Universals (pp. 321349).
Oxford: Blackwell.
Halle, Morris (1990). An approach to morphology. In J. Carter, R.-M. Dchaine, B. Philip, &
T. Sherer (Eds.), Proceedings of the North East Linguistics Society 20, Vol. 1 (pp. 150184).
Amherst, MA: GLSA, University of Amherst, Massachusetts.
Halle, Morris, & Alec Marantz (1993). Distributed Morphology and the pieces of inection. In
S. J. Keyser & K. Hale (Eds.), The View from Building 20 (pp. 111176). Cambridge, MA:
The MIT Press.
Halle, Morris, & Alec Marantz (1994). Some key features of Distributed Morphology. In A.
Carnie, H. Harley, & T. Bures (Eds.), MIT Working Papers in Linguistics 21: Papers on
Phonology and Morphology (pp. 275288). Cambridge, MA: Department of Linguistics and
Philosophy, MIT.
Halle, Morris (1997). Distributed Morphology: Impoverishment and Fission. In B. Bruening,
Y. Kang, & M. McGinnis (Eds.), MIT Working Papers in Linguistics 30: PF: Papers At the
Interface (pp. 425450). Cambridge, MA: Department of Linguistics and Philosophy, MIT.
Halpern, Aaron (1992). Topics in the Placement and Morphology of Clitics. Doctoral
dissertation, Stanford University.
References
Harley, Heidi (1994). Hug a tree: Deriving the morphosyntactic feature hierarchy. In A. Carnie &
H. Harley (Eds.), MITWorking Papers in Linguistics 21: Papers on phonology and morphology
(pp. 275288). Cambridge, MA: Department of Linguistics and Philosophy, MIT.
Harley, Heidi, & Rolf Noyer (1999). State-of-the-article: Distributed Morphology. Glot
International, 4 (4), 39.
Harley, Heidi, & Elizabeth Ritter (2002). Structuring the bundle: A universal morphosyntactic
feature geometry. In H. J. Simon & H. Wiese (Eds.), Pronouns Grammar and
Representation (pp. 2339). Amsterdam: John Benjamins.
Harnisch, Rdiger (1989). Die sogenannte sogenannte Flexion der Konjunktionen. Ein
Paradigma aus der Bavaria thuringica. In E. Koller, W. Wegstein, & N. R. Wolf (Eds.),
Bayerisch-sterreichische Dialektforschung (pp. 283290). Wrzburg: Knigshausen &
Neumann.
Harris, Alice (1981). Georgian Syntax: A Study in Relational Grammar. Cambridge: Cambridge
University Press.
Harris, Alice (1994). Ergative-to-accusative shift in agreement: Tabassaran. In Howard I.
Aronson (Ed.), Linguistic Studies in the Non-Slavic Languages of the Commonwealth of
Independent States and the Baltic Republics (NSL 7) (pp. 113131). Chicago: Chicago
Linguistic Society.
Harris, Alice, & Lyle Campbell (1995). Historical Syntax in Cross-linguistic Perspective.
Cambridge: Cambridge University Press.
Harris, Alice, & Ricks Smeets. (Eds.). (2003). The Languages of the Caucasus. London: Routledge
Curzon.
Harris, Martin (1978). The Evolution of French Syntax. A Comparative Approach. London:
Longman.
Haspelmath, Martin (1995). The growth of afxes. In G. Booij & J. van Marle (Eds.), Yearbook
of Morphology 1994 (pp. 129). Dordrecht: Kluwer.
Hawkins, John A., & Gary Gilligan (1988). Prexing and sufxing universals in relation to basic
word order. Lingua, 74, 219259.
Heine, Bernd, Ulrike Claudi, & Friederike Hnnemeyer (1991). Grammaticalizazion: A Con-
ceptual Framework. Chicago: University of Chicago Press.
Heine, Bernd, & Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge:
Cambridge University Press.
Heine, Bernd, &Mechthild Reh (1984). Grammaticalization and Reanalysis in African Languages.
Hamburg: Helmut Buske.
Heller, Karin (1975). Syntaktische Einsse des Italienischen im Zimbrischen. Incontri
Linguistici, 2, 165176.
Hetzron, Robert (1976). The Agaw languages. Afroasiatic Linguistics, 3 (3), 145.
Hewitt, George (1996). Georgian: AStructural Reference Grammar. Amsterdam: JohnBenjamins.
Hock, Hans H. (1991). Principles of Historical Linguistics (2nd ed.). Berlin: Mouton de Gruyter.
Hoekstra, Jarich, & Laszlo Marcz (1989). On the position of inection in West Germanic.
Working Papers in Scandinavian Syntax, 44, 7588.
Hoekstra, Eric, & Caroline Smits (1999). Everything you always wanted to know about
complementizer agreement. In E. van Gelderen &V. Samiian (Eds.), Proceedings of WECOL
1998. Fresno, CA: California State University Press.
Hhle, Tilman N. (1978). Lexikalische Syntax: Die Aktiv-Passiv-Relation und andere Innit-
konstruktionen im Deutschen. Tbingen, Niemeyer.
Holmberg, Anders, & Christer Platzack (1988). On the role of inection in Scandinavian syntax.
Working Papers in Scandinavian Syntax, 42, 2542.
Holmberg, Anders, & Christer Platzack (1991). On the role of inection in Scandinavian syntax.
In W. Abraham, W. Kosmeijer, &E. Reuland (Eds.), Issues in Germanic Syntax (pp. 93118).
Berlin: Mouton de Gruyter.
Holmberg, Anders, & Christer Platzack (1995). The Role of Inection in Scandinavian Syntax.
Oxford: Oxford University Press.
Holmberg, Anders (2000). Scandinavian Stylistic Fronting: How any category can become an
expletive. Linguistic Inquiry, 31 (3), 445484.
Hopper, Paul, & Elizabeth Traugott (1993). Grammaticalization. Cambridge: Cambridge
University Press.
Hotzenkcherle, Rudolf (1971). Die sdwalserisch-ernetbirgischen Mundarten im Spiegel ihrer
Verbalformen. In M. Bindschedel, R. Hotzenkcherle, & W. Kohlschmidt (Eds.), Festschrift
fr Paul Zinsli (pp. 7998). Bern: Francke.
Humboldt, Wilhelm von (1822). ber das Entstehen der grammatischen Formen und ihren
Einu auf die Ideenentwicklung. Abhandlungen der Akademie der Wissenschaften zu Berlin.
Reprint: Humboldt 1972, 3163.
Humboldt, Wilhelm von (1836). ber die Verschiedenheit des menschlichen Sprachbaues und
ihren Einu auf die geistige Entwicklung des Menschengeschlechtes. Reprint: Humboldt
1972, 368756.
Humboldt, Wilhelm von (1972). Schriften zur Sprachphilosophie. (= Werke in fnf Bnden, ed.
A. Flitner & K. Giel, Bd. III). Darmstadt: Wissenschaftliche Buchgesellschaft.
Hyams, Nina (1986). Language Acquisition and the Theory of Parameters. Dordrecht: Reidel.
Iatridou, Sabine (1990). About Agr(P). Linguistic Inquiry, 21, 551577.
Ingram, David (1978). Typology and universals of personal pronouns. In J. H. Greenberg (Ed.),
Universals of Human Language. Vol. 3 (pp. 213247). Stanford: Stanford University Press.
Jespersen, Otto (1922). Language: Its Nature, Development and Origin. London: Allen and
Unwin.
Julien, Marit (2002). Syntactic Heads and Word Formation. Oxford: Oxford University Press.
Kachru, Yamuna (1990). Hindi-Urdu. In B. Comrie (Ed.), The Major Languages of Southeast
Asia, the Middle East and Africa (pp. 5372). London: Routledge.
Kayne, Richard (1975). French Syntax. The Transformational Cycle. Cambridge, MA: The MIT
Press.
Kayne, Richard (1984). Connectedness and Binary Branching. Dordrecht: Foris.
Kayne, Richard (1989). Facets of past participle agreement in Romance. In P. Beninc (Ed.),
Dialect Variation and the Theory of Grammar (pp. 85103). Dordrecht: Foris.
Kayne, Richard (1994). The Antisymmetry of Syntax. Cambridge, MA: The MIT Press.
Kayne, Richard (2002). Pronouns and their antecedents. In S. Epstein & T. D. Seely (Eds.),
Derivation and Explanation in the Minimalist Program (pp. 133166). Oxford: Blackwell.
Keegan, John M. (1997). A Reference Grammar of Mbay. Munich: Lincom Europa.
Keenan, Edward (2002). Explaining the creation of reexive pronouns in English. In D. Minkova
& R. Stockwell (Eds.), Studies in the History of English: A Millennial Perspective (pp. 325
355). Berlin: Mouton de Gruyter.
Kemenade, Ans van (1987). Syntactic Case and Morphological Case in the History of English.
Dordrecht: Foris.
Kemenade, Ans van, & Nigel Vincent (Eds.). (1997). Parameters of Morphosyntactic Change.
Kenstowicz, Michael (1994). Phonology in Generative Grammar. Oxford: Blackwell.
References
Kenstowicz, Michael (1997). Base-identity and uniform exponence: Alternatives to cyclicity. In
J. Durand & B. Laks (Eds.), Current Trends in Phonology: Models and Methods (pp. 363
394). CNRS and University of Salford Publications.
King, Tracy Holloway (1994). SpecAgrP and case: Evidence from Georgian. In H. Harley & C.
Phillips (Eds.), MIT Working Papers in Linguistics 22: The Morphology-Syntax Connection
(pp. 91110). Cambridge, MA: Department of Linguistics and Philosophy, MIT.
Kinyalolo, Kasangati (1991). Syntactic Dependencies and the Spec-head Agreement Hypothesis
in Kilega. Doctoral dissertation, UCLA.
Kiparsky, Paul (1973). Elsewhere in phonology. In S. Anderson & P. Kiparsky (Eds.), A
Festschrift for Morris Halle (pp. 93106). New York, NY: Holt, Rinehart and Winston.
Kiparsky, Paul (1983). Word-formation and the lexicon. In F. Ingemann (Ed.), Proceedings of the
1982 Mid-America Linguistics Conference (pp. 329). Lawrence, University of Kansas.
Kiparsky, Paul (1996). The shift to head-initial VP in Germanic. In H. Thrinsson, S. D. Epstein,
& S. Peter (Eds.), Studies in Comparative Germanic Syntax II (pp. 140179). Dordrecht:
Kluwer.
Kiparsky, Paul (1997). The rise of positional licensing. In A. van Kemenade & N. Vincent
(Eds.), Parameters of Morphosyntactic Change (pp. 460493). Cambridge: Cambridge
University Press.
Kirby, Simon (1999). Function, Selection, and Innateness: The Emergence of Language Universals.
Kitahara, Hisatsugu (1997). Elementary Operations and Optimal Derivations. Cambridge, MA:
The MIT Press.
Kollmer, Michael (1987). Die schne Waldlersprach. Bd. IIII. Prackenbach: Kollmer
(Eigenverlag).
Koopman, Hilda (1992). On the absence of Case chains in Bambara. Natural Language and
Koopman, Hilda (1996). The spec head conguration. Ms., UCLA.
Koopman, Hilda (2003). Inside the Noun in Maasai. In A. Mahajan (Ed.), Syntax at Sunset
3: Head Movement and Syntactic Theory (UCLA Working Papers in Linguistics 10) (pp. 77
115). Los Angeles, CA: UCLA Linguistics Department.
Koopman, Hilda, & Anna Szabolcsi (2000). Verbal Complexes. Cambridge, MA: The MIT Press.
Kornlt, Jaklin (1990). Turkish and the Turkic languages. In B. Comrie (Ed.), The Major
Languages of Eastern Europe (pp. 227252). London: Routledge.
Kosmeijer, Wim (1986). The status of the nite inection in Icelandic and Swedish. Working
Papers in Scandinavian Syntax, 26, 141.
Krifka, Manfred (1995). Swahili. In J. Jacobs, A. von Stechow, W. Sternefeld, & T. Vennemann
(Eds.), Syntax: An International Handbook of Contemporary Research (pp. 13971418).
Berlin: Walter de Gruyter.
Kroch, Anthony (1989). Reexes of grammar in patterns of language change. Language Variation
and Change, 1, 199244.
Kroch, Anthony (1994). Morphosyntactic variation. In K. Beals, J. Denton, B. Knippen, L.
Melnar, H. Suzuki, & E. Zeinfeld (Eds.), Proceedings of the Thirthieth Annual Meeting of the
Chicago Linguistics Society. Vol. II (pp. 180201). Chicago, IL: Chicago Linguistics Society.
Kroch, Anthony (2001). Syntactic change. In M. Baltin & C. Collins (Eds.), The Handbook of
Contemporary Syntactic Theory (pp. 699730). Oxford: Blackwell.
Kroch, Anthony, & Ann Taylor (1997). Verb movement in Old and Middle English: Dialect
variation and language contact. In A. van Kemenade & N. Vincent (Eds.), Parameters of
Morphosyntactic Change (pp. 297325). Cambridge: Cambridge University Press.
