Beruflich Dokumente
Kultur Dokumente
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may
be from any type of computer printer.
In the unlikely event that the author did not send UM I a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
O rd e r N u m b e r 9129663
UMI
300 N. ZeebRd.
Ann Arbor, M I 48106
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
X-BAR THEORY A N D M ORPHOLOGICAL JUNCTURE
DOCTOR OF PHILOSOPHY
IN
LINGUISTICS
M A Y 1991
By
Philip A. Bralich
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
W e certify th a t w e have read this dissertation and th a t, in our opinion, it is
Philosophy in Linguistics.
S '/ Chairperson
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
° Copyright by Philip Andrew Bralich 1991
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
D EDICATIO N.
and to my muse.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
v
ACKNOW LEDGEMENTS
changed. W hile I began playing w ith this idea w ith some of the data of SPE I was
only able to w rite a paper on English inflectional endings for a class w ith Fred
Eckman before the financial realities of m y life forced me to w ith draw from school.
linguistics I alw ays thought it would be interesting to pursue this idea. W hen I
idea had not been investigated. I investigated it myself and w as able to w rite the
encouraged, and guided my efforts along the w ay and I would like to acknowledge
them here.
the University of Wisconsin in M ilw aukee. His presentation of the material made
the subject seem both fascinating and approachable. A fter my eight year hiatus
from university w ork, Ken Rehg, here at the University of Hawaii w as the first to
w ork with me on this project, encouraging me to continue when others said the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
idea th at a graduate student could propose a replacem ent for the theory of juncture
w as at best presumptuous. His com m ents and suggestions have been invaluable at
all levels of this project. Later W illiam O'Grady provided his insights, and
suggestions. His w ide ranging knowledge of the field of linguistics and an uncanny
facility for finding counterexam ples helped me hone the ideas and pursue directions
Tucson M ark A ronoff very generously took time from his schedule to discuss this
morphology and the w ays in which m y ideas fit into current theories.
Finally, working w ith Derek Bickerton as the UH working papers editor and
the chairman of m y com m ittee has been invaluable in bringing this w o rk to its
present stage. D erek's comm and of linguistic theory and understanding of the
precision and depth to the ideas I am trying to express. Besides the benefits
provided by his linguistics skills Derek w as also able to com m unicate important
The other members of my com m ittee, David Stam pe and Byron Bender have
Classes and discussion w ith Stan Starosta here at the University of Hawaii
has been instrumental in dem onstrating the boundaries and limitations on particular
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ABSTRACT
elem ents called boundaries or junctures between words and morphemes to explain
the variable application of phonological rules in identical strings of segm ents across
boundaries by using the theory of a multistratal lexicon to explain the facts of the
relevant rule applications. W hile these theories provide adequate accounts of the
on the extension of phrase-level X-bar theory to word-level and that accounts for
the facts of the variable application of morphologically sensitive rules w ith out
but are hindered by the asymm etrical distributions of the Class I and Class II
word-level are significantly different from phrase-level X-bar theory and can only
possible to account for the distribution of the Class I and Class II affixes
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
viii
morphology. Once the distribution of the Class I and Class II affixes is accounted
data from past analyses of English (Chomsky and Halle 1 9 6 8 ) and M alayalam
lexical strata.
second chapter discusses the facts of the distribution of the Class I and Class II
affixes and proposes the new principle of locality. Chapter three demonstrates the
extension of phrase level X-bars to word level and chapter four introduces the
theory of X-bar juncture. Chapter five compares this theory to the other theories of
juncture and the final chapter illustrates the ability of this theory to account for the
facts of a the phonology of a language other than English. The language chosen is
Malayalam .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
A C K N O W L E D G M E N T S .................................................................................. v
A B S T R A C T ..................................................................................................... vii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
X
3 .1 . IN TR O D U C TIO N ............................................................................ 76
3 .2 . THE EXTENSION OF X-BAR THEORY TO W O RD LEVEL . . 79
3 .3 . HEADEDNESS .................................................................................. 93
3 .4 . OTHER THEORIES ......................................................................... 102
3 .5 . S U M M A R Y A N D CONCLUSIONS ........................................ 103
CHAPTER FIVE ... COM PARISON W ITH OTHER THEORIES ... 152
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER S IX ...X -3 A R BRACKETS AND M A LA YA LA M 169
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
xiii
PREFACE
The proposal put forth in this thesis presents a new analysis for an old
Chom sky and Halle (1 9 6 8 ) and the theory of Lexical Phonology of Mohanan (1 9 8 6 ).
reconsider the treatm ent of several aspects of the theory of morphology. This
morphological juncture, the X-bar theory and the order of attachm ent of the
different kinds of affixes. Currently held attitudes about these m atters are
questioned and a new analysis is proposed. This new analysis makes it possible to
necessary to discuss the distribution of affixes, the X-bar theory, and previous
theories of juncture. This becomes a rather complicated task which requires the
clarify this, this preface outlines the basic problems and theories to be addressed in
The new theory of juncture proposed here accounts for the facts of variable
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
M ohanan (1 9 8 6 ). All of these previous theories have been w idely criticized yet no
difficulties.
other theories by isolating and solving a problem in the data th a t other theorists
have been forced to include in their accounts of juncture. Particularly, this theory
included an account of the order and distribution of the subgroups of English affixes
(the Class I and Class II affixes) as part of the theory of morphological juncture. It
affixes contribute to the variations in rule application, then the theory of juncture
should also account for the order and distribution of these affixes.
The present proposal explains the order and distribution of the Class I and
Class li affixes outside the theory of juncture by demonstrating th a t these facts are
level X-bars to word-level m akes it possible to improve the generality and precision
of the word form ation rules th at attach affixes. Furthermore, it is possible to use
the word-level X-bar theory to predict the application of the various, phonological.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
XV
morphological junctures.
analysis for the facts of the distribution of the Class I and Class II affixes; it
morphological juncture based on the word-level X-bar theory. The last of these
proposals is the m ost important and the focus of this dissertation, but it crucially
theory is not possible w ithout the reanalysis of the Class I and Class II affixes, and
the new theory of juncture depends on the new version of word-level X-bar theory.
Previous attem p ts to extend X-bar theory to word level are either significantly
include a description of the Class I and Class II affixes, or they do not address the
issue of the Class I and Class II affixes at all (Walinska de Hackbeil 1 9 8 5 , Jensen
accounting for a large amount of the data; that is, they arrive at a description of
some miscellaneous problems of morphology but these theories still require some
the literature and theories of these three distinct areas. Thus the discussion of the
organization w e have chosen for this thesis begins w ith a presentation of the data
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and the history of the discussion of morphological juncture. This first chapter
describes the theoretical preliminaries of the present thesis and gives a history of
the three theories th at seek to explain the facts of morphological juncture. Much of
the history of the treatm ent of the Class I and Class II affixes is mixed in w ith the
description of the history of this problem at the end of this chapter. In the
following chapter w e present the analysis of the Class I and Class II affixes and
propose a new version of the locality principle that accounts for the order and
distribution of these affixes independently of the theory of juncture and the X-bar
theory. Once the Class I and Class II affixes are described, the next chapter
theory of morphological juncture based on the distribution of the Class I and Class II
affixes. The last tw o chapters of this thesis present a discussion of the different
this theory is capable of describing the facts of at least one language other than
English.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER ONE
REVIEW OF LITERATURE
1 .1 . INTRO D UCTIO N
Chom sky and Halle (1 9 6 8 ) and the theory of Lexical Phonology of Mohanan (1 9 8 6 )
account for this data a t a rather high cost to the theory of grammar and are subject
to considerable criticism. The proposal presented in this thesis can account for the
theories and w ith o u t the addition of complex theoretical mechanisms. In fact, the
theory proposed here actually lessens the amount of formalisms that are required.
The new theory of juncture proposed in this theory is based on the extension
of X-bar theory to morphology. Other theories have proposed X-bar fram ew orks for
morphology but w ere either unable to formulate an X-bar fram ew ork th at was
consistent w ith phrase-level X-bar theory or w ere unable to account for all the facts
of morphology. The X-bar fram ew ork presented in this thesis is both consistent
w ith phrase-level X-bars and able to account for all the morphological facts.
review the theory of juncture; second, it is necessary to discuss the facts of English
affixation in order to illustrate and solve the main problem in formulating a theory of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2
juncture: the first, is the problem of the status of the different elem ents placed
phonetic realization, but have been unable to avoid them except w ith the equally
asym m etrical distribution of the tw o subgroups of affixes in English (the Class I and
that makes it very difficult to predict their distribution. The Class I affixes attach to
both bound stem s and free stem s while the Class II affixes attach only to free
stems. The overlap of distribution of the tw o groups of affixes in the case of free
theories have discussed the account of the Class I and Class II affixes, but they
have not noted th at the asym m etrical distribution com plicates the formulation of a
theory. This problem is brought out more clearly in the review of the previous
This first chapter outlines the data to be considered, review s the literature,
and describes the obstacles to the formulation of a theory of juncture. In the last
part of this chapter it is possible to discuss the facts of the Class I and Class II
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3
affixes in a m anner th at illuminates the difficulties these affixes pose for the theory
of juncture.
morphological junctures on phonology is the contrast betw een night rate, nitrate
and nye-trait (Trager and Smith 1 9 5 6 , Aronoff and Kean 1 9 8 0 , Stanley 1 9 7 3 , etc.).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
characteristics of their own but are signalled phonetically only by their effect via
scope from structuralism through the American Descriptivists to the most current
W hile the theories of the structuralists and of generativists like Chomsky and
theories attem pted to account for this data w ithout the use of boundaries or
could replace boundaries and Mohanan (1 9 8 6 ) argued that dividing the lexicon into
multiple levels could explain the facts. W e return to a discussion of these theories
in 1 .2 .2 .2 and 1 .3 .2 .0 .1 .
fram ew ork of the theory of generative linguistics. The work presented here is
M IT and other institutions. H ow ever, the formulations provided here are also
compatible w ith other theories that use the X-bar fram ew ork such as Lexical
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
This thesis accepts the w ork of Aronoff (1 9 7 6 ) as the basis of the theory of
(W BH), and the notions of base, stem , and W ord Formation Rule (W FR). These are
defined in Chapter 2 as they are not directly relevant to the present discussion.
theory m ust account for; 1) the application of stress rules, 2) the application of
morphological rules th a t attach affixes and form compounds and the correct
distribution and order of the different types of affixes w ith the different kinds of
stem s. Some theorists argue th at certain rules of syntax and sem antics also need
to refer to these same boundaries, but the use of boundaries by these components
of the grammar is not w idely accepted and is also of a rather limited nature. For
reasons of length this thesis is limited to a discussion of the problems that occur in
morphology and phonology. This thesis also limits itself to a discussion of English
and M alayalam examples. Of course, w e assume that the theory is applicable to all
sensitive rules are chosen primarily from the discussions of Chomsky and Halle
next.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
based on w hether or not certain rules are either triggered or blocked by their
presence. These are the Class I and Class II affixes. Typical examples of Class I
affixes are: in-, -ity, -ic; typical examples of Class II affixes are: un-, -ness, -less
(Siegel 1 9 7 4 , Selkirk 1 9 8 2 , and Mohanan 1 9 8 6 ). One exam ple of this is the w ell-
known rule of stress shift in English. Example (1) below illustrates the fact that
stress shift occurs in words w ith Class I affixes but not in words w ith Class II
affixes. The affix -ity is a Class I affix and the a ffix -ness is a Class II affix.