Kuen, Heinrich (1957). Die Gewohnheit der mehrfachen Bezeichnung des Subjekts in der
Romania und die Grnde ihres Aufkommens. In G. Reichenkron (Ed.), Syntactica und
Stilistica. Festschrift fr Ernst Gamillscheg zum 70. Geburtstag (pp. 293326). Tbingen:
Niemeyer.
Kuryowicz, Jerzy (1964). The Inectional Categories of Indo-European. Heidelberg: Carl Winter.
Laenzlinger, Christopher (1998). Comparative Studies in Word Order Variation. Amsterdam:
John Benjamins.
Laka, Itziar (1990). Negation in Syntax: On the Nature of Functional Categories and Projections.
Doctoral dissertation, MIT.
Lambrecht, Knud (1981). Topic, Antitopic and Verb-agreement in Non-Standard French
[Pragmatics and beyond Vol. II: 8]. Amsterdam: John Benjamins.
LaPolla, Randy (1992). On the dating and nature of verb agreement in Tibeto Burman. Bulletin
of the School of Oriental and African Studies, 55 (2), 298315.
LaPolla, Randy (1994). Parallel grammaticalization in Tibeto-Burman languages: Evidence of
Sapirs drift. Linguistics of the Tibeto-Burman Area, 17 (1), 6179.
Lechner, Winfried (2001). Reduced and phrasal comparatives. Natural Language and Linguistic
Theory, 19, 683735.
Legate, Julie (2003). Some interface properties of the phase. Linguistic Inquiry, 34 (3), 506516.
Lehmann, Christian (1988). On the function of agreement. In M. Barlow &Charles A. Ferguson
(Eds.), Agreement in Natural Language (pp. 5565). Stanford, CA: CSLI.
Lehmann, Christian (1993). Kongruenz. In J. Jacobs, A. von Stechow, W. Sternefeld, & T.
Vennemann (Eds.), Syntax: An International Handbook of Contemporary Research Vol. 1
(pp. 722729). Berlin: Walter de Gruyter.
Lehmann, Christian (2002). Thoughts on Grammaticalization (2nd edition). Arbeitspapiere des
Seminars fr Sprachwissenschaft der Universitt Erfurt.
Lehmann, Winfred (1993). Theoretical Bases of Indo-European Linguistics. London: Routledge.
Lessiak, Primus (1963). Die Mundart von Pernegg in Krnten. Marburg: N. G. Elwert.
Lightfoot, David (1979). Principles of Diachronic Syntax. Cambridge: Cambridge University
Press.
Lightfoot, David (1991). How to Set Parameters: Arguments from Language Change. Cambridge,
MA: The MIT Press.
Lightfoot, David (Ed.). (2002). Syntactic Effects of Morphological Change. Oxford: Oxford Uni-
versity Press.
Linder, Karl Peter (1987). Grammatische Untersuchungen zur Charakteristik des Rtoromanischen
in Graubnden. Tbingen: Gunter Narr.
Liver, Ricarda (1999). Rtoromanisch: Eine Einfhrung in das Bndnerromanische. Tbingen:
Gunter Narr.
Longobardi, Giuseppe (2001). Formal syntax, diachronic minimalism, and etymology: The
history of French chez. Linguistic Inquiry, 32 (2), 275302.
Ldtke, Helmut (1980). Auf dem Wege zu einer Theorie des Sprachwandels. In H. Ldtke
(Ed.), Kommunikationstheoretische Grundlagen des Sprachwandels (pp. 182252). Berlin:
de Gruyter.
Lhr, Rosemarie (1984). Reste der athematischen Konjugation in den germanischen Sprachen.
In J. Untermann & B. Brogyanyi (Eds.), Das Germanische und die Rekonstruktion der
Indogermanischen Grundsprache (pp. 2590). Amsterdam: John Benjamins.
MacKenzie, D. N. (1990). Pashto. In B. Comrie (Ed.), The Major Languages of Southeast Asia, the
Middle East and Africa (pp. 132150). London: Routledge.
References
Manzini, Rita, & Leonardo Savoia (2002). Parameters of subject inection in Italian dialects.
In P. Svenonius (Ed.), Subjects, Expletives, and the EPP (pp. 159200). Oxford: Oxford
University Press.
Marantz, Alec (1984). On the Nature of Grammatical Relations. Cambridge, MA: The MIT Press.
Marantz, Alec (1988). Clitics, morphological merger, and the mapping to phonological
structure. In M. Hammond & M. Noonan (Eds.), Theoretical Morphology: Approaches in
Modern Linguistics (pp. 253270). San Diego, CA: Academic Press.
Marantz, Alec (1992). Case and licensing. In G. Westphal, B. Ao, & H.-R. Chae (Eds.),
Proceedings of ESCOL 1991 (pp. 234253). Columbus: Department of Linguistics, Ohio
State University.
Marantz, Alec (1997). No escape from syntax: Dont try morphological analysis in the privacy
of your own Lexicon. In A. Dimitriadis, L. Siegel, C. Surek-Clark, & A. Williams (Eds.),
Proceedings of the 21st Annual Penn Linguistics Colloquium. Penn Working Papers in
Linguistics 4.2 (pp. 201225). Philadelphia, PA: University of Pennsylvania,
Martin, Samuel. E. (1961). Dagur Mongolian Grammar, Texts, and Lexicon. Based on the Speech
of Peter Onon. Bloomington, IN: Indiana University Publications.
Matthews, Peter H. (1991). Morphology (2nd edition). Cambridge: Cambridge University Press.
Meillet, Antoine (1912). Levolution des formes grammaticales. Scientia (Rivista di Scienza 12,
No. 26.6). Reprinted in A. Meillet 1958, Linguistique historique et linguistique gnrale
(pp. 130148). Paris: Champion.
Meinhof, Carl (1936). Die Entstehung ektierender Sprachen. Berlin: Dietrich Reimer.
Merchant, Jason(2001). The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis. Oxford:
Merkle, Ludwig (1975). Bairische Grammatik. Mnchen: Heimeran.
McCloskey, James (1996). On the scope of verb movement in Irish. Natural Language and
McCloskey, James (1997). Subjecthood and subject positions. In L. Haegeman (Ed.), Elements
of Grammar (pp. 197235). Dordrecht: Kluwer.
McCloskey, James, & Kenneth Hale (1984). On the syntax of person-number inection in
Modern Irish. Natural Language and Linguistic Theory, 1, 487533.
Means, Nathalie, Gordon P. Means, & Paul B. Means (1986). Sengoi-English, English-Sengoi
Dictionary. Toronto: Joint Centre on Modern East Asia, University of Toronto/York
University.
Meyer-Lbke, Wilhelm (1890). Italienische Grammatik. Leipzig: O. R. Reisland.
Meyer-Lbke, Wilhelm (1894). Grammatik der romanischen Sprachen. Bd. II: Romanische
Formenlehre. Leipzig: O. R. Reisland.
Mitchell, Erica (1994). When Agro is fused to Agrs: What morphology can tell us about the
functional categories. In H. Harley & C. Phillips (Eds.), MIT Working Papers in Linguistics
22: The Morphology-Syntax Connection (pp. 111130). Cambridge, MA: Department of
Linguistics and Philosophy, MIT.
Mithun, Marianne (1991). The development of bound pronominal paradigms. In W. P.
Lehmann & H.-J. Jakusz Hewitt (Eds.), Language Typology 1988: Typological Models in
Reconstruction (pp. 85104). Amsterdam: John Benjamins.
Moscati, Sabatino, Anton Spitaler, Edward Ullendorff, & Wolfram von Soden (1969). An
Introduction to the Comparative Grammar of the Semitic Languages. Wiesbaden: Otto
Harassowitz.
Moulton, William (1944). Scribe of the Old High German Tatian translation. PMLA, 59 (2),
307334.
Mller, Friedrich Max (1875). Vorlesungen ber die Wissenschaft der Sprache, Bd. 1. Leipzig:
Julius Klinkhardt.
Mller, Gereon (2004). Verb-second as P-rst. The Journal of Comparative Germanic Linguistics,
7 (3), 179234.
Newmeyer, Frederick (1998). Language Form and Language Function. Cambridge, MA: The MIT
Press.
Newmeyer, Frederick (2001). Deconstructing grammaticalization. Language Sciences, 23, 187
229.
Niyogi, Partha, & Robert C. Berwick (1998). The logical problem of language change: A case
study of European Portuguese. Syntax, 1 (2), 192205.
Noonan, Michael (1992). A Grammar of Lango. Berlin: Mouton de Gruyter.
Nordlinger, Rachel (1995). Split tense and mood inection in Wambaya. Berkeley Linguistics
Society, 21, 226236.
Nordlinger, Rachel (1998). A Grammar of Wambaya, Northern Territory (Australia). (Pacic
Linguistics C-140). Canberra: Pacic Linguistics.
Noyer, Rolf (1997). Features, Positions, and Afxes in Autonomous Morphological Structure. New
York: Garland.
Nbling, Damaris (1992). Klitika im Deutschen. Schriftsprache, Umgangssprache, alemannische
Dialekte. Tbingen: Gunter Narr.
Nbling, Damaris (1995). Kurzverben in germanischen Sprachen. Unterschiedliche Wege
gleiche Ziele. Zeitschrift fr Dialektologie und Linguistik, 62 (2), 127154.
Oetzel, Annette (1992). Markierte Wortstellung im Bndnerromanischen. Frankfurt am Main:
Peter Lang.
OGrady, William (1997). Syntactic Development. Chicago, IL: Chicago University Press.
Ouhalla, Jamal (1990). Sentential negation, Relativised Minimality and the aspectual status of
auxiliaries. The Linguistic Review, 7, 183231.
Ouhalla, Jamal (1991). Functional Projections and Parametric Variation. London: Routledge.
Paul, Hermann (1880). Prinzipien der Sprachgeschichte. (5th edition, published 1920, Konzepte
der Sprach- und Literaturwissenschaft, 6). Tbingen: Niemeyer.
Paul, Hermann (1952). Deutsche Grammatik II.3: Flexionslehre. Halle: Max Niemeyer.
Paul, Hermann (1998). Mittelhochdeutsche Grammatik (24th edition). Tbingen: Niemeyer.
Pesetsky, David, & Esther Torrego (2001). T-to-C movement: Causes and consequences. In
M. Kenstowicz (Ed.), Ken Hale: A Life in Language (pp. 355426). Cambridge, MA: The
MIT Press.
Pesetsky, David, & Esther Torrego (2004). The syntax of evaluation and the interpretability of
features. Ms., MIT and University of Massachusetts/Boston.
Pfalz, Anton (1918). Sufgierung der Personalpronomina im Donaubairischen. Republished
1983 in P. Wiesinger (Ed.), Die Wiener dialektologische Schule. Grundstzliche Studien aus
70 Jahren Forschung [Wiener Arbeiten zur germanischen Altertumskunde und Philologie 23]
(pp. 217235). Wien: Peter Lang.
Pintzuk, Susan (1999). Phrase Structures in Competition: Variation and Change in Old English
Word Order. New York, NY: Garland.
Pintzuk, Susan, Geroge Tsoulas, & Anthony Warner. (Eds.). (2000). Diachronic Syntax. Models
and Mechanisms. Oxford: Oxford University Press.
Platzack, Christer (1988). The emergence of a word order difference in Scandinavian
subordinate clauses. McGill Working Papers in Linguistics: Special Issue on Comparative
Germanic Syntax, 215238.
References
Poletto, Cecilia (1995). The diachronic development of subject clitics in North-Eastern Italian
dialects. In A. Battye & I. Roberts (Eds.), Clause Structure and Language Change (pp. 295
324). Oxford: Oxford University Press.
Poletto, Cecilia (1997). Pronominal syntax. In M. Maiden &M. Parry (Eds.), The Dialects of Italy
(pp. 137145). Cambridge: Cambridge University Press.
Poletto, Cecilia (1999). The internal structure of Agrs and subject clitics in the Northern Italian
dialects. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 581620). Berlin:
Mouton de Gruyter.
Poletto, Cecilia (2000). The Higher Functional Field. Evidence from Northern Italian Dialects.
Polinsky, Maria (2000). American Russian: An endangered language? Ms., University of
California, San Diego.
Polinsky, Maria (2003). Non-canonical agreement is canonical. Transactions of the Philological
Society, 101 (2), 279312.
Polinsky, Maria, & Bernard Comrie (1999). Agreement in Tsez. Folia Linguistica, 33, 109130.
Polinsky, Maria, & Eric Potsdam (2001). Long distance agreement and topic in Tsez. Natural
Language and Linguistic Theory, 19 (3), 583646.
Pollock, Jean-Yves (1989). Verb movement, UGand the structure of IP. Linguistic Inquiry, 20 (3),
365424.
Poppe, Nicholas (1954). Grammar of Written Mongolian. Wiesbaden: Otto Harrassowitz.
Poppe, Nicholas (1960). Buriat Grammar. Bloomington, IN: Indiana University Publications.
Prasada, Sandeep, & Steven Pinker (1993). Generalization of regular and irregular
morphological patterns. Language and Cognitive Processes, 8, 156.
Ramat, Anna Giacalone, & Paul Hopper (Eds.). (1998). The Limits of Grammaticalization.
Rezac, Milan (2004). Elements of Cyclic Syntax: Agree and Merge. Doctoral dissertation,
University of Toronto.
Renzi, Lorenzo, & Laura Vanelli (1983). I pronomi soggeto in alcune variet romanze (Subject
pronouns in some Romance varieties). In P. Beninc et al. (Eds.), Scritti linguistici in onore
die Giovan Battista Pellegrini (Linguistic Papers in Honor of G. B. Pellegrini) (pp. 121145).
Pisa: Pancini.
Richards, Norvin (2001). Movement in Language: Interactions and Architectures. Oxford: Oxford
University Press.
Rizzi, Luigi (1986). On the status of subject clitics in Romance. In O. Jaeggli & C. Silva-Corvalan
(Eds.), Studies in Romance Linguistics (pp. 391419). Dordrecht: Foris.
Rizzi, Luigi (1997). The ne structure of the left periphery. In L. Haegeman (Ed.), Elements of
Grammar: Handbook in Generative Syntax (pp. 281337). Dordrecht: Kluwer.
Rizzi, Luigi, & Ian Roberts (1989). Complex inversion in French. Probus, 1, 130. Republished
1996 in A. Belletti & L. Rizzi (Eds.), Parameters and Functional Heads (pp. 91116). Oxford:
Roberts, Ian (1993a). Verbs and Diachronic Syntax. A Comparative History of English and French.
Dordrecht: Kluwer.
Roberts, Ian(1993b). Aformal account of grammaticalizationin the history of Romance futures.
Folia Linguistica Historica, 13, 219258.
Roberts, Ian (1994). Second position effects and agreement in Comp. Ms., University of Bangor.
Roberts, Ian (1995). Object movement and verb movement in Early Modern English. In H.
Haider, S. Olsen, &S. Vikner (Eds.), Studies in Comparative Germanic Syntax (pp. 269284).
Dordrecht, Kluwer.
Roberts, Ian (1997). Directionality and word order change in the history of English. In A.
van Kemenade & N. Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 397427).
Roberts, Ian (1998). Have/Be raising, Move F, and Procrastinate. Linguistic Inquiry, 29 (1), 113
125.
Roberts, Ian (1999a). Verb movement and markedness. In M. DeGraff (Ed.), Language Creation
and Language Change: Creolization, Diachrony and Development (pp. 287329). Cambridge,
MA: The MIT Press.
Roberts, Ian (1999b). Agreement marking in Welsh and Romance. In H. van Riemsdijk (Ed.),
Clitics in the Languages of Europe (pp. 621638). Berlin: Mouton de Gruyter.
Roberts, Ian, & Anna Roussou (1999). A formal approach to grammaticalization. Linguistics,
37, 10111041.
Roberts, Ian, & Anna Roussou (2003). Syntactic Change. A Minimalist Approach to Gram-
maticalization. Cambridge: Cambridge University Press.
Robinson, Orrin W. (1997). Clause Subordination and Verb Placement in the Old High German
Isidor Translation. Heidelberg: C. Winter.
Rohlfs, Gerhard (1949). Historische Grammatik der italienischen Sprache und ihrer Mundarten.
Bd. II: Formenlehre und Syntax. Bern: A. Francke.
Rohrbacher, Bernhard (1994). The Germanic VO languages and the Full Paradigm: A Theory of
V-to-I Raising. Doctoral dissertation, University of Massachusetts, Amherst.
Rohrbacher, Bernhard (1999). Morphology-driven Syntax: A Theory of V-to-I Raising and Pro-
drop. Amsterdam: John Benjamins.
Ross, JohnR. (1969). Guess who? In R. Binnick, A. Davison, G. Green, &J. Morgan(Eds.), Papers
from the 5th regional meeting of the Chicago Linguistic Society (pp. 252286). Chicago, IL:
Chicago Linguistic Society.
Rowley, Anthony R. (1994). Morphologie aus Syntax natrlich. Zur Flexion der Nebensatz-
einleiter in nordostbayerischen Dialekten. In W. Viereck (Ed.), Verhandlungen des Inter-
nationalen Dialektologenkongresses Bamberg, 29.7. 4.8.1990). Band 3: Regionalsprachliche
Variation, Umgangs- und Standardsprachen (Zeitschrift fr Dialektologie und Linguistik,
Beiheft 76) (pp. 488497). Stuttgart: Franz Steiner Verlag.
Sabel, Joachim (1996). Restrukturierung und Lokalitt. Berlin: Akademie Verlag.
Sapir, Edward (1921). Language. New York, NY: Harcourt, Brace and World.
Saneev, Garma Dancaranovic (1973). The Modern Mongolian Language. Moscow: Nauka.
Sauerland, Uli (1996). The late insertion of Germanic inection. Ms., MIT.
Savoia, Leonardo (1997). Inectional morphology of the verb. In M. Maiden & M. Parry (Eds.),
The Dialects of Italy (pp. 7586). London: Routledge.
Schlegel, August Wilhelm von (1818). Observations sur la langue et la littrature provenales.
Paris: Librairie grecque-latine-allemande.
Schulze, Wolfgang (1998). Person, Klasse, Kongruenz. Fragmente einer Kategorialtypologie des
einfachen Satzes in den ostkaukasischen Sprachen. Band II: Die Person. Ms., University
of Munich.
Schtze, Carson (1994). Serbo-Croatian second position clitic placement and the phonology-
syntax interface. In A. Carnie, H. Harley, & T. Bures (Eds.), MIT Working Papers in
Linguistics 21: Papers on Phonology and Morphology (pp. 373473). Cambridge, MA:
Department of Linguistics and Philosophy, MIT.
Schweizer, Bruno (1952). Zimbrische Gesamtgrammatik. Fnfter Teil: Syntax der zimbrischen
Dialekte in Oberitalien. Ms.
Shlonsky, Ur. (1994). Agreement in Comp. The Linguistic Review, 11, 351375.
References
Sievers, Eduard (Ed.). (1961). Tatian. Darmstadt: Wissenschaftliche Buchgesellschaft.
Siewierska, Anna, & Dik Bakker (1996). The distribution of subject and object agreement and
word order type. Studies in Language, 20 (1), 115161.
Siewierska, Anna (1999). From anaphoric pronoun to grammatical marker: Why objects dont
make it. Folia Linguistica, 33 (2), 225251.
Siewierska, Anna (2004). Person. Cambridge: Cambridge University Press.
Sigursson, Halldr A. (1996). Icelandic nite verb agreement. Working Papers in Scandinavian
Syntax, 57 146.
Sigursson, Halldr A. (2000). The locus of case and agreement. Working Papers in Scandinavian
Syntax, 65, 65108.
Sigursson, Halldr A. (2001). Inectional features and clausal structure. In J. Niemi & J.
Heikkinen (Eds.), Nordic and Baltic Morphology: Papers from A NorFA Course, Tartu, June
2000 [Studies in Languages 36] (pp. 99111). University of Joenssu.
Sigursson, Halldr A. (2002). Agree and agreement: evidence from Germanic. Working Papers
in Scandinavian Syntax, 70, 101156.
Simpson, Andrew, & Tanmoy Bhattacharya (2003). Obligatory overt wh-movement in a wh-in-
situ language. Linguistic Inquiry, 34 (1), 127142.
Simpson, Andrew, & Zoe Wu (2002). Agreement, shells and focus. Language, 78 (2), 287313.
Sommer, Thomas (1994). Flexionsmorphologie des Verbs im althochdeutschen Tatian. Mnchen:
Tuduv.
Speas, Margaret (1991). Functional heads and inectional morphology. The Linguistic Review,
8, 389417.
Spencer, Andrew (1991). Morphological Theory. Oxford: Blackwell.
Spencer, Andrew (1997). Inectional morphology and functional heads. In W. Dressler, M.
Prinzhorn, & J. Rennison (Eds.), Advances in Morphology (pp. 3149). Berlin: Mouton
de Gruyter.
Spiess, Federico (1956). Die Verwendung des Subjektpersonalpronomens in den lombardischen
Mundarten [Romanica Helvetica 59]. Bern: A. Francke.
Sportiche, Dominique (1998). Partitions and Atoms of Clause Structure. London: Routledge.
Steele, Susan (1977). Clisis and diachrony. In C. N. Li (Ed.), Mechanisms of Syntactic Change
(pp. 539579). Austin, TX: University of Texas Press.
Steele, Susan (1978). Word order variation: A typological study. In J. H. Greenberg, C. A.
Ferguson, & E. A. Moravcsik (Eds.), Universals of Human Language IV: Syntax (pp. 585
623). Stanford, CA: Stanford University Press.
Steele, Susan (1995). Clisis. In J. Jacobs, A. von Stechow, W. Sternefeld, & T. Vennemann (Eds.),
Syntax: An International Handbook of Contemporary Research Vol. 2 (pp. 12261235).
Berlin: Walter de Gruyter.
Steriade, Donca (2000). Paradigm uniformity and the phonetics/phonology boundary. In M.
Broe & J. Pierrehumbert (Eds.), Papers in Laboratory Phonology V: Acquisition and the
Lexicon (pp. 313334). Cambridge: Cambridge University Press.
Stowell, Tim (1996). The phrase structure of tense. In J. Rooryk & L. Zaring (Eds.), Phrase
Structure and the Lexicon (pp. 277292). Dordrecht: Kluwer.
Stump, Gregory T. (1998). Inection. In A. Spencer & A. Zwicky (Eds.), The Handbook of
Morphology (pp. 1343). Oxford, MA: Blackwell.
Suer, Margarita (1988). The role of agreement in clitic-doubled constructions. Natural
Language and Linguistic Theory, 6, 391434.
Suer, Margarita (1992). Subject clitics in the Northern Italian vernaculars and the matching
hypothesis. Natural Language and Linguistic Theory, 10, 641672.
Swan, Toril (1994). Old English and Old Norse initial adverbials and word order. In T. Swan,
E. Mrck, & O. J. Westvik (Eds.), Language Change and Language Structure: Older Germanic
Languages in Comparative Perspective (pp. 233270). Berlin: Mouton de Gruyter.
Szemernyi, Oswald (1989). Einfhrung in die vergleichende Sprachwissenschaft (3rd edition).
Darmstadt: Wissenschaftliche Buchgesellschaft.
Taylor, Ann (1994). Variation in past tense formation in the history of English. In R. Izvorski,
M. Meyerhoff, B. Reynolds, & V. Tredinnick. (Eds.), University of Pennsylvania Working
Papers in Linguistics 1 (pp. 143159). University of Pennsylvania, Philadelphia.
Thrinsson, Hskuldur (1996). On the (non-)universality of functional categories. In W.
Abraham, S. D. Epstein, H. Thrinsson & J.-W. Zwart (Eds.), Minimal Ideas. Syntactic
Studies in the Minimalist Framework (pp. 253281). Amsterdam: John Benjamins.
Traugott, Elizabeth, & Bernd Heine (Eds.). (1991). Approaches to Grammaticalization.
Travis, Lisa deMena (1984). Parameters and Effects of Word Order Variation. Doctoral
dissertation, MIT.
Trips, Carola (2002). From OV to VO in Early Middle English. Amsterdam: John Benjamins.
Torrego, Esther (1995). On the nature of clitic doubling. In H. Campos &P. Kempchinsky (Eds.),
Evolution and Revolution in Linguistic Theory: Essays in Honor of Carlos Otero (pp. 399418).
Dordrecht: Kluwer.
Torrego, Esther (2002). Arguments for a derivational approach to syntactic relations based on
clitics. In S. Epstein & T. D. Seely (Eds.), Derivation and Explanation in the Minimalist
Program (pp. 249268). Oxford: Blackwell.
Tortora, Christina (2003). The Syntax of Italian Dialects. Oxford: Oxford University Press.
Tyroller, Hans (2003). Grammatische Beschreibung des Zimbrischen von Lusern. Stuttgart: Steiner.
Ura, Hiroyuki (2000). Checking Theory and Grammatical Functions in Universal Grammar.
Ura, Hiroyuki (2001). Case. In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary
Syntactic Theory (pp. 334373). Oxford: Blackwell.
Uriagereka, Juan (1995). Aspects of the syntax of clitic placement in western Romance. Linguistic
Inquiry, 26, 79123.
Vance, Barbara (1997). Syntactic Change in Medieval French: Verb Second and Null Subjects.
Dordrecht: Kluwer.
Vanelli, Laura (1997) (with Lorenzo Renzi). Personal pronouns and demonstratives. In M.
Maiden & M. Parry (Eds.), The Dialects of Italy (pp. 106115). London: Routledge.
Velasco, J. B. de (1737). Arte de la Lengua Cahita. (Edited and reprinted by Eustaquio Buelna,
Mexico, 1890).
Vikner, Sten (1994). Finite verb movement in Scandinavian embedded clauses. In N.
Hornstein & D. Lightfoot (Eds.), Verb Movement (pp. 117147). Cambridge: Cambridge
University Press.
Vikner, Sten (1995). Verb Movement and Expletive Subjects in the Germanic Languages. Oxford:
Vikner, Sten (1997). V-to-I movement and inection for person in all tenses. In L. Haegeman
(Ed.), The New Comparative Syntax (pp. 189213). London: Longman.
Vogel, Irene (1999). Subminimal constituents in prosodic phonology. In S. J. Hannahs &
M. Davenport (Eds.), Issues in Phonological Structure (pp. 251267). Amsterdam: John
Benjamins.