/
. *. / .
curious curiosity curiousness
In these form s stress shifts to the right in the presence of the Class I affix -
ity, but stress does not shift to the right in the presence of th e Class II a ffix -ness;
th at is, the rule th at shifts stress to the right only operates w hen Class I affixes are
present. Thus the application of the stress rules needs to refer to different classes
of affixes. Other rules of stress also need information about the hierarchical
correctly triggered and blocked. In some cases the rules are sensitive to differing
morphological structures and in others the beginnings and ends of words are
relevant to the application of a particular rule. These are illustrated in (2) and (3).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7
the variable deletion of the stem final /g / in pairs like, loq ~ loqger, and log ~
logiq.
g - > 0 1 ___#
R2 R2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 .1 .5 . Morphological Rules. Morphological rules also need to be sensitive to
morphological juncture. This is discussed in some detail in Chapter 2 , but for the
present, it is enough to point out that the morphological rules that attach affixes
1 9 7 6 , Allen 1 9 7 8 ). For instance, affixes need to know a t least the category of the
word to which they attach and inflectional affixes need to distinguish the difference
betw een ends of words internal to compounds from ends of words. This is
necessary to insure th at the truck of firetruck not the fire receives a plural marker
confusing since the theories define these terms differently. The terms in question
are: juncture, boundary, form ative, morpheme, and morph. I begin w ith a
There is some ambiguity and overlap that has occurred in the literature
regarding the terms juncture and boundary. The ambiguity and overlap result from
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9
included in the list of phonemes. Later linguists (Chomsky and Halle 1 9 6 8 , Stanley
1 9 7 3 , Siegel 1 9 7 4 , etc.) used the term boundary to refer to their conception of this
phenomena. For these authors, boundaries are introduced either in the lexicon or
by rules w hich operate on the surface structure syntactic string. Chomsky and
Halle (1 9 6 8 ) actually use the term s juncture and boundary synonymously (Chomsky
This paper uses the term juncture to refer to the shared edges of
concatenated morphemes; that is, I simply use the term to refer to the meeting
place of any tw o morphemes of any sort. The word boundary is used to refer to
Both term s are used in order to discuss the different w ays of describing the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
brackets (this thesis). Thus, in this thesis, the term juncture is used to refer to the
'boundary' is used to refer to the particular theoretical construct that has been
The term s morph, m orphem e, and form ative all occur throughout the
literature and require a statem ent as to their use in this thesis. From the tim e of
the early structuralists the term s morph and morpheme referred to the minimal units
and morph is used as the morphological equivalent of a/lophone. For exam ple, the
plural markers Iz l, is/, and Idzl are morphs of the plural m orphem e /z/. In later
discussions the term a/lomorph is used. Chomsky and Halle introduce the term
form ative which is equivalent to m orphem e. This thesis refers to the term
morpheme as in conventional use. This thesis uses it throughout and avoids use of
Finally, the term s 's te m ', 'bound stem ', 'free s te m ', and 'w o rd ' are used
The term 'stem ' refers to any morpheme that a word form ation rule can a ffe c t. A
'bound stem ' refers to stems th at cannot stand as words in the syntax. This
includes words like to/er- of tolerable and requite of requited. A 'free stem ' refers
to any morpheme that can stand as a word in the syntax. This refers to both
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
11
simple, monomorphemic items such as table or sw im and com plex words such as
tolerable or theatricality. W ord and 'free stem ' are coterm inus but the term 'free
stem ' is used w hen a morphological process is being discussed such as 'the prefix
un- attaches only to free s te m s ...' The word 'w ord' is used in all other
distinguish 'bound stem s' which are potential words, such as requite of unrequited,
from 'bound stem s' w hich are not potential words, such as to/er- of tolerable. This
morphological juncture.
the boundaries provided by certain theories; this is because in some theories these
segm ents, w hile at the same tim e indicates that these boundaries are seen as
present in such a manner th at rules can refer to them . Segm ent-like is formally
defined in (6).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
tike w hen it has the property of being a unit of structure but lacks any phonetic
characteristics of its ow n.
present. This section is divided into three subsections: 1) the structuralists, 2) the
boundaries. W ithin each of these subsections the theories discussed are ordered
chronologically.
not attem p t to explain these phenomena and simply introduced the junctures as
parts of the phonemic string. Throughout the history of this discussion linguists
morphological juncture. Specifically, these linguists argued that the nature of the
theory required that the different components of the grammar be rigidly separate
(c.f. Aronoff and Sridhar 1 9 8 3 ). Thus in order to allow the phonology access to
string.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
13
phenomena; 1) are these junctures segments of some sort, and if they are, why
Aronoff (1 9 8 0 ), discussing the developm ent of these ideas among the American
anything th at could a ffe c t the phonology had to be part of the phonological string.
This w as their working assumption but also 'their greatest problem' (Aronoff
1 9 8 0 :3 1 ). The junctures had an obvious connection w ith the morphology and syntax
but they w ere unable to express this. This forced the conclusion that the boundaries
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
w ere phonemes of some sort even if they did not have any phonetic features of their
ow n.
This problem of the phonemic notion of the junctures persisted until Chomsky
and Halle (1 9 6 8 ) rejected the strict notion of autonom y of the com ponents and simply
categories of the syntactic tree structure and the labeled brackets of the phonemic
1 .2 .2 . The Generativists. This section discusses the developm ent of the theory
of juncture w ithin the theory of generative linguistics. There are three proposals that
seek to explain these phenomena in generative linguistics: 1) the word boundary theory
in th at order.
SPE) w ere the first to propose a theory of word boundaries in generative linguistics.
Their theory has been used in the majority of generative studies since it w as first
introduced either by a series of insertion rules or in the lexicon, and w ere then modified
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15
... of a sentence is precisely the proper bracketing of a string of fo rm a tiv e s w ith the
bracketed substrings (the phrases assigned to categories selected fro m a certain fixed
(6)
s
i
—I
VP
I
i
V NP
if
JL
I r
STEM
I I
+we+ +establish+ +pa&t+ +tele+ +graph+ +ic+ +communic ate+ +ion+
(6) illustrates the principle of bracketing that is assumed in SPE and this
labeled string served as the basis for the rule th a t inserts the word boundaries. The
' + ' boundaries w ere present in the lexicon. The boundaries w ere later inserted
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
16
(7)
# # ft + w e + # # # # # + establish + # + p a s t+ # # # # + te le + + g ra p h + #
+ ic + # # # + communicate + # + io n + # # # #
(8)
... a string of form a tiv e s (one or m ore) contained in th e c o n tex t tttf # # and
containing no occurances o f # # .
SPE theory uses three boundaries; and ' = '. These boundaries are
##).
... each lexical category (e.g. noun, verb, adjective) and each category th a t dom inates
a lexical categ ory (e.g. sentence, noun phrase, verb phrase) au to m atically carries a
boundary sym bol # to th e le ft and to the right of the string th a t belongs to it (i.e ., th a t
SPE's (4) and (5) are repeated here above as (6) and (7). Later (p. 3 6 1 ) they
reformulated this definition but its consequences remained the same. The
morpheme boundary (' + ') (formative boundary in SPE terms) w as a part of the
The + sym bols represent form ative boundaries w h ic h , by convention, auto m atically
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17
instance of the main stress rule of SPE in words such as: p erm it (N or V ), concur,
com pel and deter. This boundary w as of rather limited use and w as criticized as ad
separate from one another and from phonemes. Thus all boundaries w ere distinct
from phonemes because of the feature [-segm ent]. The features for formative
boundary [ + FB] and word boundary [ + WB] distinguished the different boundaries:
mentioned in SPE deleted the occurrences of the boundaries before the phonetic
... the "readju stm en t rules" relating sy n ta x and phonology m ake various other
This passage is cited here not only to illustrate the nature of readjustment
rules but also to give some indication of the level of psychological reality these
gram m atical devices had in the theory of SPE. They also state that there are rules
th at convert '# ' to ' + ' (p. 3 6 8 ). The need for deletion rules indicates, that the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
18
structures, nodes, and boundaries w ere taken as units even if they w ere specified [-
segm ent]; th at is, they had some kind of psychological reality as elem ents in the
string. This point becomes more im portant later in the critique of the different
Some insight into the possible power of the readjustm ent rules is given when
(p. 3 6 7 ),
k n o w of no clear exam p les of this; but there are as w e shall see, rules th a t delete # in
various places.
rules. It w as a simple m atter to w rite any kind of readjustm ent rule given these
hesitation.
On the nature of the boundaries them selves SPE stated (p. 3 7 1 ), 'In our
treatm en t, boundaries are units of a string, on a par in this sense w ith segm ents.'
(labelled brackets) in th a t they play a role in determ ining the application of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19
The rules of phonology in SPE were applied cyclically in complex words; that
is, all rules w ere allowed to apply within every occurrence of '# # ' and '# # '
beginning w ith the innermost and continuing outward until the outer limits of the
approach' to word boundaries based on his investigation of the facts. Stanley first
and 2) 'the w ays in which the phonological rules are sensitive to these boundaries'.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
20
th a t the six different boundaries posited in his 1 9 6 9 w ork on Navajo must also be
ranked hierarchically.
lexical items w ith complex internal structure, boundaries w ere listed in the lexicon
w ith the word (the only boundaries assigned in the lexicon in SPE are the form ative
based on the surface structure of the sentence (though his rules differed from those
of SPE1; and 3) Stanley, again like SPE, had an elaborate set of adjustm ent rules to
While S tanley's paper added much needed theoretical detail to the theory of
word boundaries proposed in SPE, it also suffered from some of the same problems
as SPE. Particularly, the readjustment rules w ere remarkably powerful and made
the theory unfalsifiable. Stanley's decision to limit the insertion of word boundaries
to lexical categories but not to phrases probably made this theory incapable of
dealing with certain rules of stress, but it would have been easy enough to reextend
the word boundary insertion rule to include phrasal categories w ithout damaging
Stanley's approach. Actually, the only substantive changes Stanley made in the
SPE formulation were; 1) the addition of word boundaries in the lexicon for words
w ith complex internal boundaries (this may actually have been SPE's original
1 Actually, he differs by not assigning word boundaries on either side of the higher
level categories such as NP and AP. This is sufficient for the data covered in Stanley's
paper: how ever there are stress rules that need to refer to the NPs as opposed to N's
and these require boundaries
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
21
difference betw een Class I and Class II affixes w ith readjustm ent rules (this w as
probably necessary for an account of these affixes in SPE as w ell, but they did not
important departure from the SPE theory. SPE recognized the role these affixes
played in the application of rules, but the method of assigning boundaries to the
affixes w ere attached to a bound stem the appropriate ' + ' boundary would be
present from the lexicon and the rule that inserts '# ' would not apply because
bound stems do not receive category labels. The Class II affixes w ere preceded by
a '# ' in all cases because they always attached to free stem s (not bound stems).