References
Vogelaer, Gunther de, Annemie Neuckermans, & Guido vanden Wyngaerd. (2002).
Complementizer agreement in the Flemish dialects. In S. Barbiers, L. Cornips, & S. van der
Kleij (Eds.), Meertens Institute Electronic Publications in Linguistics (MIEPiL) II: Syntactic
Microvariation (pp. 97115). Amsterdam.
Wartburg, Walther von (1970). Einfhrung in die Problematik und Methodik der Sprach-
wissenschaft (3rd edition). Tbingen: Niemeyer.
Watanabe, Akira (1996). Case Absorption and WH-agreement. Dordrecht: Kluwer.
Watanabe, Akira (2000). Feature copying and binding: Evidence from complementizer
agreement and switch reference. Syntax, 3 (3), 159181.
Watkins, Calvert (1969). Indogermanische Grammatik III.1: Geschichte der indogermanischen
Verbalexion. Heidelberg: Carl Winter.
Weber, Albert (1987). Zrichdeutsche Grammatik. Ein Wegweiser zur guten Mundart. Zrich:
Hans Rohr.
Wei, Helmut (1998). Die Syntax des Bairischen. Studien zur Grammatik einer natrlichen
Sprache. Tbingen, Niemeyer.
Wei, Helmut (2002). Agr-in Comp and partial pro-drop. Ms., Universitt Regensburg.
Wei, Helmut (To appear). Inected complementizers in Continental West Germanic dialects.
Zeitschrift fr Dialektologie und Linguistik.
Werner, Otmar (1988). Mundartliche Enklisen bei Schmeller und heute. In L. M. Eichinger &
B. Naumann (Eds.), Johann Andreas Schmeller und der Beginn der Grammatik (pp. 127
147). Mnchen: Oldenburg.
Wexler, Kenneth, & Peter Culicover (1980). Formal Principles of Language Acquisition.
Cambridge, MA: The MIT Press.
Wexler, Kenneth, & Rita M. Manzini (1987). Parameters, Binding theory, and learnability.
Linguistic Inquiry, 18, 413444.
Widmer, P. Ambros (1959). Das Personalpronomen im Bndnerromanischen in phonetischer und
morphologischer Schau. (Romanica Helvetica 67). Bern: A. Francke.
Wiesinger, Peter (1989). Die Flexionsmorphologie des Verbums im Bairischen. Wien: Verlag der
sterreichischen Akademie der Wissenschaften.
Wischer, Ilse, & Gabriele Diewald (Eds.). (2002). New Reections on Grammaticalization.
Wright, Joseph (1925). Old English Grammar. (3rd edition). Oxford: Oxford University Press.
Wu, Zoe (2004). Restructuring and the development of functional categories. In E. Fu &
C. Trips (Eds.), Diachronic Clues to Synchronic Grammar (pp. 191217). Amsterdam: John
Benjamins.
Wurmbrand, Susi (2001). Innitives: Restructuring and Clause Structure. Berlin: Mouton de
Gruyter.
Zanuttini, Raffaella (1997). Negation and Clausal Structure. Oxford: Oxford University Press.
Zehetner, Ludwig (1985) (unter Mitarbeit von Ludwig M. Eichinger, Reinhard Rascher, Anthony
Rowley, & Christopher J. Wickham). Das bairische Dialektbuch. Mnchen: Beck.
Zeller, Jochen (1994). Die Syntax des Tempus. Opladen: Westdeutscher Verlag.
Zeller, Jochen (2001). Particle Verbs and Local Domains. Amsterdam: John Benjamins.
Zwart, Jan-Wouter (1993a). Dutch Syntax: A Minimalist Approach. Doctoral dissertation,
University of Groningen.
Zwart, Jan-Wouter (1993b). Clues from dialect syntax: Complementizer agreement. In W.
Abraham & J. Bayer (Eds.), Dialektsyntax (Linguistische Berichte Sonderheft 5) (pp. 246
270). Opladen: Westdeutscher Verlag.
Zwart, Jan-Wouter (1997a). Morphosyntax of Verb Movement. A Minimalist Approach to the
Syntax of Dutch. Dordrecht: Kluwer.
Zwart, Jan-Wouter (1997b). Words and features. On lexicalization in Generative Grammar. Ms.,
University of Groningen.
Zwart, Jan-Wouter (2001). Syntactic and phonological verb movement. Syntax, 4 (1), 3462.
Zwicky, Arnold, & Geoffrey Pullum (1983). Cliticization vs. inection: English NT. Language,
59 (3), 502513.
JB[v.20020404] Prn:8/09/2005; 14:07 F: LA81IND.tex / p.1 (45-137)
Index
A
A-movement :,, :o, o,, IoI,
II8I:o, I:,
feeding agreement II8
A-position o8, ,,, IoI, Ioo, I:,
A/A-distinction Ioo
A-agreement II
A-movement :,, :o, ,o, o,, IoI,
II8I:o, I:,, ::
Ablaut :88
Abney, Steven ,
Abraham, Werner ix, o, II:,
I,8, Io,, Ioo
absolutive, see Case
Accessibility
marker I:, I,
theory ,, I:, I,
accusative, see Case
Ackema, Peter ,,, Io,Io,, II:,
II,, I:,
activating Case feature, see Case
adjacency between clitic and
nite verb, see clitic
adjacency effect, see
complementizer agreement
adjective I,, ,, 8, ,,, I,o, I,I
inection ,:
adjunction I,, ,o, ,, 8I, 8:, Ioo,
Io:, Io, III, II:, I:II:,,
I:,, I,,, :I,, :8I, :,o, ,oI
head adjunction ,o, o:, ,,,
,o,, II,, I::, :I8, ,oo
adverb oI, ,o, ,,, Ioo, Io,,
I:oI::, I,I, I8I, I,,, ::o
VP-adverb ,,, oo
afx-hopping ,o
Afro-Asiatic ,I, II
agglutinative ideal I:
Agree Io, :,:8, ,, ,o, ,,, ,o,
o,, o8, ,, ,,, 8I, 8,8,,
8,8,, ,,, ,, ,8, Io,,
Io,Io,, IIIIo, II,, I:o,
I:I, I:, I:,, I,,, IIIo,
I8I,o, Io8, I,o, I8,, :o8,
:I,, ,oo
loss of an Agree relation I:
multiple Agree Io
agreement (Agr)
3rd person agreement, lack of
,, :o, ::,, :,, :,
as a relational category ,8, 8
canonical (via Agree) o,, 8:,
8,, ,, Io8, I:o, I8,
Io8I,I, I,8, :o:, :I,, ,oo,
,o:
checking :,, ,,, ,, ,8, 8,
context-sensitive II
controller :, ,, I,, Io, I,, ,,
,o, o,,, 8I, 8,, IoI, IIo,
II8, II,, I::, I,,
default Io, ,:, ,I, o,, II8, :,,
::, :,I
defective paradigm I, I:,,
:I8, :,,, :,,, :8,, :8o, :,:
domain o8, ,,, ,, II,
ergative/absolutive agreement
,o,
features :, ,I, ,,, ,o, ,,, 8:,
IoI, Io:, Ioo, II8, I:I, I:,
I,,, Io,, :,, :,8, :o,, :o,
:o8, :,,, :8,
features, interpretability of
I,
fused marking , :o, o, II,,
II8, I:I
grammaticalization path ,
,o, o, I,8, :I,, ,o:
head I, ,,, ,o, o,, 8I, ,I, ,,,
IoI, II, I,o, :,o, :,o, :,:,
:,
in gender :,,, :,,, :o, :,I,
:8, :,,
in noun class ,o
in number ,, II, o,
in person ,, ,II, I,, I,, I,,
:o, I, IoI, Io:, Io,, ::,
:,, :,, :,,:,,, :o, :o8,
:o,, :,, :8, ,o
in tense features IIo
licensing of :8, 8, 8o, 8,, 8,,
,,, II,
loss of , ,,, :,,:,,, :,:,
:,o, :8:, :88
morphological realization of
oIo,
multiple agreement Io, I,,
o:, o,, 8,, ,, ,,, II, IIo,
I:, I8, I,o, I,o, :I,, :o,,
:,,, :8:, :8,, ,oo, ,o:
object agreement ,, I,, o,
8,, 8, 8,, II,, II,, I,, I,,,
:I8
participle agreement ,o,
8,8,, ,,, ,, II,
phrase-structural
representation of ,oo,
prexal , ,, 8, o,, I,o, I,8,
I, I,, I,I, I,:, I,, I,,,
I,,, :o,, :o, :o,, :o8, :I,
::,, :,,, :,o, :,I, :,, :,,,
:o, :8I, :8:, :,, ,o:
sufxal :, ,, 8, ,, I8, II8, I,o,
I,I, I,,I,,, Io:, Io,, I,o,
I,,, I8,, I8,, :o,:I:,
:I:Io, ::I, :::, ::,, ::,,
:,:, :,, :,,, ::, :,, :o,
:,, :,,, :,, :,o, :,8, :,,,
:o, :o,, :o,, :,,, :,,, :,I,
:,, ,oI, ,o:
and sufxing preference ,,,
I,:, I,,, I,,, :I,, :Io, ,oI
via stem vowel alternation
:,8-:o, :o, :o,, :8o, :88,
:,,
via consonant alternations (in
Skou) :8I, :,o
weakening (in Dutch) II:, II,
agreement morpheme
Index
Agr-on-C Io:, Io,, Io,,
Io,III, II,, I:,, I:o, I:,,
Io:, Io,, Io,I,:, I,, I,,,
I,8, I8,, I,,, :oI, :o:,
:Io:I8, ::o, :o, :8, ,o:
Agr-on-C, distribution of (in
Sutselvan) I,,, :oI, :o:
Agr-on-T I:,, I,, I8, I,o,
I,I, I,o, :o,, :I8, :, :o,
:o,, :,,, :,o
linearization of oo,, ,o,
parasitic on other functional
heads Io, ,o, o,, 8o, 8:,
I:I, I,,, :I,, ,oo
agrammatism II8
AgrC ,8, IoIIo,
adjunction to AgrCP Io:
AgrC-to-C movement Io:
SpecAgrCP IoI, Io:, I:,
agreement weakening (in
Dutch), see agreement
AgrO ,, oo, ,, II,, II,, II8
SpecAgrOP ,,
Agr-on-C, see agreement
morpheme
Agr-on-T, see agreement
morpheme
AgrP I, oo, oI, o,, ,:, Ioo, IoI
AgrS ,, ,, o:, oo,, 8o, ,,,
,8, Io,, II,, II8, I,, :,,
AgrS-to-C movement
,8Ioo
SpecAgrSP , ,,, ,,, I:
Albright, Adam :,8:o, :88
Alemannic I,I, I,,, :I,, :I
Alexiadou, Artemis :o
Algonquian o,, II,
Allen, Cynthia ,,, :88
Altmann, Hans ,, ,, Io:, III,
I,8Ioo, Io, Io,, :,o
American Russian, see Russian
Ampezzan I, I8,
Anagnostopoulou, Elena :o
analogical change ,, I,, :I, Io:,
I,o, :::, :,,, :,8, :oo, :o,,
:8, :88
analogical extension :,,, :,,,
:8,
analogical leveling :,,:I,
:,, :o, :88
Anderson, Stephen :,, ,:, o,,
I:I, I,, :,I, ,o,
anti-agreement ,I, I,,, I,,, I,I
aorist I8, 8,, 8o
argument licensing ,8
argument order ,,
argument structure :, II, I,,
I,I, Io8, I,,, I8, I8,, I,o, :o8,
:I,, :I,, ,o
Ariel, Mira ix, III,, I8, :,,
:8, :8o
Aronoff, Mark :,, ,,, :,I
Arregi, Karlos ,, I8, ,I
Ashby, William ,, o, I, :,,, :,,
:,o
Aspect (Asp) ,, 8, ,,, ,8,
o:o,, o,, ,,, ,8, ,,, 8o8:,
88,I, II, I:I, I:,
asymmetric V2, see V2
Athabaskan :o
Auger, Julie ,, I,,, I,8, :,,, :,o
Austronesian , II, :,,, ,o
Author in Speech Event (Auth)
:8, :,, :,,:,,, :,8,
:,::,, :,I, :,:
auxiliaries 8, ,, :o, o:, ,
,,oI, o,, I:,, ::o
B
Badiot I,
Baker, Mark ,, ,,, oI, o:, 8,,
,o, I:,, :8, :,o, :,I
Bantu ,, I8, I
Bare Phrase Structure ,, I:I
Barlow, Michael I8
Basque ,, ,, I8, ,I
Bavarian ,, ,o,8, Ioo, Io:, Io,,
I:8, I,8, IooIo8, I,oI,:,
I,, I,8, I,,, I8,, I8,, I8,,
I,o, I,8, :oI, :Io:I,,
:,o:,:, :,,, :,,, ::, :,
:o, :,:, :,, :8,, :8,
Lower Bavarian ,o, I:,, I:,,
Io,Io,, :I8, :o,, :,o
Northern Bavarian :I,, :,:
Bayer, Josef ,, ,, Io:, I:, I:,,
I,8, IooIo:, Io,, Io8, :8
Beard, Robert ,o
Bejar, Susana o,, 8o, II, I::
Belletti, Adriana ,, ,,, o,
Beninc, Paola ,, I,,I,,, I,
I,, I,,, I,,, I8o, I8,I88, I,:,
I,,, I,,, ::I::,, :,,, :o,, :,I,
:,:, :,
Bennis, Hans ,, I:
Benveniste, Emile 8, :,
Berber ,I
Berwick, Robert ,o
Besten, Hans den I:
Bickerton, Derek :,8
big DP, see D
bilingualism ,8
Blackfoot II,
blocking effects, see
morphological blocking
effects
Blocking Principle :,o, :,I,
:,,:,,, :,,, :,8, :I:,,
:,, :,:,o, :,8, :o, :o8,
:,I, :,, :,8, :8,:8,, :8,,
:,, :,,, ,o:, ,o,
and elimination of
syncretisms :,o, :,,
and phonemic distinctions
:I
operates in a local fashion
Io, :
Bobaljik, Jonathan ,o, ,:, ,,, o,,
o8, ,,, ,8, 8,, ,o, II, IIo, II8,
I:oI:,, I8:, :8, :,I
Bonet, Eulalia ,:, II:, I,I, :oo
Bopp, Franz :, o
Borer, Hagit ,, ,o, I,,
Bosco Gurin I,,
Brabants ,o
Brandi, Luciana ,, ,I, I,,, I,,
Branigan, Phil o,
Bresnan, Joan ,, I8, Io,, I:,, I:o,
I
Brinkmann, Hennig I8, Io:, :Io,
:,,
Brown, Samuel I:I, ::o
Bruening, Benjamin o8, II
Brugmann, Karl :, o
Buryat ,, :o,:I:, ::,
Bybee, Joan ,,, I8, I,, ,,, ,I,
I:o, I,,, :,8, :,, :8,
C
C I,, Io, :o, :,, ,:, ,, ,, ,, 8,
,o, ,,, ,o, oo, o,, o8, ,,, ,,
8:, 8,, ,8II,, I:, I:,, I:,,
I:8, I,I,o, I,:, IooIo,,
Io,, Io,, Io,, I,o, I,, I,,,
I,8, I8o, I8,, I,, I,o:o:,
:o8, :I,, ::, :o, :,,, :,,
CP 8, :,, ,:, ,, ,, ,o, o,, o8,
8,, 88, ,, ,8, Ioo, IoI, Io,,
II8, I:,, I:8, I,o, Io,-Io,,
I,:, I8I
Index
SpecCP ,:, 8,, II,, I,I, I,,,
::8
Campbell, Lyle ,8, ,:
Cardinaletti, Anna :I:
Carinthian Ioo, :Io
Carstens, Vicki o,, 8,, ,8, Ioo,
Io, Io,, III, I:o, :,
Case ::o, :8, ,o, o, ,o,
,o,8, o,, ,8, 8I, 8,8,, IoI,
II,I::, I,, :,o
absolutive ,, o,,, 8,, I,o,
I,,
accusative :, ,, ,o, 8, 8,,
,, II,, I:I, ::I
checking :8, ,,, 8,, 8,
dative ,, 8,, I:I, ::I
ergative ,, ,, 8,, 8o, ,,, II,,
I,o, I,,, I,
feature :8, ,o, 8o, 8,, Io,, I::
feature, activating :o, 8,, Io
feature, marked for deletion
:8, 8,, 8o, 8,, Io, I:I, I::
genitive ,:, ,,, :o,, :Io
licensing of :8, ,,, 8,, 8, 8o,
8,, ,,, II,, ,oo
nominative :, ,, ,, ,,
8,8o, ,,, Io:, II,, II,, I:o,
I,8, :o,, :Io, ::,, :8
Caucasian ,, I8, ,o
Cecchetto, Carlo I,
Celtic Io, I,, 8o, Io
Chafe, Wallace :o, ,o
Chamorro II
Chichewa ,, I8, I,
Chinese :II
Chomsky, Noam I, I,I,, :,:,,
,,8, ,, ,, ,, ,o, ,,,8, o,,
o8, ,, ,o, ,8, 8I8,, 88,o,
IoI, Io, Io,, III, IIo, II,, II,,
I:I, I::, I:, I:,, I:,, II, :I,,
:,, ,oo
Chung, Sandra ,8, o,, 8, II,
,o,
Cimbrian I,,, I,:I,, I,8, :I8,
:,o, ,oI
Cinque, Guglielmo ,, ,, ,,
,,, o, I:o, I:I
Clark, Eve :o,, :8o, :8,, :,,
Clark, Robin ,,, ,8, II, :,, :8,,
,o,
Classical Aztec I,o, :o,, :o,
::, ::o
Classical Mongolian :Io, :II,
::8
clause structure ,:, ,, ,, ,,,
,8, II,
clause type ,, ,,, Ioo, :,
clause union II,
clitic
adjacency between clitic and
nite verb Io, I8, I,,,
I8,, I,o, :o:, :o8, :o,, :I,
::,, :,,, ,oI, ,o:
complementary distribution
of clitic and full forms
I,o, I,I, I,:, I,,, :o, :I,
C-oriented I,, I:,, I,,
I,I,o, I,,, I,8, Io:, Ioo,
Io,, I,,, I8,, I8,-I,I, I,,
I,o, :oI, ::I, ::, ::,, :,,,
,oI,o,
C-oriented clitics, reanalysis
of I,, I,:, I,,, I,,, Io,,
Io, I,, I8,, :o:, :o,, :I,
:I,, ::,, :o,, ,oI, ,o,
In-oriented clitics, reanalysis
of I,I8, I,:, I,o
movement of 8,, :I,, :::, ::
non-distinctive (in Vicentino)
:o:, :o,
obligatorily present I,, ::,,
:,, :,o:oo, :o:, :o,, :o,,
:,I, :8,
reanalysis of I, I,I,, :I, I:,,
Io, II, I,, Io, I,I, IoI,
Io8, I,o, I,I, I,, I,,, I,,,
I,8, :o,, :o,, :II, :I, :I,,
::,, :,,, :, :,8, :,, :8:,
:8,, :,,, :,,, ,oo
second position clitics , ,I,
I8, I,, I,I, I,:, I,,, I,o,
:o,:o8, ::,::,, ,o:
Wackernagel clitics 8, ,, I,8,
:I:
clitic climbing ::I
clitic cluster ::o
clitic doubling IIo, ,,, I:,,
I,,, I,8, I,, I,I,, I,,,
I,, I,,, I8,I,o, I,8:o:,
:I, :I,:I,, :::::, ::,,
:,I, :,,, :o,, :,I, :,o, :,8,
:8,, ,o:
deniteness/specicity
restriction on I,,, I,,,
I,,I,,
double I, I,I, I,:, I,, I,,,
I,8, I, Io, I,, I,I,:,
I,, I,,, I,o, :oo, :o:, :oo,
:o8, :I, :I,, ::, ::o, ::,,
,oI
emphatic function of o, I,
I,:, I, Io, I,, I,:, I,:,
I,, :o:, :o,, :o,, :o8, :II,
:,, :,o, :,8, :,:, ,oI
loss of stylistic force I, I,:,
I,,, I,,, :o8, ::,
object clitic doubling I,,,
I,,, I,
restrictions on (in Swiss RR
dialects) I,:, I,,:oI
clitic left dislocation I,,, II,
I, I,:
clitic object pronouns III
cliticization I8, I,,, I,, IoI, I8o,
I8,, :o, :I:, :,,
loss of I,
closest c-command Io, ,, ,,,
,,, 8I, 8,, ,,, Io, II, I,,
coalescence :
Collins, Chris :I,, :,
Comanche :o, :o,, :o,
comparatives Ioo, Io,, Io,, I:,,
I:o, :o,, :,
complementizer agreement Io,
I,, ,o, ,o, ,,, 8:, ,II,,
I:,I:8, I,,, I,o, IooIo:,
Io, I,,I,,, I,o, I,,, :I,,
::o, :,I, :,,, :,,, :o, :8,,
,oo, ,o:, ,o,
adjacency effect Io,, Ioo,
IIoII,, I:, I:,, I:8, I,,
parasitic on verbal agreement
Ioo, I:o
replaces verbal agreement
,o, ,8, Io,, II:, II,, I:o, I:,,
IoI, Io,, Ioo, Io,, I,I, I8,,
:I, :I,, :I8, :, :,, :,o
restrictions on ,o, Io, I:,,
IoI, :,I, :,,, :,,, :o, :8,,
,o,
strict adjacency between
inected C and the subject
(West Flemish) I:,, I:8
Comrie, Bernhard ,, ,, o,, 8o,
I,o, :o,, :Io, :I:, ::,
concord I,, I
condensation
conditioned allomorphy I:,
conjugational allomorphy :,:,
:,,
context-sensitive agreement, see
agreement
Index
Cora :o,, ::o
Corbett, Greville ix, :, I,, I8,
I,o, :,,
Cordin, Patrizia ,, ,I, I,,, I,,,
I,,
core functional categories ,,
,,, ,, 8, ,,, 8:
covert movement :, ,, ,o, ,,
,,
covert syntax :,, ,
Cree II,, II,
creole languages :, I, :,8
cue (in language acquisition)
,,, ,,, ,8, ,:, I:,, I,o, I,,, I,I,
:,, :8,, ,oI
Culicover, Peter ,o
cumulative exponence ,I, ,I
Cupeo :oo
Cutler, Anne ,, I,,
cyclicity :,, ,
Cysouw, Michael , ,, :,, :8,
,o
D
D ,o, ,:, ,, ,,, ,,, ,, 8, ,I,
,,, 8:, 8,, I:,, I,o, I,, :oo,
:,,
DP II, I, I,, :o, :8, ,, ,,
,o, ,,, ,o,, 8,8,, ,,, IoI,
II, II,, I::, I,oI,:, II,
Io, I8:, I,o, I,:, I,8, :,:,
,o
big DP I,, I, I,:, I,,
I,I,o, I,,, :o,, :oo, ::
big DP, loss of I,, I, I,,
I,, I,o, :o:, :o8
Dagur ,, I8, ::,
Dal Negro, Silvia I,,, I,o, :I8,
:I,, :o,, :,8
Danish 8o
dative, see Case
de Haan, Germen ix, ,
declarative o, I,o, I8o
Decurtins, Alexi :,,
default agreement, see agreement
defective intervention effect ,o
Degree-0-Learnability Io
degree phrase I:o
demonstrative I8, ,, I,,, :II
diffusion ,,, I, ,:, :88, :,o
dissociated agreement
morpheme Io, I,, ,o, 8I,
8:, ,, Io8IIo, I,,, II, I,,
I,, I8, I,o, I,:, I,,, I,,,
Ioo, Io:, I,o, I,I, I,8, I8,,
I,oI,8, :oI, :I, :I,, :,,,
:, :,, :o,, :8,, ,oo, ,o:
insertion procedure Io,, IIo,
II,
parasitic on syntactic
Agr-morpheme Io,, Io,,
:o,, :I, :o,, ,oo, ,o:
similarities with clitics I,I
dissociated morpheme ,o, ,,,
8I, 8:, Io8, Io,, I:o, II, I,,,
I,o, :I,, ,oo
Distributed Morphology (DM)
:,, :8, :,, ,:, o, ,, ,o, ,I, I::
Dixon, R. M. W. , :,I
do-support ,:, I::
Donohue, Mark ix, :,,, :8I,
:8:, :,o
double, see clitic doubling
doublets, see morphological
doublets
Dresher, Elan ,,
Dutch ,, ,o, II:, II,
E
E-language I
East Franconian :I,
East Netherlandic ,,, I:,
economy principles :,, ,o, ,,,
I,,, II, :,,, :o, :8,, :8,,
,o,
Least Effort ,,, :,
Einheitsplural :I
elsewhere :,, ,I, :,, :I, :8,
:,,, :,, :,o, :,8, :,::,,
:8,, :88, :,I
Embick, David Io, ,o, ,I, 8,, ,o,
Io8, I::, :oo, :,
Emonds, Joseph ,,, ,8
English , I8, :8, ,o, ,I, ,,, ,,, ,,,
:, ,:, ,,, ,8, I:o, I:I, ::o, ::,,
:,,, :,8, :o, :,, :8, :,,,