H ow ever, in those cases where the Class I affixes w ere attached to free words
they would be preceded by the '# ' boundary just like the Class II affixes. This was
not explicitly discussed in SPE and there were some problems w ith it, but the
readjustm ent rules could presumedly be used to avoid problematic cases. The
problem cases would have been those words that were made up of a Class I
morpheme and free stem . These pose a problem because they overlap w ith the
Class II affixes which also attach to free stems. This is discussed thoroughly in
Chapter 2. Stanley's main contribution was his description of the w ays th at rules
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
22
morphology; the theory of case in English and the nature and distribution of word
boundaries. The form er does not concern the present thesis and is not discussed,
the latter is important in the history of the developm ent of the theory of
morphological juncture.
In Siegel (1 9 7 4 ) and siegel (1 9 8 0 ) she argued that the ' = ' boundary of SPE
could be eliminated by reanalyzing the rules of stress in the case of lexically derived
words. A fter establishing th at there are only tw o types of boundaries, she argued
th at the Class I and Class II affixes w ere marked as such in the lexicon by the
inclusion of the ' + ' boundary w ith Class I affixes and the boundary w ith the
Chapter 2; for present purposes, it is enough to note that this is a departure from
the SPE treatm ent of ' + ' and that leads to a simpler account of the stress rules
of English.
There w ere three reasons that Class I and Class II affixation needed to be
separated: 1) the affixes separated into tw o groups based on the facts of their
subcategorization (Class I affixes attached to both free words); and bound stems
while Class II affixes attached only to free words. 2) the Class I affixes w ere never
found outside of Class II affixes, which suggested that all Class I affixes w ere
attached prior to all Class II affixes. 3) certain phonological rules w ere either
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
23
to a w ord. Thus, in order to account for the distribution of these affixes, Siegel
Siegel pointed out th a t Class II affixes w ith certain exceptions did not attach
to stem s, w hile Class I affixes attached to either stems or words. She says.
She therefore presented the first proposal for the reordering of the lexicon
(9)
A. In English, Class I a ffix a tio n precedes Class II a ffix a tio n .
B. T h e cyclic stress assignm ent rules fo llo w Class I a ffix a tio n and precede Class II
a ffix a tio n .
form ation in morphology and the manner in which these principies interact with
other components of the grammar. Allen's thesis presented four proposals for
called the 'Strong Boundary Condition;' 2) a principle for the ordering for
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
24
the application of word form ation rules called, the 'Adjacency Condition'. While
the notions expressed in (3) and (4) are pertinent to the discussion of morphology
in Chapter 2 of the present thesis, they do not contribute to the discussion of the
description of morphological juncture. Numbers (1) and (2) how ever, are of
Hypothesis provided some of the groundwork for dividing the lexicon into strata as
( 10 )
constraint insured th at phonology and semantics w ere able to refer to the same sort
manner that w as slightly more constrained than readjustment rules of SPE (p. 7 5 ).
W e say 'slightly' more constrained because the rules of affixation had to be applied
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
in a similar manner in each instance that they w ere applied, how ever, there w as
little independent motivation for the type or number of boundaries that w ere
(11)
Concerning the Strong Boundary Condition for Negative Prefixes Allen says
7 4 -5 :
'u n -', and 'n on -' in each of three com ponents of the g ram m ar; th e phonology, the
be m ade simply by listing inform ation about the phonological, m orphological and
w ould not expect to find any sys te m atic correlations b e tw e e n th e d iffe re n t aspects of
th e behavior of each prefix. But w e do find s y s te m a tic ity ... C o nsequently, I claim th a t
prefix...B oundaries seem to be good candidates for th e d istinctive 'stru ctu ral p ro p erty'
of negative p refixes...
( 12)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
Thus while Allen did not replace any of the theoretical mechanisms
principles of word form ation. He described the facts of morphology in English and
proposed a theory to account for them . His theory did not accept the description of
word boundaries in Siegel (1 9 7 4 ) and sought to form ulate the definition of word
(13) (Aronoff 1 9 7 6 :8 2 )
(A) T here are as m any types of boundaries as there are points in phonological
the affixes. For exam ple, he argued that there must be tw o affixes in -able one
associated w ith the ' + ' boundary and the other associated w ith the boundary.
The evidence cited in Aronoff is too complicated and lengthy to be repeated here,
h ow ever, he used evidence from stress, allomorphy, and truncation to support the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
27
dem onstrated th a t the differences betw een the tw o -able's is not 'one of kind but
of degree', and therefore th at the 'sound and the meaning of the boundaries are not
have no substance.' For a detailed discussion of this data the reader is referred to
A ronoff (1 9 7 6 ).
Thus A ronoff disagreed com pletely w ith the structuralists w ho thought that
these item s w ere phonemes and it seems th at he disagreed w ith the SPE position
th a t these elem ents are [-segment] units th at are present in the phonological string.
If I interpret A ronoff correctly he seems to say th at the boundary elem ents placed
in the string are m erely useful mnemonic devices for the linguist to encode w h at a
m orphemes.
investigation of m atters of juncture. Articles th at are relevant to this w ork are cited
boundaries w as w idely used but it w as not w idely accepted. The main complaints
removed by the power of the readjustm ent rules and the segment-like nature of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
28
theory of w ord boundaries of SPE but none of them w ere able to replace the
readjustm ent rules or the problems caused by the segment-like nature of the
boundaries.
1 .2 .3 .2 . GLOBAL CONSTRAINTS
fundam ental problems w ith the system of segment-like boundaries and proposed an
rules which allowed phonological rules to look back at the derivation. Specifically,
Global Constraints w ere rule specific statem ents about the morphological structure
of the words being affected by the phonological rule. This ability to look at the
derivation (the tree structure) gave the phonological rules the same information that
the boundaries did, but it avoided the problems of the w ord boundaries.
According to Pyle the use of Global Conditions avoided all of the above
problems and still accounted for all of the relevant data. W hile Pyle raised some
in them selves w ere not sufficient to dismiss boundaries and adopt Global
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
29
1 .2 .4 . LEXICAL PHONOLOGY
com pletely diverged both from the word boundary fram ew ork as described in the
Mohanan began his book w ith a discussion of the order of attachm ent of
affixes in English. He accepted the division of the lexicon into ordered levels as in
morphological process....
Mohanan sought to explain the distribution of the Class I and Class II affixes,
and its relation to compounding and inflection by dividing the morphology into four
levels. In particular this theory proposed that the rules of morphology w ere divided
into four blocks of application represented as the four strata of the lexicon and that
these four blocks were sequentially ordered. A fter motivating the first tw o strata
based on the facts of the Class I and Class II affixes, Mohanan then demonstrated
that it w as possible to account for all the distribution and ordering facts of English
morphology using stratal theory. Mohanan also claimed that the word boundary
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30
phonological rules to different blocks. Part of the motivation for replacing the
theory of word boundaries w ith the theory of Lexical Phonology w as the fa c t that
and their insertion by rules and their rearrangement by readjustm ent rules appeared
ad hoc and unm otivated. Mohanan claimed th at the ordering of the rules into the
different strata avoided these problems. Mohanan, like SPE, assumed th at bound
stem s, w ords, and affixes all received individual listings in the lexicon and th at the
Class I and Class II affixes w ere marked as such. A fter the application of all the
rules of a particular stratum have applied, the brackets w ere erased; that is,
Mohanan posited four strata and a postlexical level for the rules of English
(Mohanan 1 9 8 6 :2 8 ).
(14)
Stratum 3: Compounding ##
Stratum 4: Inflection ## ##
applied any tim e the phonemes of their structural description w ere present
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
31
regardless of w h eth er or not any boundaries w ere or were not present. This level is
of SPE as in the right column in (1 4 ). SPE did not accept the notion of a
The correct order of the Class I and Class II affixes w a s guaranteed by the
attach m en t of all Class II affixes after all Class I affixes. M ohanan's four strata are
given in (1 5).
( 15)
did not justify these levels, he merely accepted them on the basis of Siegel's
original suggestion. Siegel's argum ent in favor of this division was solely based on
the fa c t th at the order and distribution of the Class I and Class IIaffixes was
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32
predictable. She fe lt that by merely noting this she missed a generalization about
the grammar; consequently, she claimed that this predictable order of the affixes
affixation into the first tw o strata. Since Class I derivational affixes do not attach
third stratum . Also the rule of 'stem final tensing' (Mohanan 1 9 8 6 :2 6 ) in some
the fourth stratum because inflectional endings attach to all words of all previous
Mohanan (1 9 8 6 ) correctly accounted for the main facts of the distribution of the
the word boundaries com pletely. In order to account for the application of the
application of these rules within the lexicon was accomplished by assigning the
different phonological rules to particular strata as their 'dom ain' of application. The
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
33
T hat is, all words had to pass through all strata in sequential order, and further,
once a word had passed through one level it could no longer return to th at level.
Mohanan states (1 9 8 6 :4 8 );
... a form is scanned fo r rule application a t every lexical s tra tu m , independently of the
num ber of brackets it contains. Thus, even a m onom orphem ic form in English passes
through fiv e cycles of rule application, four lexical and one postlexical.