:8o, :8,
EPP (feature) :o, :o, 8,, II8, I:,
I,, I,, Io, I,, I8:, I,,
:I,, :::, ::o
loss of an EPP feature I,,
Io, I,o, :I,, :I,
equidistance 8
ergative, see Case
Estonian ,:
expletive :o, IIo, Io,
Extension Condition I:I
external argument II,, I,
F
f-morpheme, see functional
morpheme
Falk, Cecilia ,,, :88
Fanselow, Gisbert ,o
Faroese 8o
Fassa I, :oo
feature
checking :,, ,:, ,,, Io,, Ioo,
I:,
formal :, :,, ,, I:, I:,
I,, ::o
geometry ,o, :,,, :,,, :8
matching :o, ,,,,, o,, o8,
8I, 8,, ,,, Io,, II,, IIo
syncretism I:, I,, I,
Fin ,,, 8o, 8:, Io
Finnish II,, I,
Fintel, Kai von o,, ,, ,, 8,
,,
Fiorentino I,,
Fission ,o,,, I:I
xation ,
focus (Foc) ,, I,I,, I,, o, o,,
,o, II:, I,8
FocP I,, I, IIoII:
Fodor, Janet ,,
formal features, see feature
Fox II,
French I,,, I,8, I,,, :,o, :,I
Friedemann, Marc-Ariel I,8
Friedmann, Naama II8
Frisian ,o,8, Io:, I:, I:,
Friulian I,, I,,, I, I8,, :::
Fukui, Naoki ,
full pronoun, see personal
pronoun
functional categories :,, ,,,,,
,,, ,,, ,8, o:, o, o,, o,,
,8, 8o8:, ,I, II, II8, I,,,
Io, :oo, :I,
phonological deciency of ,
I, :, ,,
universal hierarchy of o, ,8
universal properties of ,,
,oo
functional morpheme
(f-morpheme) :,, ,o, o, ,I,
,,, ,,, 8,, ,, I,, I,8, :8o, ,o:
Index
functionalist explanations I,,
:,, :,,, :8:
Furlan I,
Fusion , :I, ,I, ,I
future, synthetic 8, o, 8
G
Gabelentz, Georg von der o,
:8o
Gartner, Theodor I,,, :::
Gelderen, Elly van ix, o, 8I, 8:,
I:8
gender o,, o,, ,o, ,:, :::, :8,
:,, :oI, :o, :oo, :o8, :,I,
:8o, :8, :88, :,,, :,,
features I8, ,:, :,,, :,,, :,:
gender agreement, see agreement
genitive, see Case
Georgian ,o, 8,, 8o, ,,, II,, I::
Gerlach, Birgit ,, I,,, I,8, I,
I,,, :,,, :,o, :,:, :,,, :,o:,8,
:,:
German I,, I8, ,:, :, ,, ,,, ,,,
,, ,o, II8I:o, I:, I,,, I8,
I,,, I,,, I,,, Io:, Io, Io,, I,I,
I,:, I,,, I,o, I8:, I8,, :I,, :I,,
::o, ::,, :,o, :,,, :,,, :o,
:,, :o,, :,,, :,8, :8,, :8,, :88
Germanic :, , ,, Io, I8, :I, ,o,
,:, o, ,:, 8o, 8:, ,, ,,, Ioo,
IoI, Io,Io,, Io,, II,, I:o, I:8,
I,,, I,:, I,8, I8I, I8,, I,, I,o,
I,,, :I:, :I, ::o, ::, :,,, ,oo
Giazza I,:, I,,
Gibson, Edward ,8
Gilligan, Gary ,, I,, I,,
Givn, Talmy I, ,, ,,, II, I8, II,
I,, I,,, I,, I,,, ::,, ,o,
goal :,, ,, ,o, 8o, ,,, Io,, ,oo
active :o, 8,, 8,, Io
grammar change ,,I, 8, :,,,
:,o
grammar competition ,,, I,
8, Io,, I,, ::,
grammaticalization
cline/path , ,o, o, ,:, I,8,
:I,, ,o:
formal analysis of o,
of 1st and 2nd person
agreement ,, :o, I, :8
of 3rd person agreement ,,
I,, :o, :,, :8
of agreement, complementary
distribution of old and
new marker Ioo, :o
of agreement, pioneering role
of 1st and 2nd person ,,
I:, I,, I,, I,, :,o, :,, :,,
:o,, :8, :,o, ,oo, ,o,
of agreement, syntactic
conditions I:,, I,,, II,
I,, Io, I,,, Io8, I,,, I,,,
:8,
of agreement prexes , 8,
I,o, I,8, I, I,, I,I, I,:,
I,, I,,, I,,, :o,, :o,, :o8,
:Io, :I, ::,, :,,, :,, :,,,
:o, :8:, :,, ,o:
of agreement sufxes 8, ,,
I,o, I,:I,,, I,o, :o,:II,
:I:Io, :,,, ,oI, ,o:
of inectional morphology
I, :,o, :o8
of object agreement I:, I,
as transition from lexical to
functional elements :,
,, o, 8, ,I, ,,
unidirectional process ,:
Greek I8
Greenberg, Joseph ,, ,, ,,, :,:
Grewendorf, Gnther ix, IIo,
II,, I:o, I,, I,,, Io,, I,,, :I,,
::,, ::
Groat, Erich ,,
Grodzinsky, Yosef II8
Guiraud, Pierre :,o, :,:
H
Haas, Mary :I
Haeberli, Eric ix, ,,, ,:, ,,, II8,
I:,, :88, :,I
Haegeman, Liliane ix, ,, ,,,
I:,, I:
Haider, Hubert II8, :I,
Haiman, John ,, I,,, I,, I,8,
I,,, I,,, I8o, I8,, I8,I88, I,o,
I,:, I,,, I,,, :o:, ::I::,, :,,,
:,:,I, :o,, :,I, :,:, :,
Hale, Kenneth Io, :o, ,,, ,8, I,
,:, ,I, Io
Hale, Mark Io, :o, ,,, ,8, I, ,:,
,I, Io
Hall, Christopher I,,
Halle, Morris I, :8, ,o,:, ,I, o,,
,8, 8I, 8:, 8,, ,o, Io8, Io,, I::,
:,I, :8, :8,, :,I, :,,
Halpern, Aaron II:, I,I, :oo
Harley, Heidi :,, ,:, ,o, ,I, I::,
:8, :,,, :,:
Harnisch, Rdiger I,8, :I8
Harris, Alice ,, I8, ,8, 8,, ,o
Harris, Martin ,, :,,, :,, :,o
Haspelmath, Martin :8o
Hawkins, John ,, I,, I,,
head adjunction, see adjunction
head complex ,,, o,, 8,, 8, 8,,
:I,, :,
head domain ,:, ,,, II,
head movement :, ,o, ,,, oo,
o:, ,,,8, 8,, ,:, ,8, Ioo,
Io,, I:, ::o
interarboreal I:I
Head Movement Constraint
(HMC) oo, 88
Hebrew I8, II8
Heine, Bernd o
Hellendoorn ,,, Ioo, Io,Io,,
IIo, III
Hindi :o, II,
Hoekstra, Eric ,, I:,, I:o
Hoekstra, Jarich ,, ,8
Holmberg, Anders I,, I8:, :I,
Hopper, Paul ,, , ,8, o, I, ,:,
:8o
Huichol ::o
Humboldt, Wilhelm von ,, o
I
Iatridou, Sabine oo, oI
Icelandic ,, IIo, I8I, ::
Identication of feature content
II, I,I,o, Io8, I,o, I,o, :o8,
:I,, :I,
Identity Generalization I:,
I-language ,,
imperfect I8, ,:, ::I, :::, :,:,
:8,
Impoverishment rules ,I, ,:, ,I,
Io,, II:II, :I, ::, :o,
:8
expansion of :88, :8,
loss of :88
incorporation I,, :o
Indic :
Index
indicative ,I, Io:, I8,, :,o, :,,,
:,,, :,8, :oo, :o:o,,
:,I:,, :8,, :88, :,I, :,
Indo-European :, I,, :8, :,,
In 8, :,, ,, ,o, ,,, oI, I:
In-to-C movement ,8
IP ,,, IoI, Io,, IIo, I:I
In-oriented clitics, see clitics
inectional head I, ,, oo, Io,
I,o, :oo, :I, ::
inectional markers , ,, I,, o:,
8I, II, I:o, I,, ::, :,,,
:,o, ::, :o8, :8o, :,,, ,o
linear ordering of oo,,
,,8o, ,o,, II,, II,, I::,
I:,, ,oI
non-attested sequences of
o, ,8
non-bound :, o,, :oo
prex/sufx distinction 8I,
,I, ,, II,, I::, ,oo
selectional properties of ,
,I, I,o, I,I, :I,
inectional morpheme, see
functional morpheme
inectional morphology I, ,,,
IIo, I::, I:,, :,o, :,,, :,8,
:,o, :,,, :o8, :,,, :,8, :8,
acquisition of I,, :I, :,,
:8o, :8,, :,,, ,o:
loss of ,,, :,8
rich verbal inection I8:
inectional paradigm :,,, :o,
:,, :,,
acquisition of :,,, :8,
information structure IIo, ::,
Innu-aimn o,
interarboreal head movement,
see head movement
interface level :, :,, ,8
interrogative ,, :,I, :,
intervention effect ,o, ,, 8:,
Io,, I:o
inversion ,, 8, I8, ,o, ,,, Io,,
Io, II:, I:,, I,,, I,8, I,, I8,
I,o, I,,, Io:Io, Io,, Io,, I,:,
I,oI,8, I8o, I8I8,, I8,I,I,
I,o, I,8, I,,, :oI, :o:, :I,::I,
:,,, ::, :, :,, :,I,
:oo:o:, :o,, :o,:,,, :,,,
:,8, :8,, :8,, :,o, :,,, :,,
Irish Io, I,,
Italian I,, II8, I,oI,,, I,I, I,:,
I,,, I8, :8
J
Jacaltec I,
Jahai ,
Japanese ,8
Jespersens cycle , I
Julien, Marit ,, I8, ,o, o:, oo,,
,o,,, 8I, 8:, ,o, ,I, II, II,,
I:o, I:I, I:, I,o, :I:, :,
K
Kalmyk ,, I,o, :o,, ::,
Kaye, Jonathan ,,
Kayne, Richard I, ,,, o,, ,o,8,
8,, I,, I,,, I,, :I:
Keenan, Edward ,,
Kemenade, Ans van ,,, ,,, ,:,
:88
Khalkha ::,
Khinalug :,,
Kim, Shin-Sook ix, I::
King, Tracy Holloway 8,, I8:
Kinyalolo, Kasangati I:o
Kinyarwanda ,
Kiparsky, Paul :,, ,,, I,,, :,I,
:8,
Kipsigis II,
Kirby, Simon I,,
Kisar , :,,
Kitahara, Hisatsugu :,
Kitja :,I
Kollmer, Michael ,, Io,, Ioo, :I,,
:,o
Koopman, Hilda o8, ,:,, 8,,
8,, IoI
Korean :o
Kornlt, Jaklin ,
Krifka, Manfred I8
Kroch, Anthony ix, ,,, ,8, ,,,
,:, I,, ::o, :,I, :,, :8,, :,o
Kuen, Heinrich ,, I, :,,, :,,
:,o, :8o
Kurzverben I,o, :o,, :oo, :8,,
:,,
L
l-morpheme, see lexical
morpheme
Lhr, Rosemarie I8, :I, Io:
Laenzlinger, Christopher :,,
Laka, Itziar o:
Lambrecht, Knud ,, I,8, I,,,
:,o, :,I
Lango o:
language acquisition I, ,o,,,
I,,, :,o, :,I, :,,, :,,
:,,:,, :,, :o,, :,, :,8,
:8,, :8,:8,, :8,, ,oI,o,
logical problem of ,o
language change Io, :,, ,,,,,
II, :,,, :8,, :8,, :,,
abrupt ,, ,o, ,8, ,,, ,
gradual ,o, ,8, ,,, I, 8, ,
logical problem of ,o
language contact :,, :, :8,
language loss :,,, :,,
LaPolla, Randy ,
Lasnik, Howard Io
Late Insertion I, o, 8, ,,, Io,,
Io,, I:I, I:, :,,
Late Linearization Hypothesis ,o
Latin ,I, I, ,I
Least Effort, see economy
principles
left periphery ,o, ,:, ,, ,,, IoI,
II,, I:o, I,:, I,, I,8, I,I, :o,,
:oo, ::,
Legate, Julie ,o
Lehmann, Christian ,, , I,, I8,
o:, , ,:, :8o, ,o,
Lehmann, Winfred :, I8:o
lexical category :, ,o, o:,
, o8, 8
lexical morpheme (l-morpheme)
:,, ,o, o, :8o, ,oo
Lexical Parametrization
Hypothesis ,
Lexicon :, ,, ,o, ,I, I,:, :,,,
:,, :o, :8,, :8,, ,o,
Encyclopedia ,o
LF :, :,, ,, ,o, II
Lightfoot, David ,,, ,,,8, o,
I, ,:, I,,, Io, :,8, :88, :8,
Linder, Karl Peter ,, I,,, I8o,
I8:I8,, I8,, I,o, I,:, I,,, I,,,
I,8, ::o::, :,,, :o:,
:o,:,I, :,
Linear Correspondence Axiom
(LCA) ,o
linearization
of X
0
-complexes ,o, o, ,8,
8o, ,o,, I:I, I::, ,oI
of words I::
linear ordering of inectional
markers, see inectional
markers
Index
linguistic variation ,8
Livonian o,
Local Dislocation I::
locality
of syntactic operations
:,:,, o,, o8, ,I, ,, 88, 8,,
II,, I:o, I:8
of morphological operations
Io,, IIo
logical meaning :,, ,, ,,
Lombardo :,o
long passive, see passive
long-distance agreement (LDA)
:o, :,, ,, ,o, o,,,, 8:, 8,
II, II,, II,, I:o
Longobardi, Giuseppe ,,, I:
Lower Bavarian, see Bavarian
Luiseo :o,, :o,, :oo, ::
Lusern I,:I,
M
Macushi I,o, I,I, I,,
Mainland Scandinavian, see
Scandinavian
Malagasy ,
Mandarin ,
Manzini, Rita :,, ,, ,o, ,,
Marcz, Laszlo ,, ,8
Marantz, Alec I, :,, :8, ,o,:,
,o, ,I, o,, ,8, 8I, 8:, 8o, 8,, ,o,
Io8, Io,, I:o, I::, :,I
Maricopa :,,
Matthews, Peter ,I, ,I
Mbay o:
McCloskey, James Io, :o, Io
McFadden, Tom ix, :I,
Mchombo, Sam ,, I8, I
Meillet, Antoine :, o
Merge :, :,, :,, :8, ,:, , ,,,
8,, I:I, I:, I,o
Merge over Move :,
Meyer-Lbke, Wilhelm ,, I8,,
:,o, :o:, :,
Middle English (ME) :,, :8,
minimal domain 8
Mirror Principle oI, o:, o, ,,,
,I, ,, IIo, ,oo
Mitchell, Erica ,8, o:, o,, 8I, 8:,
8
Mithun, Marianne , ,, :o, :,,
,o
modal particle III, II:, I:8, I,o,
I,,
modal verb :, ,, :oo
Moena I,
Moghol ,
Mon-Khmer ,
Mongolian ,, ,, I,, :o, I,o, I,:,
I,,, I,,, :o,, :o,:I:, ::8, :,o,
:,,, ,o:
Mono :o, :o,
mood 8,, :,o, :,,, :oo, :o,, :,:,
:8,
morphological blocking effects
I, I,, ,,, ::,:,,, :,,, :I:,,
:o, :,, :,, :,,, :o,, :,,
:,8, :8,, :8, :8o, :,,, ,o:, ,o,
morphological component :8,
,, ,,, ,, ,,, 8,, I:, ::,
morphological doublets ,,, :,o,
:,I, :::o, :8, :8,
Morphological Economy I:o, IoI,
:,o, :,
Morphological Merger ,o, ,I,
8,, Io,, I:o, I::, Io, I,o, I,,,
Io,, :I,
morphological operations ,o,
,:, ,,, 8I, Io,, Io,, I::, I,
Morphological Structure (MS)
,,, 8I, 8:, 8,, ,o, Io,Io,, I::,
I,,, II, I,, I, I,o, Io,, :oo,
:I,, :: :,o
morphosyntactic features :8, ,I,
,:, ,o, oI, I::, :,,, :,,, :,8,
:I, ::, :, :,,, :8,, :,, ,o,
Move ::8, ,:, ,, ,o, ,,, I:I,
I:
Move > Merge reanalysis
movement, loss of ,,, 8, I:,
Io, I,, I,:
Mller, Friedrich Max :, o
Mller, Gereon ,:
multiple Agree, see Agree
multiple agreement, see
agreement
multiple exponence ,,, :o,
multiple speciers ,8
N
Nbling, Damaris III, I,8, I,,,
:Io, :I8, :o,, :,,
Nakh-Dagestanian :,, ,o, o,
Native American languages , ,,
:o
Neeleman, Ad ,,, Io,Io,, II:,
II,, I:,
Negation (Neg) ,8oo, o:, o,,
8o, 8I, II, I::, I,,, Io:, :oo,
::o, ::I, ::,, :o:, :o,, :,,
negative concord I,
neutralization (of phonemic
distinctions) :,,, :o
Newmeyer, Frederick o
Nilotic ,, ,
Niyogi, Partha ,o
nominative, see Case
Non-Standard French I,8, I,,,
:::, :,o, :,,, :,o:,o, :o,
Northern Bavarian, see Bavarian
Northern Italian dialects ,, I,,
,I, ,,, I,,, I,, I,,, I:, I,
I,, I,,, I,,, I,o, I8,, :::, ::,,
:,o, :,,, :,o:o8, :,,, :,8,
:8, :8o, :,:, :,, ,o,
noun class I8, II, ,o
Noyer, Rolf I, :,, ,I, ,:, ,o, ,I,
8,, ,o, Io8, II, I::, :oo, :,I,
:I, :8, :,,, :o,, :88, :8,, :,
NP-detachment ,,, ,, Io, I,,
:,,
NP-movement ,, 8,
number :, ,II, I,, ,:, ,,, o,, o,,
II,, I::, ::,, :,:,o, :,8,
:,,, :oI:o,, :oo, :,:, :,,
:8,, :,I:,,, ,o,
features :8, :,, :,,, :8,
:,:
NumP ,,
numeration 8:, Io8
O
object agreement, see agreement
object movement ,,, 8,, 8,8,,
II,, I:I
obligatorication ,
Oetzel, Annette I,,I8:, I,,,
::o
Oirat :o,, ::,
Old English (OE) I8, ,,, :, ::o,
:,8, :8,, :,o
Old French :,
Old High German (OHG) I8,
:, Io:, Io,, :Io, :I,, :I,, ::o,
:,,, :8,
ONeil, John ,,
Optimality Theory :88, :,:
Ortmann, Albert ix, :8o
OSV ::o
Ouhalla, Jamal ,,, ,,, o:, o,
Index
OV ,,, ,:, 8o, I,o, I,,, :I,, :
OVS :,:
P
Papago :o,
Papuan , I,, :,o, :,,, :,,, :8,,
,o,
paradigmatization o
paradigm leveling :I, ::, :o,
:8
parameter :,, ,,, ,, ,,
change ,,o, ,8, ,,, ,
default setting ,o, ,:
doublets ,,
parametrization ,, ,,, o,, ,,
Pari ,
Participant in Speech Event
(PSE) :8, :,, :,,:,,, :o,,
:,:, :,,, :,I, :,,
participle 8, ,o, oo, oI, 8,8,,
,,, ,, II,, I,8, :,
participle agreement, see
agreement
Pasamaquoddy o,, II,
Pashto ,, :o
passive 8, ,o, II,, I,
impersonal passive Io8
long passive (in German)
II8I:o
past tense 8, ,o, :,8, :,, :,o,
:8,
Paul, Hermann I8, ,,, ,,, o, 8,,
I8, :I
penultimate stress rule (in Swiss
RR dialects) I8o, I8,, :oI,
::I, ::,, :o,, :,o, :,:, :,,, :8,,
:,o, :,
percolation ,,, ,, ,8
permutation invariance ,
person :, ,:, ,,, o,, II,, I::, I,,,
I,, ::,, :, :,:,,
:,::oo, :o8, :8,, :,o
1st ,, ,, III,, I,, I,, I,, :o, ,I,
8o, II:, I, I,, Io,, I8,,
I,o, I,:, :oI, ::,, ::, :,o,
:,:,, :,:, :,, :,8, :oo,
:,:, :,, :,o, :8, :,o:,:,
,oo, ,o,
2nd ,, ,, III,, I,, I,, I,, :o,
,I, ,o, Io:, I:,, I:, I, I,,
I,,Io:, Io,, Io,, I,,,
I8,I8,, I8,, I8,, I,o, :oI,
:Io, :I8, :::::, :,o, :,,,
:,, :,, :,, :,8, :o, :o,,
:,::,, :,o, :8, :,o, :,:,
:,,, ,oo, ,o,
3rd ,I,, I,, I,, I8, :o, ,I, IIo,
I, I,,, IoI, I,o, I,:, I,,,
:o, :::, ::, ::,, :I,
:,:,, :,::,o, :,8, :,,,
:oI, :o:oo, :o8, :o,, :,o,
:8, :,o:,:, ,o,, ,o
binary feature system
:,:,, :,I
features :o, :,, :I, ::,
:,:,, :,, :,8, :o, :8,
:,o, ,o,
PersP ,,
personal pronoun (free form)
I:, ,, ,,, IoI, I,8, I,,, Ioo,
Ioo, I,o, I8,, :o:o,, :II,
:I,:I,
extraposition of :I:
weak , o, I,8, I8,, :I:, :I,,
,o:
Pesetsky, David 8, 8,, 8,, Io
PF :, :,, ,,, ,8, ,o, 8,, Io,Io,,
I:,, I,, I,
mapping to II:, :oo
PF feature checking Io,, Ioo,
I:,
PF movement II:
PF rules II,
Pfalz, Anton ,, I,8, Ioo, Io, Io,
phase :,, ,o, ,, 8,, 8,8,, ,,
Io, II,, II,
Phase Impenetrability Condition
(PIC) :,, 88, ,
phi ()-features I,I,, ::o,
:8, ,, ,,, ,,, ,8, o,, ,,
8,8,, 8,, ,,, IoIIo,, Io,,
I:I, I::, I:, I,o, I,,, I8,
I,o, :,, :,,, :o,, :oo, :,o,
:8I, :8, :,,
checking of 8,
interpretable :,, o8, ,,, 8,,
8, ,,, Io, Io,, II,, IIo, II,
,oo
marked for deletion :,, :o
non-interpretable I,, I8,
::o, :8, ,, ,o, ,8, o8, 8,,
Io, Ioo, Io8
phonological erosion , I:, I8,
:o, ,,, ,o, I, :, I:, :o,, :o,,
::I, :,:, :,, :,8, ::, :,, :,,
:,,, :8o, :88, :,,
Piattino ::,, :,,, :,o:,,, :o,,
:o8, :8, :,,
pied-piping ,:, ,,, I:o
Pinker, Steven :8,
Pintzuk, Susan ,,, ,:, :88
Platzack, Christer ,,, I8:
Pochutla :o,, :o,, ::o
Poletto, Cecilia ix, ,, ,:, ,,, I,,,
I:, I,,, I,,, :,,, :8, :,o, :o:,
:,:, :,,
Polinsky, Maria o8,:, ,, II,,
I:o, :,,:,,, :,,
Pollock, Jean-Yves ,, ,,, ,,oI,
,,, ,8, 88
Pomattertitsch I,o, :I8, :I,, :,,
Pontresina I8,
Poppe, Nicholas ,, I,o, :Io, :II,
::8
portmanteau ,I, o, o,, II,
possessive ,:, :o,, :Io, ::,
Potsdam, Eric o,,:, ,, 8,, I:o
PP Ioo, Io,, IIo, I8:, I8,, I8o
prex/sufx distinction, see
inectional markers
prexation ,,, ,8
present tense 8o, I:,, :,,, :o,
:88, :,I
Preservation of argument
structure II, I,I,o, Io8,
I,o, I,o, :o8, :I,, :I,, ,o
preterite-presents :I, Io:, :,,
Primary Linguistic Data (PLD)
:,o, :,,, :,8, :I, :,, :, :,o,
:8,, :8,, ,o,
Principles and Parameters theory
Io, :,, ,,, ,, ,,, ,o, ,,, IIo
pro IoI:, ,, ,,, Io:, I:, I:,,
I,I, I,:, II, Io, I8I,o,
I,:, Io,, I,o, :I,, :I,:I,,
::I, :::
referential Io, I,, I,, Ioo,
Io8, I8,
pro-drop I,, Io, ,,, ,,, ,8, Io:,
I:,, II, I,, I8, I,o, I,:, I,,,
Ioo, IoI, Io,, Io8, I,oI,,,
I8,I8,, I8,, :II:I, :I,,
::I::,, ::o, :,I, :,,, :,,, :o,
:,, :,o, :8,, ,oI,o,
pro-drop parameter ,:
probe :,:,, 8, 8,, 8,
Prosodic Inversion II:
prosodic phrase Io,, Ioo, II:, II,
Proto-Indo-European :, ,, :88
Index
Pullum, Geoffrey I:,, I,,I,8,
I,I
Puter I,,I8,, I,o, I,,, ::o:::,
::, :o,, :,
Q
Quechua II,
quirky subject ,
R
raising :o, ,o,:, II8, II,
Ramat, Anna Giacalone o
realizational model of grammar
:8, ,, :,I, ,o:
reanalysis of clitics as agreement
markers, see clitics
Reh, Mechthild o
relative pronoun Io,, Io, I,
remnant movement ,8, ::
Renzi, Lorenzo ,, :,,, :,o, :o:
restructuring ,o, II,, :,o
resumptive pronoun o, II, I,,
I,,,:o,, ::,, :,o
reanalysis of ,, ,, :,,
Rezac, Milan II
Rhaeto-Romance ,, I,, ,, I8,
I,,, I,o, :,o, :,,, :o:
Northern Italian dialects I
Swiss dialects Io, I,,, I,,, I,,,
I,,I8,, I,:I,, I,8:o:,
:o,, :I, ::o::,, :o8, :o,,
:,o, :,I, :,, :,,, :8,, :,o,
:,, ,oI, ,o,
Richards, Norvin I:I