(attached at stratum 3) are affixed w ith Class II affixes (attached at stratum 2). In
order to account for these problematic cases Mohanan (1 9 8 6 ) used a device called
Mohanan described the loop as, 'A loop is a device that allows the output of
stratum n to move back to stratum n -1 .' He does limit looping to adjacent strata
but he leaves open the possibility of more powerful looping if the evidence should
require (Mohanan 1 9 8 6 :5 2 ).
ungram m aticality (Selkirk 1 9 3 2 , Aronoff and Sridhar 1 9 8 3 ). In this word the Class
I affix -ity has been attached after the Class II affix un- . Since Class I affixes are
impossible for this to occur. Since there is nothing unusual about the attachm ent
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
34
of -ity to ungram m atical this is a clear counterexam ple to the claims of Lexical
boundaries could be eliminated in Mohanan w as a point in its favor. W hile the word
boundaries presented in SPE w ere useful to account for the data, they were
A ronoff and Kean 1 9 8 0 for discussion). W hile the theory of Lexical Phonology
some rather serious problems of its own. Specifically, critics have charged that the
subdivision of the lexicon into different levels w as ad hoc and counterintuitive. The
In sum m ary, the main reason for the separation of the lexicon into the four
strata w as the facts of the distribution of the different kinds of affixes w ith respect
to the sorts of stems they affected and the rules th at applied to them . Particularly,
the contrastive distribution of the Class I and Class II affixes and the predictable
order of attachm ent of these tw o affixes motivated the first tw o strata. The latter
cases where the strata theory broke down (Mohanan 1 9 8 6 :5 0 ), the device of the
loop allowed a return to earlier levels. Finally, there w as also a postlexical level that
contained all the phonological rules that did not refer to morphological information.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
35
morphological structure.
1 .3 . S U M M A R Y A N D C O NCLUSIO NS
discussion of the developm ent of the theories of morphological juncture, the central
role of the Class I and Class II affixes became apparent. The discussion of the
different groups of affixes goes back to the structuralist days, but it w as not until
affixes. Before this they w ere merely referred to as the stress neutral and stress
recognized, but early structuralist accounts did not need to consider them since the
junctures they postulated w ere phonemes th at could be posited whenever the rules
indicated. SPE assumed th e readjustment rules would take care of any difficult
cases th a t might occur and thus did not discuss them . Siegel (1 9 7 4 ) fixed the
boundaries of these groups and proposed a partial listing of them . She also
proposed the first division of the lexicon based on the behavior of these affixes.
Besides playing a role in the application of phonological rules, the attachm ent
of these affixes raises questions about the nature of the word formation rules and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
There are three main theories th at seek to explain the distribution of the
affixes. The proposals take three different approaches. SPE sought to explain the
betw een the various morphemes. In the SPE theory the fact th at there w ere tw o
not discussed in SPE, but Siegel (1 9 7 4 ), working within the SPE fram ew ork,
argued th at the order needed to be described. This resulted in the first subdivision
of the lexicon.
bar theory. In her theory the separation of the affixes into tw o groups w as simply
stipulated, but she claims that the order of the tw o Classes of affixes relative to
one another follows from her theory. In Selkirk's theory morphemes are divided
into 'roots' and 'w ords' and 'affixes'. The Class I affixes attach to 'roots' and
Class II affixes attach to 'w o rd s'. The fact that Class I affixes never attach outside
of Class II affix is thus a result of the fact that the attachm ent of Class II affixes
alw ays creates a word which creates a form that is unacceptable for the Class I
rules. For Selkirk all the 'w ords' which take Class I affixes are simply converted to
'roots' by a rule. This rule which converts words to 'roots' is circular and ad hoc.
In a word like lucidity (see exam ple (1) above) the word lucid has undergone the
'w o rd ' to 'root' in order to be available for the attachm ent of -ity. In the case of
lucidness the word lucid has not undergone the 'w o rd ' to 'root' rule. The 'w o rd ' to
'root' rule does generate the correct results, but it is clearly ad hoc and circular.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
The theory proposed in this paper does not have this problem, since the attachm ent
The theory proposed in Chapter T w o predicts the facts of the Class I affixes
w ithout either a rule like the 'w ord' to 'root' rule of Selkirk (1 9 8 2 ), or the need to
subdivide the application of the rules into sequential strata. The facts of
From the preceding discussion it is apparent that the role of the distribution
of the Class I and Class II affixes has been discussed but the details of the problem
have not received much attention. The facts of the Class I affixes are the main
theories proposed so far has satisfactorily accounted for the data. The SPE
fram ew ork w as able to account for this data but it did so w ith the overly powerful
have solved the problem w ith the different strata, how ever, there w as problem with
that besides marking all affixes as to the stratum at which they attach, these
theories had to also mark all words; that is, the words th at are affixed with Class II
affixes had to be marked in some w ay to prevent them from being affixed at the
first level (the Class i level). If the words that take Class II affixes must be marked
anyw ay, it would be better just to mark all words as to which class of affixes they
take rather than subdivide the lexicon. In a theory th at does not subdivide the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
lexicon the affixes are in competition for any one form and thus the need to mark
the w ords does not occur. How ever, in the theories w ith the divided lexicon all the
Class II affixes are unavailable at the first level, so some kind of marking is required
to m ake sure those th at do not take Class I affixes do not. The words must be
marked for particular strata in order to prevent their affixation at level one.
accounted for and 2) the history of the theories that are form ulated to account for
the asym m etries in the distribution of the Class I and Class II affixes.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER T W O
2 .1 . INTRODUCTION
The discussion in the previous chapter indicated that the major obstacle to
distribution of the Class I and Class II affixes. This chapter presents an analysis of
the Class I and Class II affixes th at is independent of the theory of juncture and
morphology it is possible to explain the fact that the English affixes separate into
tw o groups and th at the order of these tw o groups of affixes relative to each other
words, the WFRs must be sensitive to the category of the word to w hich it
from the distribution of the affixes un-, in-, -m ent, -ary and noun forming -al
miss a generalization about the distribution of these affixes. Since the iocality
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
(1 9 7 6 ). The terms 's te m ', and 'b ase', and the notion of the 'W ord based
hypothesis' are accepted w ithout modification. These term s are repeated here as
by applying a regular rule to single already existing w o rd s . Both the n e w w o rd and the
The base of any word is the w ord from which it is derived. (Emphasis
added.)
For any word, th at which consists of the word minus the affix is the
stem.
In addition this paper introduces the term 'fragm ent' to distinguish a bound
'stem ' that is not a potential word from a bound stem which is a potential word.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
41
'potential w ord' is assigned a category in the lexicon, while the bound stems that
are not 'potential w ords' are not assigned a category. I take the term 'potential
(4) Fragment.
Those stems which have neither a meaning nor a lexical category as in (5).
(5)
toler able
in im it able
sub m it
serendip ity
simple) and 'analyzable' (synonymous w ith complex). These are defined in (6) and
( 8 ).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
42
clear
even
fair
W innebago
W eyauw ega
secure
toler able
sens itive
w ait ed
black board
dis band ed
In a theory th at accepts the W BH, words made up of a fragm ent and an affix
such as tolerable are entered into the lexicon in tw o w ays. In some cases these
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
words are borrowed into the language from French or Latin. In other cases WFRs
operate on the borrowed forms and are followed by truncation. For example,
assuming th a t tolerate is the form that is borrowed from Latin, the word tolerable is
created by the WFR th at attaches -able operating on tolerate and follow ed by the
( 10 )
general m anner:
1 2 3 ==>
1 J6 3
w h e re X and Y are m ajor lexical categories.
All th e rules of truncation w hich I have found in English apply e x a c tly like the
abo ve schem a; th a t is, they apply before su ffixes, and only w ith + boundary affixes.
The introduction of the notion 'fragm ent' along w ith the Im m ediate
accurately then current theories. The separation of affixes into Class I and Class II
which the Class I affixes do not have. The order of the tw o groups of affixes can
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Class I affixes are those affixes which can be attached to w ords th at are
made up of either a fragm ent and an affix, or words th at are made up of a stem
(11) ( 12 )
/
F
A\ af af N af af
the attach m en t of a Class II affix, (13)b illustrates the affects of the Class II
restriction and (1 3)c dem onstrates a WFR that attaches un-. This WFR includes the
Class II restriction that prevents a WFR from attaching the affix to a word
(1 3)a (1 3)b
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
(14 ) - (1 7 ) provide some examples of Class I and Class II affixes. This list is
partially taken from Siegel (1 9 7 4 :1 0 7 ) and Selkirk (1 9 8 2 :9 4 )'. The suffix un- is
included in the Class II affixes in this list. Siegel and Selkirk do not include this
(1 4 ) Prefixes
(1 5 ) Suffixes
Class I: l-ary], [-al], [-m ent], [-ous], [-ity], [-ive], [-ate], [-ory].
[-al], [-ify], [-ate], -ion], [-ic], [-able]3, noun forming [-y] ...
' There is some disagreement about these affixes. For this thesis I accept these
as a starting point. I am aw are of no w ork that makes an exhaustive and definitive list
of either of the tw o groups of affixes.
2 These exam ples are listed in conventional English orthography. The brackets are not
m eant to indicate a transcription.
3 Following A ronoff (1 9 7 6 ) w e separate the affix -able into tw o . One which is Class I and
the other Class II. The argum ent for this separation is quite com plicated but is not crucial to
the present discussion. The reader is referred to A ronoff (1 9 7 6 ) for a complete discussion.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
46
(16) Prefixes
Class I Class II
(17)
Class I Class II
2 .2 . LOCALITY RESTRICTIONS
Principles of locality are designed to restrict the ability of rules to refer to the
internal structure of the items they affect. A principle of strict locality is one which
principle of strict locality would prevent a WFR from referring to any of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
47
(18)
u n ^ J^likeJ^] able ] ]
A V V af af A
(19)
in [ [ comprehense ] [ible] 1
af A
any of the information that is contained in the internal brackets. For exam ple a rule
which attaches the affix un- to a word can only state that this affix attaches to
adjectives4. The WFR th at attaches the affix in- also can only refer to the fact that
correct it does not allow one to capture all the facts about the regularities of the
distribution of un- and in-. The information th at (19) is a latinate w ord and thus is
negated by in- rather than un-is not available in a word like incomprehensibility
because the principle of strict locality prevents the WFR from referring to the word
internal feature [ + latinate] of the word comprehend. The type of affix which
4 I am only discussing un- which negates an adjective. The un- which contradicts the word
to which it attaches as in unbind and attaches to a variety of categories is n o t being considered
here.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
48
creates the adjective (in com plex words) can not be relevant information for the
W FR nor can the category of the word contained within the complex word nor any
other information contained inside the outermost brackets. The only information
th at is available to the WFR in a theory that assumes strict locality is the outermost
Using examples similar to the case of un- and in- just discussed, authors
allow WFRs to refer to tw o levels of structure rather than just one. In this theory a
distribution of the Class I and Class II affixes is the fact th a t the Class I affixes
attach to both fragments and stems. Class II affixes attach only to words. If Class
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
49
I affixes attached only to fragm ents and Class II affixes attached only to stems the
explanation could be w ritten using one rule which referred to 'fragm ents' (in the
case of Class I affixes) and one rule which referred to 'stem s' (in the case of Class
II affixes). Since this is not the case it is necessary to find a means of writing rules
Tables (1) and (2) illustrate the distribution of the Class I and Class II affixes.