Rima I,,
Rimella I,,
Ritter, Elizabeth ,o, :8, :,,, :,:
Rizzi, Luigi ,, ,,, ,, ,, Io,, IIo,
I,,, I,,, I,8, I,,
Roana I,:I,
Roberts, Ian ix, ,, 8, Io, Io, :I,
,,,8, o,, ,, 8, ,:, ,,, ,8,
IoI, Io,, I:,, I,,, I,8, II, I:,
I,, Io, I,I, I,, I,,, I,8, I8:,
:I,, :::, ::o, :,, :,, :8,,
:88, :,I, ,o:, ,o,
Rohlfs, Gerhard ,, :,o, :o:, :,
Rohrbacher, Bernhard ,8, I8:
roll-up movement ,8, I:o
Romance :, ,, 8, I, , o, 8o,
I8,, :oo, ::o, :,, :o,, :8o,
:,,, :,
root ,I, 8,, ,o, ,:, I:,, :I, ::,
:o, :,:, :,,:8I, :8:, :8, :,o
root-out insertion ,o
Roussou, Anna , ,, 8, Io, :I, :,,
,o, ,,, o,, ,, 8, ,,, I,,,
II, I:, I,, Io, I,I, I,, I,,,
Io,, I,o, I,8, :I,, :,, ,o:
Rowley, Anthony I,8, Io,, :I,
Russian :,,:,8
American Russian I,, :,o,
:,,, :,o:,8, :8,, :,,
S
Sabel, Joachim ix, II,, II,
Sanskrit I8, :o
Sapir, Edward ,, ,,, ,,
Sauerland, Uli ,:, :,I, :8,
Savoia, Leonardo ,,, :,:, :,
Scandinavian ,,, I:, :,, :8,
Mainland Scandinavian 8o,
:,,
Schtze, Carson ,I, II:, I,I, :oo
Schulze, Wolfgang ix, ,, I8, ,o
Schweizer, Bruno I,:I,, :I8
scrambling Ioo, I8,, ::
disrupting structural
adjacency Io,, IIoII,
semantic content/features :,
loss of I,
Semitic ,, 8
Sengoi o,
Separation Hypothesis ,o
Serrano :o,, :oo
Shlonsky, Ur ,,, ,, ,,, IoIIo,,
I:,, I:,
short verbs, see Kurzverben
Siewierska, Anna I, ,, , ,, ,, Io,
I:, I,, I,, :o, ,,, I,,, :,,, :,
:,, :8o, ,o
Sigursson, Halldr ix, ,:, ,
simplicity metric, see structural
simplication
Simpson, Andrew ix, :, ,, I,I,,
,:, I:o, I, I,,, ::,
Siouan ,
Skou :,,, :,,, :8:, :8,, :,o, :,,
Slavic 8
sluicing Io,
Smits, Caroline ,, I:,, I:o
Sommer, Thomas Io:, Io,, :Io
Southern Paiute :oo, :o,
SOV ,,, I,, o,, ,,, ,8, I:o, I,
I,, I8, I,oI,:, I,,, I,,, Io,,
:o8:Io, :I:, :I, ::o::8,
:,,, ,o:
Spanish I,I, I,:, I,, I,,, I,, :I8
Speas, Margaret ,8, o:, o,, 8I,
8:, 8, II,, I:,
specier-head conguration :,,
:,, ,o, ,,, o,, o8, ,:,, 8,,
IoI, Io:, II,, I:o
Spell-out :, :,, :,, :8, ,,, ,o,
8,, Io,, I:I, I:o, :I,
at LF ,,
Spencer, Andrew o:, I:, :o,
Spiess, Federico ,, :,,, :,o
split ergativity 8,
Sportiche, Dominique o,
Standard Arabic I,,
Starke, Michal :I:
Steele, Susan , I,o, :o,, :o,
:o8, ::::o
stem vowel alternation :,8-:o,
:o, :o,, :8o, :88, :,,
loss of :8
Stowell, Tim ,
Stray Afx Filter 8,
structural adjacency Io, Io,II,,
II,, I:,, I:8, Io, II, I,, I8,
I,,, I,o, I,,, I8,, I,o, I,,, :o:,
:o,, ,oo
structural simplication I,, ,,,
I:, I,I,, I,, I,o, I,:,
I,, Io8, I8,, I,o, :o:, :o8,
:I,, :I,, :I,, :::, :,,, ,o
simplicity metric I:, I,:, I,
Stump, Gregory ,o, ,I
Suer, Margarita I,,, I,, I,,,
:I8
subject-initial clauses Io,, I:,,
I88, :oo:o:, :,,
subjunctive :,o, :oo, :o,, :,:,
:,
Subset Principle
in language acquisition :,,
,o, ,:
governing Vocabulary
Insertion :,, ,o, :,I,
:::,, :8o
sufxation ,, ,,, ,8, I:I
sufxing preference ,,, II, I,
I,, I,, IIo, I,:, I,,, :I,, :Io, ,oI
suppletion I,o
Surmeiran I,,I8,, I8,, I,o, I,I,
I,,, I,,, :oo, ::I, :::, :o,
Surselvan I,,, I8o, I8:, I8,, I,o,
:::, :o,
Index
Sutselvan I,,I8:, I,:, I,,,
I,,I,,, :oI, :o:, :o8,
:::::, :o,, :,,, :8,, :8,, :,,
SVO ,,, o,, I, I,:, I,,, I8o,
I8, I,,, :o8, :I, ::o, ::,,
:,:, ,o:
Swabian :I,
Swahili ,, I8, o:, II,
Swedish ,,
Swiss Rhaeto-Romance dialects,
see Rhaeto-Romance
syncretism I:, I,, :,o, :,:
syntactic change :,, ,,,,, ,,
,
Szemernyi, Oswald :, ,
T
Tagalog ,,
Tarahumara :oo, ::,::,
target grammar ,o, ,8, ,, I:,
:I, :,8, :,I, ,oI
Taylor, Ann ::o, :,, :8,
Tense (T) ,o, ,I, ,, ,o, ,,,
,,oo, o:o,, ,,, ,,8o, 8:,
8, 8,, ,o, ,,, Io, Io,, Io8,
II,, I::I:, I:8, I,I,o,
I,o, I8,, :oo, :o:, :o8,
:I,:I,, ::,, ,oo, ,oI
defective T :o
T+Agr o,, Io,, III, I,, Io,
I8, I,,, Io8
TP :,, IIo, II:, II8, I:,, I,o,
I,:, I,,, :o,, :I,, :I, ,o:
TP fronting I,o, I,:, :I,, ::8
SpecTP I, :o, o8, ,,, ,8, IIo,
II8, II,, I:,, I,, I,, I,o,
I,, I8:, I,, I,o, :oo, :I,,
:I,, :::, ::, ::o
Tepecano ::o, ::,
Tepehuan ::o
theme vowel :,8, :o,, :o,, :,:
Theta ()
position I,, ::o
role Io, ,, I,I, II, I,I,:,
I,, Io8, I,o, I,,, I,, I,o,
:o:, :o8, :I::I,, :,,, ,oI
theory I,,, II, Io8
Thrinsson, Hskuldur ,,, II8
Tibeto-Burman ,, ,
topic o, ,, II, I,, I,, ,o, ,, 8,,
Io,, II:, II,, I:o, I,:, II,
I,, ::,, :,o
deniteness/specicity
restriction I,,
TopP IIoII:
topic-drop I8
topic left dislocation o, II, Io,
:o,, :,o, :,I, :,,, :,o, :,,
of 1st and 2nd person forms
I,, I
loss of topic movement I,
Torrego, Esther 8, 8,, 8,, Io,
I,
Traugott, Elizabeth ,, , ,8, o,
I, ,:, :8o
Travis, Lisa deMena o:, Io,, I:,,
I,,, :oo
trigger experience ,,, ,o, ,,, Io,
:,, ,o,
Trips, Carola ix, ,,, ,o
Tsakhur II,
Tsez o,, ,I, ,:, ,, ,,, 8,, II, II,
Tubatulabal :o, :o,
Turkana ,, II,
Turkish ,, ,,
Tyroller, Hans I,:, :I8
U
Umlaut ,o
unaccusative ,o
underspecication :,, ,:, ::,,
:,I, :,,, :, :o, :,8, :o,,
:8,, :8,:,:, :,,, ,o,
for agreement features :,o,
:,,, :,8
for person I,, :,, :8, :,,
:,,, :,, :o:, :,, :8, :,I,
:,:
for number :,, :,,, :o:,
:,, :,I
Universal Grammar (UG) ,,,
,,, ,,, ,,, II, :,, :,8, :8,
Ura, Hiroyuki ,8
Urdu :o
Uriagereka, Juan I,,, I,, I,,,
I,, :oo
Uto-Aztecan , I,, I,o, I,, I,,,
:o,, :o,, :o,:o,, ::, :,o,
:,,, :,,, ,o:
V
V1 ,:, I,,, I,:, I,o, I8, :,
V2 I,, ,,, ,:, ,, III, I:8, Io,
I,I,:, I,,, I,,, I,,, Io:,
Io,, Ioo, Io,, I,o, I,:,
I,8I8I, I8,I8,, I8,, I,o,
I,,, I,, I,o, I,,:o:, :o,,
:Io, :I, :I,, ::o, ::I, ::,
:,, :,, :,:, :,,, ,oI, ,o:
asymmetric analysis of Io,,
:o:
embedded V2 I:, :I,
V3 I,,, I,:, ::o
Vallader I,,I8:, I8, I8,, I,o,
I,,, I,, I,,, ::I, :::, :o,
Vanelli, Laura ,, :,,, :,o, :o:
Veneto I,, :,,, :,,
verb movement ,,, ,,, ,,, oo,
,o, ,8, ,:, Io,, I,8, Io, I,o,
I,,, I8:, :o8, :I,, :I, ::o,
::I, :,I, :,,, ,o
loss of verb movement ,,,
8, ,:
V-to-In movement ,,
V-to-In movement, loss of
,,
V-to-C movement III, I:8,
IoI, Io:, Io,, I,o
V-to-T movement III, I,
I,,
verb stem ,, ,,,,, o:, o, ,,,
,o,:, II,, I::, I,,, :,,, :88,
:,o
verb-initial languages 8, I,, I,,
Vicentino :,o, :,,:o8, :8,
Vikner, Sten I8:
VO ,,, ,:, I,, I,:, I8:, I,, :::
Vocabulary Insertion ,I, ,:, o,
,, ,o, ,,, Io8, I:I, I:o, Io,,
:,I, :,,
proceeds in a bottom-up
fashion II,, I:o, ,oo
Vocabulary items :8,:, o, ,o,
,I, ,I, ,, II,, I::, :,I, :,:,o,
::, :8, :,, :,,, :,, :8o,
:8,, :8,, :,I, ,oo, ,o,
Voice ,I, ,, ,I, ,,, 8:, I:I
VOS :::, ::o, ::,
VP :,, ,,, , ,, ,,, 8,, II,, I8I
I, :,:,, ,:, ,, ,,, ,,I, ,,,
,8, ,,, 8:, 8,, 8, 8,, 8,, ,,,
II,, I,,, :I,, ,oo
P :,, ,o, ,,, ,8, 8:, 8,, 88,
8,, ,, II,, II,, I:I, :I,
SpecP I, ,, o8, ,,, 8,, Io,
Io,, II,, I,Io, I8I,o,
I,, Io8, I,, :oo, :o8, :I,
VSO Io, I,:, I,,, :Io, ::o, ::,
Index
W
Walser German I,,, I,,I,8,
:I,, :,o, :,,, :o,, :,,, :,8, :8,,
:,,, ,oI
Wambaya ,, I,
Wartburg, Walther von ,, o, I,
:,, :,o, :8o
Watanabe, Akira o,, I:
Weerman, Fred ,, I:
Wei, Helmut ix, ,, ,, I,8, Ioo,
Io,, Ioo, Io8
Welsh Io, II, Io
West Flemish ,,,8, Io:, Io,,
I:, I:,, I:8
Wexler, Kenneth :,, ,, ,o, ,8
wh-agreement ,o,
wh-movement 8,
wh-phrase I:o, I8o, ::I
wh-questions I:o, I8o
Widmer, P. Ambros I,,, I8,, I8,,
I88, ::I, :::
Wiesinger, Peter ,, I,8, Io:,
IoIoo, :,:
Word building constraint Io,
I,I,o, I,, Io8I,o, I8,, I,o,
:o8, :I,, :I,, :,,
word formation ,,,,, ,,, 8I,
8,, ,I, I:I
Word Formation Rules I:I
word order ,, 8, ,o, ,8, ,,,
ooo, ,8, II,, I::, I:,, I,:,
Ioo, I8, :Io:I:, ::, ::,,
:,,
basic I8:, :Io, ::o
free ,o
word order change :Io
from OV to VO ,,, ,:
from SVO to SOV I,,
Wu, Zoe :, ,, I,, I, 8, I, I,,,
::,
Wurmbrand, Susi o,, o8, ,,,
II8I:o
X
X-bar theory Io
Y
Yaqui :oo, :o,, ::,::,
Yiddish 8o, :,,, :o, :88
Yuman ,, :,,
Z
Zanuttini, Raffaella o:
Zeller, Jochen ix, IIo, I:,
Zeneyze I,
Zimmermann, Ede ix, I:o, ::,
Zurich German :I,
Zwart, Jan-Wouter ,, ,o,,,
Io,, I:,, I:, I:,, Ioo, I,,, :oo
Zwicky, Arnold I:,, I,,I,8, I,I

The Rise of Agreement A Formal Approach To The Syntax and Grammaticalization of Verbal Inflectio

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

The Rise of Agreement A Formal Approach To The Syntax and Grammaticalization of Verbal Inflectio

Hochgeladen von

Copyright:

Verfügbare Formate

The Rise of Agreement

Agr []] may enter into an Agree relation

Das könnte Ihnen auch gefallen