Fragments Words
NO YES
(2) shows w hether or not the category of the word to which an affix can attach is
allowed to im m ediately dom inate a word or fragment. The Class I affixes are
divided into Class la and Class lb to portray a variation in distribution that forms a
subclass of the nonneutral affixes. This is discussed in the next section of this
chapter.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
Fragment + af W ord + af
(noncategory) (category)
Class lb YES NO
Class II NO YES
The separation of Class ! into la and lb does not change the membership of
Class I. Class II affixes (the stress neutral affixes) can still be described as those
affixes which have a restriction which prevents them from attaching to a word
described as 'all others', but now this group includes tw o types. Class lb affixes
are those which have a restriction that prevents them from applying to words
Class la affixes are the affixes which have no such restriction. Since the fact that
the subclasses within Class I do not change the distribution of affixes in terms of
stress shift and other phonological rules, the major distinction betw een Class I and
Class II is maintained.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
51
2 .4 . THE IM M E D IA TE D O M IN A N C E C O N DITIO N
In order to account for the distribution of the Class I and Class II affixes as
for the earlier locality restrictions. The Im m ediate Dominance Condition allows the
the word being affixed. The separation of affixes into tw o classes is a result of the
A WFR can refer to the lexical category to which it attaches plus the tw o
For exam ple, in the case of the Class II affixes there is a stipulation th at says
(24) af - > [ [
X Y
(W here Y must be a category)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
52
(25) (26)
\A
af V
' \ af af
I I I
un2 believe able *u n 2 toler able
un- cannot attach to tolerable because in this word the 'A ' node immediately
plus the stipulation for Class II affixes prevents a Class II affix from being attached
in a form such as (2 6 ) where the 'A ' to which it attaches im m ediately dominates a
noncategory5.
The Class lb affixes are those which have a restriction th a t prevents them
good exam ple of a Class lb affix involves words ending in -m ent and followed by --
at, (e.g. ornamental) and words ending in -m ent and followed by -ary, (e.g.
problems for her theory, since the notion of adjacent elements does not allow for an
explanation of the variable attachm ent of -al. This is because the affix -m ent is the
adjacent elem ent to -a l in all cases. These affixes, however, provide a very clear
5 T h e a ff ix in - c a n n o t a tta c h t o t h e fo r m in ( 2 5 ) b e c a u s e , in g e n e r a l, in - c a n n o t a tta c h to
a w o r d w h o s e c a te g o r y im m e d ia te ly d o m in a te s a c a te g o r y .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
53
(27)
a b
c d
In (a) through (d), note th a t -ary and -al can attach to some but not all words
ending in -m ent. Particularly, w h a t w e find is that -ary and -al can attach to words
m ade up of a fragm ent plus -m ent as in (a) and (c), but they cannot attach to words
attachm ent of these affixes must be extrinsically ordered to prevent the generation
include a restriction on the WFRs that attach -ary and -al to forms like F]m ent]N that
dominates a category (as in [N[v employ ]v [„, -ment]a,]N. Thus examples that are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
counterexamples for Siegel (1 9 7 7 ) are actually support for the theory proposed
here. Also the need to order the affixes is unnecessary in this theory.
w as stated before, the Class I affixes are rarely found outside of Class II affixes.
Forms like ungram m aticality make it difficult to say that all Class I affixation
precedes all Class II affixation since in these cases the affix -ity, a Class I affix, is
attached after un- which is a Class II affix. Since un- attaches only to adjectives it
These forms are actually correctly predicted by the proposal put forth in this
paper, un- correctly attaches to gram m atical and -ity can be attached to
ungram m atical because it is not prevented from applying by strictly ordering all
Class I affixes after all Class II. The affix -ity is a Class la affix in the theory
proposed here; th at is, it attaches to both fragm ents and words. It is thus not
w hether or not gram m at should be treated as a fragm ent separately from the word
gram m ar or w hether gram m at should be treated as the form gram m ar th a t has had
the /r/ changed to a It/ by an idiosyncratic phonological process. For this exposition
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
g ram m at is listed as a word. Since this paper accepts the W ord Based
gram m atical be listed as a separate entry in the lexicon along side gram m ar. This is
impossible in a theory th a t accepts the W ord Based Hypothesis because fragm ents
cannot receive a separate listing in the lexicon. This also opens the door to claims
th a t many other rephonologized forms are also fragm ents such as sensu- of
sensual. In th at case sensual would also have to be given a separate lexical entry
from sense. Thus in keeping w ith the W BH w e argue th at sensu- and gram m at- are
rephonologized forms and the lexical entries are gram m ar and sense.There is little
danger of fragments being mixed in w ith these rephonologized words because there
is no lexical entry in such cases; that is, forms such as toler- do not have lexical
(28)
/\
un2 gram m at ical ity 1
Class II affixation (even though forms like ungram m aticality m ake it impossible to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
56
(29)
th at there m ust be some explanation of this almost universal constraint on the order
of affixes in English. They therefore argued th a t the order of attachm ent of the
Given the theory proposed here, how ever, the fact that Class I affixes
seldom occur outside of Class II affixes can be shown to follow from the
the Class II affixes (those w ith the restriction on noncategories) apply before all the
of Kiparsky (1 9 7 3 ). It is not possible to account for the facts of the relative order
of the Class I and Class II affixes w ithout the Elsewhere Condition because of
contrasts like fearlessness - *fearlessity. Both -ity a Class I affix and -ness a
Class II affix both can attach to adjectives, the fact th at ity cannot attach to
attach outside of Class II affixes. Since affixes are allowed to attach to adjectives
there m ust be some other principle that predicts their attachm ent. The Class II affix
-ness has the restriction th at prevents it from being attached to adjectives which
are made up of a fragm ent and affix. The Class II affix therefore has a structural
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
description th a t is more detailed than th a t of -ity and the Elsewhere condition
predicts th at in such situations the rule w ith the more detailed structural description
(3 0 ) (from Kiparsky 1 9 7 3 :9 4 ):
T w o a d ja c e n t rules o t th e form
A -> B / P _ Q
C -> D / R_ S
Since the attachm ent of Class II affixes specifies a more particular structural
description than the attachm ent of affixes w ithout such a restriction, the Class II
affix must be attached first. The examples in (29) above illustrate this. In the
exam ples in (2 9 ) the suffix -ness subcategorizes for tw o levels of structure and the
suffix -ity does not. Thus the suffix -ness has a more specific subcategorization
than -ity.
com petition for application on one form, the rule w ith the more specific structural
description will take precedence over the other. This explains w hy Class II
affixation, which refers to tw o levels of structure, applies before all those Class I
affixes th at refer to only one level of structure. This is a w idely accepted principle
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
restricted to phonology.
The second step in predicting the order of the affixes involves the Class lb
affixes like -merit, and -a!. The Elsewhere Condition does not predict th a t the Class
II affixes will be attached before the Class lb affixes because, rather than being less
specific than the Class II affixes, these Class lb affixes are equally as specific as the
word to which they attach can dominate just like the Class II affixes. Thus, in
these cases both the Class II affixes and the Class lb affixes are referring to tw o
that prevents them from applying to a word whose category immediately dominates
In these cases the Class lb affixes do not follow the Class II affixes because
the attachm ent of a Class II affix creates the structure th at is forbidden in the Class
I affix.
(31)
where the category im mediately dominates a category. Since Class II affixes never
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
59
attach to a fragm ent there will also never be a word made up of a noncategory
followed by a Class II affix. One hypothetical case would be a word like to/erness.
the same sort of internal structure as demonstrated above. One such example is
noun forming -a / which attaches only to fragm ent plus affix combinations. I have
only found tw o counterexam ples to this (from W illiam O 'G rady personal
(32)
dis pose al
re fuse al
re verse al
pro pose al
com m it al
be tray al
re triev al
((33)a). These words contrast w ith a number of other Ns in -m en t that cannot have
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
60
(33)
3 b
V or N N not V
regi m ent
supple m ent
w hether the V or the N is basic, for this theory it does not m atter which is basic,
the words in (a) are made entirely of fragm ents plus -m ent, while those in (b) are
m ade of verbs plus -m ent. It does not m atter which form is chosen as the basic
form (either N or V ), w h a t m atters is that the W FR that derives one from the other
needs to be sensitive to exactly the same sort of internal structure th at in-, un-, -
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
61
It might be argued th at verbs in (33) are actually generated, but they are
blocked or filtered out due to the existence of verbs such as employ or contain.
This, how ever, implies th at words like em ploym ent are actually generated and then
simpler, since there is no reason to consider overgeneration and filters. Only the
correct forms are generated in the first place. The rule of verb formation is blocked
in cases w here there is a full w ord, but allowed in a word m ade up of a fragm ent
plus a suffix. The N - - > V word form ation rule needs to distinguish betw een the
(35)
a b
1I I ]
F N X N
(where X f noncategory)
The facts of the distribution of un- and in- can also be explained by the
Im m ediate Dominance Condition. One point th at is often not clearly made is that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
62
(36)
a b
un buri ed
un wound ed
un health y
adjectives (e.g. from Latin and French) as in (37)a. In analyzable words, in
(37)
a b
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
63
unanalyzable words the prefix un- cannot attach to borrowed simple words or to
fragm ents. In analyzable w ords, un- cannot attach to words that are made up of a
general, in analyzable w ords, in- cannot attach to derived words. W ords borrowed
from Latin and French regularly take in- regardless of w hether or not the word is
As w ith Class I affixes, the distribution of un- and in- can be described in
one statem ent by allowing the rules to refer to the category of the word to which
This distribution can be stated as w ith other Class II affixes; 'The prefix un- cannot
illustrated in (3 8 ).
(38)
*u n 2 ordinF ate
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
64
The 'A ' node in the square is the node to w hich un- or in- will attach . The
underlined nodes are those im m ediately dominated by the squared 'A 's '. I use the
term 'im m ediately dom inates' because I specifically w a n t to exclude the possibility
th at th e W FRs see still deeper levels of morphological structure. That is, when the
affix attaches to the w ord, it sees only the previous derivation. Thus, in the word
intolerable, the in- is correctly chosen for intolerable because the 'A ' to which in
attaches im m ediately dom inates a fragm ent plus a suffix as in (3 9 ). In the word
unm itigated, how ever, un- is correctly chosen as the negative form because the 'A '
(4 0 ). Thus even though the first morpheme of the word is a fragm ent in both
cases, it is not the adjacent elem ent th at is the determining factor here, but w hat
(3 9 ) (40)
2 .6 . COUNTEREXAMPLES
Possible counterexam ples to the explanation offered for the distribution and
order of English affixes would be forms in which a Class II affix attached to a word
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
65
counterexam ples to this part of the principle. H ow ever, there are some
The first counterexam ples are from the prefix in-. This form behaves like an
ordinary Class lb affix in th at it usually attaches to fragm ents and to words whose
category dominates a fragm ent. H ow ever, unlike -m ent, and -al, and other Class I
affixes this affix can in some cases attach to words whose category immediately
in- constitutes som ew hat of a special case among the affixes because many
of the words th at take in- w ere borrowed with the affix; th at is at the tim e of the
introduction of the word into English it w as regularly negated w ith in-. The
counterexam ples w ith in- are the result of the preservation of the historical term.
To support this the examples below are sighted which have changed from taking in
W ords th at have changed from in- to un- to conform to this pattern: (from
Marchand 1 9 6 9 ).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
66
(41)
in changeable un changeable
in pleasing un pleasing
in popular un popular
in charitable un charitable
in certain un certain
in chaste un chaste
in ceremonious un ceremonious
in imaginable un imaginable
in utterable un utterable
in organical un organical
There w ere some counterexamples to un- in the past as well. These too
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
67
(42)
un effable in effable
un perfect in perfect
un possible in possible
un dubitable in dubitable
Thus the counterexam ples w ith in- are archaic preservations and are not
if such exam ples exist they could probably be explained in a similar manner.
order of the tw o groups of affixes. This is the case of words like m onstrosity and
word monstrousness, how ever, is not precluded. The theory proposed in this paper
claims th at the fact th at Class I affixes never occur outside a Class II affix is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
68
Class lb affixation because the attachm ent of a Class II affix creates the structure
Class I affixes attach rather generally (see Table (2 1 )). H ow ever, the form
violate that restriction since the word m onster is not a fragm ent.
(43)
A closer look at the data indicates that form s in -ous take either -ity or -ness.
In general the choice of -ity or -ness depends on the Class II and Class lb
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
69
(4 4 )b (Class la)
(44)d (counterexamples)
account for the fact th at -ness cannot attach to monstrous. T h at is, if monstr- is
viewed as some kind of fragm ent it can block the attach m ent of -ness. However,
this would require that monstrous be given its own lexical entry in order to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
70
differentiate the fragm ent monstr- from the word monster. This extra complication
theory as in (44) a and b, there are some counterexamples. The existence of these
counterexamples follows from the fact that words ending in -ous take either -ness
underlying pattern conforms to the principles discussed in this theory, but the
actual forms are often determined by analogy with -osity or -ousness. This could
also be explained as a case of the affix -ness changing its class membership. There
likely th a t the fa ct th at -ness and -ity are both applicable after -ous has opened the
door for -ness to begin to lose the Class II restriction and become more like a Class
la affix. W hile it is likely that the explanation for this is due to class change,
avoided. How ever, it is safe to predict that Class change, if it occurs, is going to
go from the more specific to the more general as in the case of -ness.
rules limited access to the internal structure of the words they affect. These
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
structure by restricting their access to only 'm aterial im m ediately adjacent to the
a ffix '. Th at is, a WFR can refer to information th at is separated from the affix by
no more than one bracket (45a and b below). W FRs are prevented from 'seeing'
any other information. For exam ple, Siegel (1 9 7 7 ) states th a t there is a condition
on un- attachm ent th a t says th at un- cannot attach to words that contain negative
content in the next syllable but only if this negative content is 'im m ediately
(45) a b
[un [[distinguishled]]
[un [[discoverable]]
9 The bracketing o f these forms is in accordance with Siegel (1977) who follows Chomsky and Halle (1968).
D ifferent bracketing conventions make this principle fail. See Chapter Four o f this thesis for a discussion of
different forms o f bracketing.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
72
She pointed out th at the forms in a, b, and c are ail generated by a WFR in
A ronoff (1 9 7 6 ) and th at the form s in a and b are then filtered out by her adjacency
condition.
H ow ever, given the theory proposed here, only the correct forms are
generated in the first place. Since un- is a Class II a ffix it is restricted from
category and thus it can apply. The trees in (46) and (47) illustrate this..
(4 6) (47)
The W FR w ith the locality condition is sufficient to generate not only the
forms in un- plus dis-, but also forms of un- and in- and other affixes, un- cannot
attach to the form s in (4 5)a and (45)b because the category of these words
un- from attaching to words like those in (45)c but could not prevent un- from
attaching to the words in (4 5 )a and b, as the dis- is not 'im m ediately adjacent to
the a ffix '. These notions are given more precision in Chapter Three.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
73
Condition', w hich also serves to restrict the am ount of word internal structure that
a WFR can refer to. For W illiam s, though, a W FR can refer only to the lexical node
to w hich it will attach an affix. H ow ever, this node can carry information th at has
percolated up from the low er lexical nodes. W hile he sets no theoretical guidelines,
set of features. Thus, he limits the WFRs while still generating the correct output
ablaut] to percolate.
(4 8)
a b
f
[stand] (jf stood P
[ + abl]
[standHed]])
(4 9)
a b
prep V V pst
i
[under] [stand] [[under] [stood] </
[ + abl]
(*[under][stand]edl)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
74
1 9 8 6 :1 7 0 -7 2 ).
1 1 I
Z Y X
Both principles recognize the fact that WFRs need to refer to or recognize
some but not all, internal structure. Scalise (1 9 8 6 ) points out that these tw o
principles m ake different predictions and, as yet, there is not sufficient evidence to
decide betw een them . This paper argues that neither is sufficient to account for
the facts of the distribution of the English affixes illustrated in this paper.
Generalizations about the distribution of the affixes -un, -in, -al, -ary, and -m ent
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
75
2 .8 . CONCLUSION
This chapter dem onstrates that w ith a slight modification in the theory of
locality it is possible to predict the contrastive distribution of the English Class I and
Class II affixes. This same theory also makes it possible to explain the order of the
tw o groups of affixes relative to each other w ithout extrinsic ordering statem ents or
ordered levels in the lexicon. It is no longer necessary to include the Class I and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER THREE
3 .1 . INTRODUCTION
and Class II affixes are best explained in terms of a new principle of locality in
once the distribution of the Class I and Class II affixes is described in terms of the
straightforward manner. Since the facts of the English affixes are described
there is still enough motivation for a word-level X-bar theory. T h at is, previous
theories (Mohanan 1 9 8 6 and Selkirk 1 9 8 2 ) that seek to extend the X-bar theory to
word level do so simply in order to describe the facts of th e Class I and Class II
affixes. Now th at these facts are dealt w ith by the Im m ediate Dominance
th a t needs description to justify the use of the X-bar theory at word level.
able to capture generalizations about the structures of words and the rules of word
formation. A WFR creates new words from existing w ords by processes of 1) zero
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
77
possible words to which an affix can attach and WFRs express the processes of
word form ation. The theory of gram m ar also needs to encode this morphological
information into the phonemic string so th a t the relevant rules of phonology can
apply correctly.
subcategorization and its W FR. A W FR just describes the attachm ent of an affix to
the affix and the structural description of the WFR are identical. The morphology
would be simplified if there w ere simply one W FR, such as A ttach o, and the
This chapter dem onstrates that it is possible to extend the phrase-level X-bar
make this extension of positive value. Specifically, the theory of word-level X-bars
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
78
accounts for the facts of headedness in morphology, and serves as the basis for a
subdivided lexicon. Since this theory merely extends phrase level X-bar theory to
word level, this is not only done w ithout an extra cost to the grammar; it actually
unexplained gap, the gap being th at there is no principled explanation for the fact
should be able to represent facts of categorial organization both within words and
in the syntax. This is especially true since both morphology and syntax represent
components of the grammar th at are based on the organization of lexical items into
hierarchical groups. It is also possible to argue that the fact that X-bar theory in
fact th a t morphology and syntax are indeed different components of the grammar.
morphology, and further, if it captures the relevant facts of word formation, then
this would be a strong argum ent in favor of X-bar theory in general and of the value
would simplify the grammar by replacing theoretical mechanisms that are particular
claim th at the extension of X-bar theory to words provides significant insight into
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
affix subcategorization, the order and distribution of affixes, and the variable
Second, it dem onstrates that this extension of the X-bar theory to morphology can
about the distribution of English affixes that are otherwise missed. This
condition; how ever, the subcategorization frames of affixes and the WFRs are also
improved by this extension of X-bar theory to morphology. The third part of this
X-bar fram ew ork for morphology, but these theories are either significantly different
from the phrase-level theory or they are not capable of handling all the facts.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
80
Kornai and Pullum (1 9 9 0 )1. Kornai and Pullum define seven basic principles to
which any well-form ed X-bar theory in a context free gram m ar (CFG) must
conform. These are: Lexicafity, Succession, M axim aiity, Uniform ity, Centrality,
Uniform ity is the principle which states th at '... the maximum possible bar
level is the same for any preterterm inal.' Since this holds of preterterminals it has
morphology but it is not obvious because in most cases heads are bound forms
such as en- of enrich, if morphemes like en- w ere allowed to be optionally present it
would mean th a t en- could stand as a word on its own. H ow ever, cases of
category change such as [N perm itlN —*• [V perm itlV indicates a case w here a head
is not present.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
81
Centrality concerns the 'start symbol' (such as the 'S ' for sentence of SPE)
Periphera/ity repeated in (1) below states th at the head of X '... alw ays
Definition: A Lexicality-observing CFG observes Peripherality iff in any rule rew ritin g X ’
as Y X °Z , e ith e r Y = e or Z = e.
of any WFR takes one lexical item (complex or otherwise) attaches it to another
lexical item , and creates a new lexical item made up of tw o parts. The one is
alw ays peripheral to the other and thus the head is alw ays peripheral to the X. In
cases of zero derivation since there is only one elem ent Peripherality is meaningless.
The principles which determine the shapes of X-bar structures in the present
syntax ' ... requires all phrasal categories be projections of lexical categories' Kornai
categories, they operate on lexical categories and create other lexical categories
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
require th at W FRs operate only on lexical categories and produce only lexical
categories. Since fragments are not lexical categories the principle of lexicality
predicts th at W FRs cannot operate on fragm ents. This is essentially a restatem ent
and the principle of Lexicality requires th at X-bars project only lexical categories
from only lexical categories. The W BH requires that fragm ents be entered in the
lexicon either as forms borrowed from another language or they are formed from
these borrowed forms by the operation of a WFR followed by truncation. For this
reason the words which contain the fragm ents not the fragm ents them selves, are
involved in word formation and thus the fragm ents are not counterexamples to
Lexicality.
(2) Lexicality.
G V T and i ) 0 .
M axim ality and Succession are more directly pertinent to the assignment of
labeled trees. These principles are repeated here as (3) and (4). Kornai and Pullum
1 9 9 0 :2 8 -3 0 ).
Maximality iff for every rule X" — ►Y X"-' Z, the strings Y and Z are in V M', w h e re V M
= {xm | x e V ).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
83
M axim ality, all lexical categories must be introduced at their highest possible bar
level. Item s listed in the lexicon w ith a category are given a 0 as their maximal bar
level. A ffixes that assign categories are also introduced a t the 0 level but the facts
assigned a -1 . Those affixes th a t are not category assigning are assigned a -2.
m aximal bar level assigned to all lexical entries th at have a category (words and
as their maximal level because they serve as elem ents of the syntactic string.
betw een the -1 and -2 levels for their maximal level. The -2 level is chosen as the
m axim al bar level for these noncategory morphemes because the -1 level generally
marks the head of a word and there is no clear reason to allow the noncategories to
(4) Succession.
rew ritin g som e nonterm inal X" has a daughter labeled X"'1.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
84
requires th at one daughter be labeled X"'1. In morphology there are forms w here
tw o daughter nodes are both of the same category as the whole word as illustrated
in (5) and (6). In these words it is not im m ediately apparent if one or both
(5) (6)
This thesis argues that the morpheme which determ ines the category of the
entire word should be X"'1 while the morpheme which does not should remain
provide evidence for this but section 3 .3 of this chapter on Headedness, and
Chapter Six on the discussion of Malayalam indicates that there are empirical
consequences in applying Succession in this manner. Thus in (5) the noun door is
the modifier of the noun 6 e //th u s cannot determine the category of the entire word.
The noun bell m ust determine the category of the entire word and is affected by
2 This also applies in words made up of tw o different categories as in overw ork. Sine the
compound overw ork is a verb the item th at gives the entire word its category must be work.
Therefore the priciple of Succession operates on the verb but not the preposition as in (A).
(A)
Pj° V;1
over w ork
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
85
succession. In (6) the category is assigned by the affix thus succession determines
this should be the X n' \ The discussion on headedness in section 3 .3 of this chapter
dem onstrates th a t there are cases in w hich Succession applies to both members of
some compounds and both members of some cases of affixation when both
m em bers are of the same category and the assignment of the category is
(7)
0
I I
un believe able
Since all of the morphemes in this example are assigned categories in the
lexicon, the principle of m axim ality requires that they all are entered as '0 's 3. The
'A ' node of believable is not demoted to -1 by Succession because it does not
assign the category to unbelievable. W ithin the word unbelievable how ever, the
affix -able assigns the category to believable and thus the affix is demoted to -1 by
the principle of Succession. The 'V ' node above believe is not affected by
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
86
Succession because it plays no role in assigning the category to the entire word.
The rules of this theory th a t assign the bar levels are phrase structure rules
similar to those of Selkirk (1 9 8 2 ), but should not be taken as replacem ents for the
W FRs presented so far. They should be seen instead as well formedness conditions
rules are:
(8)
X0
a. X 1 a f:
V
1
l
O
X
b. X 1 af*
A
1
1
■<
O
o
X
c.
■ ti
X
A
0)
1
1
d. x ° --> af'2 X
o
X
e. af*'1 X '1
A
l
l
0
f.
X
a f 1 Y°
1
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
87
The theory proposed here has three bar levels as represented in (9).
identical categories.
affixes.
Rules (8) (a) - (c) above are the rules of suffixation, while rules (8) (d) -(f) are
the rules of prefixation. There are three kinds of prefixation and suffixation:
noncategory-assigning (as in (a) and (d)) is the first, and the second and third are
both category assigning, one in w hich the category assigned is the same as the
word affixed (as in (8) (b) and (e)) and the other when the category assigned is
other than the category of the category that is affixed (as in (8) (c) and (f)).
( 10 )
NT1 a f2
king dom
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
88
Rule (8) (b) describes a category assigning suffix attaching to a word of the
(11)
A '1 afA-1
I I
[ [green] [ish] ]
(12)
A0
N° afA-1
I 1
[ [monkl [ish] ]
Rule (8) (d) describes a non-category assigning prefix attaching to word (as
in (1 3 )):
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
89
(1 3 )
I I
[ [counter] [exam ple] ]
( 15 )
a fv-1 A0
I I
[ [en] [rich] ]
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
90
(16)
accordance w ith the principle of m axim ality. Since some affixes assign the
category of the word that is created by their attach m ent these must have
(17)
(18)
th at these morphemes cannot serve as heads and th at they cannot serve as input
to the syntax. The -2 level is the maximal bar level of these morphemes and the
because the -1 level is created by the interaction of the X-bar principles on Os and
because -1s can be construed as heads and neither fragm ents nor noncategory-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
91
These rules indicate the different sorts of structures th at result from the
application of the WFRs. Rules (8) (a), (b), (d), and (e) indicate cases where a WFR
creates a word of the same category as the word it affects. In cases (8) (a) and (d)
marked as 'a f'2'. In other cases ((8) (b), (c), (e), and (f)) the affix determines the
category of the word it creates. Category assigning affixes are marked as 'af*'1'.
The 'Y 0/ indicates that a category has been changed by the operation of a WFR as
in (8) (c) and (e). In those cases where a category assigning affix is attached to a
word of the same category as the category it creates, the principle of Succession
requires th a t the category is -1 rather than a 0 . This follow s from the fact th at the
Tense inflections, which are discussed in Chapter Four are given a 0 bar
level because of their status as lexical entries in the syntax. T h at is because the
tense marker must be introduced into the string as independent item . It is given a
'0 ' bar level like any other item of the syntactic string. A later process of
encliticization attaches the tense marker to the verb. In these cases the tense
marker does not assign the category to the word and therefore it cannot be
agreem ent marker /z/ of the third person present indicative of regular verbs (as in
(1 9)). The tense marker in (1 9 ) is originally an elem ent of the syntactic string under
an 'infl' node and a later process of encliticization attaches it to the verb quit.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
92
Since the category of quit is not determined by the tense m arker, it is not demoted
to a -1 by Succession.
(19)
NP -V°
V°
\ af°
he quit s
category. T hat is, in the theory of X-bar brackets the main function that
distinguishes one word from another is the difference betw een internal junctures
created in the case of the operation of a category assigning WFR ( ]0 [-1 or ]-1 [0)
or a noncategory assigning WFR (]-2 [-1 or ]-1 [-2). All -1 's are created by the
Five.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
93
3 .3 . HEADEDNESS
subphrase having all the sam e category fe a tu re s , but one bar level fe w e r.
and reeducation, the head of the phrase must be the one which determines the
category as a w hole. This turns out to be the righthand m em ber in every case (20
a-c).
(20)
a. N b. V c. N
V
/\ A N -af af V af N
• I t • 1
[[instruct] [ion]] [[re] [instruct]] [[re] [education]]
In (21 a-c) w e see the same words w ith the bar levels assigned according to
the X-bar theory outlined here. Here it is clear that the head is determined by X-bar
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 1 I r I I
[[instruct] -ion]] [[re] [instruct]] [[re] [education]]
Using the examples in (d), W illiams points out that suffixes generally
(22)
X + -ism - - > N
X + -able - - > A
Prefixes, how ever, ordinarily do not determine the category of the word they
Prefixes, he says, are not given category membership because they are not heads.
W illiam s (1 9 8 1 ) says:
In m orphology w e define the head of a m orphologically com plex w o rd to be the right hand
m em ber of th a t w o rd .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
95
and th at prefixes in general should not, but in this theory this is a fact that follows
from the nature of the affixes and not from specific statem ents about prefixes or
suffixes. Instead w e can state th at all category assigning affixes have categories,
while others do not. Noncategory-assigning affixes and fragm ents are never heads.
This then allows for the possibility th at some suffixes do not determine the
category of the word while some prefixes do. Both such possibilities will contradict
W illiam s' claim th at suffixes determine categories while prefixes do not. The prefix
e/7- seems to be one such counterexample, en-, W illiams adm its, is a counter
exam ple and must for him be listed as exceptional to the RHR (see (23) below):
(23)
en rich
In words like enrich or ennoble, then, en- is clearly determining the category
and must be the head. While the example w ith en- is a counterexample for
W illiam s, it is actually further support for the X-bar theory presented here. For this
theory, heads are those categories that share the same features as the syntactic
node but are one bar level lower. This is precisely the X-bar relation that exists for
en- in enrich. Thus, given the notion of headedness provided by X-bar theory it is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
96
partially correct and claim th at the evidence from Spanish diminutive forms
presented in Jaeggli (1 9 8 0 ) indicates that there are some forms in which the RHR
fails to account for headedness. In forms like (17) the diminutive prefix -ita can be
determining the category of the derived word and cannot be the head.
headp of a word is the rightmost element of the word marked for the feature F.
Because the left-hand elem ents of th e form s in (9) ((2 4 ) above] are (by d efault)
the rightm ost elem ents of th e form s marked for category specification, th e y are the
"headcolooofv" (head w ith respect to categ ory), and so the w h o le m u st agree w ith th e m
in categ ory.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
97
the same term s as in phrase level X-bar theory but with none of the counterexamples
of W illiam s (1 9 8 1 ) or Di Sciullo and W illiams (1 9 8 7 ).
a com plex form (phrases in syntax and words in morphology) th a t is the same category
as the entire containing form (phrase or word) but is one bar level lower. Thus in
morphology all category assigning affixes are the heads o f the forms they create.
Fragments and noncategory-assigning affixes can not be heads because their lack of
category m akes it impossible for them to fit the definition of head. Compounds are
redhead or bluebeard but in noun - noun combinations this is less obvious. For
exam ple, the means by w hich w e determ ine the head in words w here the dominating
node and the tw o dominated nodes are of identical categories as in (26) below needs
to be described.
(26)
N '1
book shelf
Since the first noun of such a noun/noun compound is a modifier like the
adjectives in compounds like redhead ideally the X-bar fram ew ork should assign the
same bar level to these modifying nouns as to the adjectives. As described so far the
principle of Succession determ ines th at only one morpheme in every word is demoted
to a -1 . Since the first noun in a form like (26) is modifying the second it is odd to
claim that the category of the compound is being assigned by that noun. Thus it is
plausible to argue th at the second noun of such a compound assigns the category to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
98
the entire compound and thus should be affected by the principle of Succession. Thus
book in bookshelf is not affected by Succession and remains a 0 . This parallels
syntactic phrases such as Jo h n ’s book w here the noun Jo hn's is treated as a modifier
rather than a NP.
(27)
door bell
(28)
head
Clearly, the noun in (27) and (28) share the same head-modifier relation as the
adjective noun compounds and this is reflected in the tree structure by the principle
of Succession. H ow ever, there are cases in English and other languages w here a
noun/noun compound is not in a modifier/modified relation. Mohanan (1 9 8 6 )
dem onstrates th a t the language of Malayalam makes such a distinction betw een tw o
types of noun compounding. Mohanan (1 9 8 6 ) argues th a t it is necessary to
are tw o kinds of compounds th at are made of identical lexical categories, those which
have a specifier/head relation between the tw o elements (29) and those which do not
(3 0 ). (Mohanan 1 9 8 6 :8 3 ).
(30)
N° NT’ N -’ N '1 i
I I I ]
tree horse father mother
'wooden horse' 'parents'
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
99
the morpheme th at determines the category of the word. In (30) however both words are
taken together to form a noun as a kind of coordinate structure. Both the right and left
m ember of the word contribute to making the entire category similar to a coordinate
structure in syntax. Thus Succession determines th at both must be -1 . This leaves a
difference in the tw o structures, but this is good because, as Mohanan points out, there
are phonological rules which apply to one type of compound but not to the other. Thus
the facts of Malayalam subcompounding and cocompounding provide us w ith support for
W hile there is not a large selection of examples of this sort in English, one exam ple
(M . Franklin personal communication) illustrates that there is a need in English to
distinguish such forms. In the compound insect predator it is possible to get tw o
th at is a predator while the other refers to a predator of insects. These are illustrated in
(31) and (3 2 ). (31) represents the predator th a t preys on insects (sparrow) whereas (32)
(31) (32)
adjective. For example, w hen the adjective determining affix -ish attaches to a noun the
tree created is similar to a modifier head combination such as redhead in (28) above.
W hen the affix -ish is attached to an adjective however, the tree created in this case is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
(3 3 ) (34)
A"1 afA-’
i t
monk green ish
There are tw o possible explanations of the trees in (33) and (3 4 ). The first is that
one of the tw o -1 categories is the head and the second is that the addition of -ish to an
adjective creates a coordinate type form similar to a word like bluegreen. In bluegreen the
(3 5 ) (36)
/A 0 A \
/ \ / \
A '1 A '1 A '1 a f'1
I t
blue green green ish
are their ow n heads but in words such as perm it and rem it (the words which were
assigned the ' = ' boundary in SPE) there is no -1 as required by the definition of
headedness. This means these forms must be headless. This would seem to be incorrect
except for the fact th a t these words are not created by W FRs. Since these words are
made up of a fragm ent plus a noncategory-assigning affix, the only w ay they could have
been brought into English is as borrowed forms. Thus the fact that they are headless is
not problematic. Rather w e can claim that these words are their own heads just as
unanalyzable words are.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
101
Thus in this theory there are three kinds of heads: 1) unanalyzable words are their
ow n heads, 2) analyzable words w hich are not created by WFRs (such as perm it) are their
(35)
N°
table
[0 table ]0
(36)
af'2 F'2
I I
per m it
(37) ^ A ° ^
V° a f'1
r i
believe able
In some cases there are words w ith tw o heads. These forms occur in exocentric
compounds and in the case of category assigning affixes th a t are attached to words of the
same category they create. These form s are dual headed because both elem ents of the
word contribute to the category of the entire w ord. These are illustrated in (38) and (39).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
102
/ N° \
N '1 N '1
. I I
insect predator
[0 [-1 insect ]-1 [-1 predator ]-1 ]0
(39)
A
A- r
green ish
3 .4 . OTHER THEORIES
attem p t to account for the facts of the Class I and Class II affixes or of headedness.
schema is only worked out on a very small subset of Latin verbs. T h at is, basing the
assignment of bar I levels on differing kinds of affixes can not conform to the principles of
theory can not conform to the principles of X-bar established in Kornai and Pullum (1 9 9 0 ).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
103
3 .5 . CONCLUSION A ND S U M M A R Y
The theory of X-bar thus is very simply applied at word level. The complications of
this theory caused by the distributions of the affixes are avoided in this theory by the
Im m ediate Dominance Condition presented in Chapter T w o . The X-bar theory presented
here conforms to phrase level X-bar theory in these points; 1) the use of M axim ality and
Succession, 2) the description of headedness, 3) the ability to capture cross categorial
achieved by allowing the morphologically sensitive rules to refer to the labeled brackets of
the phonological string which have been marked w ith category and bar-level information.
This theory of morphological juncture is taken up in the next chapter. The next chapter
presents the theory of morphological juncture called 'X-bar Brackets.'
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER FOUR
4 .1 . INTRODUCTION
bar theory to word level in a straightforward manner once the problem of the Class
I and Class II affixes is resolved. This chapter argues that it is possible to use the
X-bar theory to account for all the facts of the variable application of morphological
Mohanan (1 9 8 6 ). The theory of juncture proposed here is able to account for all
the data of the other theories of juncture w ithout the addition of any new
required for the theory of grammar. T h at is, it is possible to account for all the
tree structure; these bar levels simply have to be included on the labelled brackets
of the phonological string. Since, the X-bar fram ew ork developed here is m erely an
Chapter Three for a full discussion) no new theoretical apparatus is added to the
gram m ar and the boundaries and strata can be eliminated. The X-bar bracket
structure. T hat is, the lexicon consists of just one level and the brackets that are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
105
marked w ith the X-bar levels are not taken as segm ents in the string; the brackets
are m erely a notational device included just like the brackets in syntactic
derivations, simply to represent the category and hierarchical strength of the word
phonetically unrealized segm ents (like ' + ' and ' = ') placed betw een morphemes
by the gram m ar. Thus, in the present thesis, the labeled brackets replace segment
the w ords rather than by positing boundary units in the phonetic string.
This chapter illustrates the ability of the theory of X -bar brackets to account
for the data, but the analyses of other theorists are not discussed (the next chapter
discusses the relative merits of the different theories in some detail). The present
sum m ary.
are listed in the phonological string with a bracket to its left and to its right. This
deviates from the usage of brackets in SPE and later authors. In SPE brackets w ere
placed to the left and right of every category node of the syntactic tree beginning
w ith the low est and continuing to the highest. They did not bracket formatives
such as affixes or stems th at did not receive a lexical category. This difference is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
106
illustrated in (1) and (2). (1) gives an example of SPE style bracketing and (2) gives
an exam ple of the same word bracketed according to Mohanan (1 9 8 6 ) and this
theory.
might reenter the theory by default; th at is, if some morphemes are bracketed while
others are not, it is possible to refer to the differences in structure in term s of the
treated as segm ents. In order for the brackets to represent only categorial
hierarchical information it is necessary for them to indicate the category and level of
every item in the string. In the present proposal the brackets are used simply to
reflect the category of the word being affected and its hierarchical strength (bar
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
107
domains. The bracketing in (4) includes the individual morphemes as w ell. (3)
illustrates the organization of words in terms of the application of WFRs but it does
not give brackets to the affixes. This implies that a speaker knows nothing about
rules according to w h a t they know about the categories of the words they are
using and the hierarchical strengths of these categories. This parallels a speakers
theory, and have no substance.' The theory proposed in this thesis agrees w ith
Aronoff on this, but the state of the theory at the time of A ronoff's comments
in (3) above) does not reflect w h at Che speaker knows about the category of the
WFRs.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
108
The rules selected for discussion in this chapter form a representative subset
to include the appropriate bar level assigned by the theory of X-bars on the labeled
bracket, and second, it is necessary to allow the relevant rules to refer to these
more detailed labeled brackets. (5) repeats example (4) w ith the brackets labeled in
this manner.
(5)
[N° [af'1 re Jaf'1 [N'1 [V° [N° organ] N° [af'1 ize] af'M V0 [af'1 ation laf'1 JN'1 ]N°
discussion of SPE style boundaries demonstrated th at there are three w ays in which
'delineating' are accepted for this theory as well except they are applied to brackets
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
109
...three formally quite distinct sorts of interactions between boundaries and phonological
rules can be distinguished: (a) a rule may apply generally except when blocked by a
strong boundary in the interior of its environment; (b) a rule may apply only when a
particular boundary is present; and (c) a rule may apply only when boundary of a certain
type or stronger appears at the (right or left) end point of the environment of the rule.
boundary, (b) above describes a rule th a t 'requires' a particular boundary, and (c)
theory and in this thesis each rule requires a statem ent th at indicates w hether the
each other as determined by X-bar brackets. The hierarchical order of the brackets
interm ediate and -2 is the w eakest. If rule (6) is 'ranked' for -1, it applies a t -1
present, and a rule 'delineated' for -1 applies at -1 and 0 , but not at -2.
In order to capture the facts of the morpheme junctures in the m ost general
w ith the variable 'X '. This illustrates that the main function of X-bar theory, which
This paper does not actually specify the variable 'X ' in every rule. Instead just the
bracket plus the bar level is listed in the rule (]X ° is replaced by ]0 or IX '1 is replaced
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 1 0
by ]-1 ). H ow ever, if a particular category is required (as in the stress rules), this
Examples (7) and (8) illustrate the actual set of possible boundaries.
(7)
internal1
10 [-1
1-1 [0
] 0 [-2
] -2 [0
]-2 [-1
1-1 [-2
1-2 [-2
]0 [0
1 Since morphem es th at are marked -2 are bound form s th e y never occur in initial or final
position. The fa c t th a t th ey are bound implies th a t th ey are contained by a higher category.
2 This typ e of boundary is rather rare in English. It occurs only w hen a noncategory
assigning affix is attached to a fragm ent as in reserve or permit.
3 No W FRs create a ]0 [0 bracket, how ever, the attachm ent of a tense marker to a verb
m ay actually create a 10 [0 bracket. This is discussed in the section on g deletion I, but for
n ow it is sufficient to point o u t this possibility. Since regular processes of word form ation do
not form a ]0 [0 bracket it is not nentioned in (7).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I l l
The bracketed pairs in (a), (b), and (c) above seem to be of the same
conceivable th at languages might refer to boundaries as groups such as (a), (b), and
(c) how ever, none of the data considered in this paper refers to brackets in that
manner. In general, rules are w ritten for a pair of brackets (one left facing and one
right facing). This pair of brackets describes the juncture of tw o morphemes. Any
reference to more than one bracket type is done in term s of the principles boundary
of M alayalam the use of these principles is more detailed than in English. In a rule
whether or not a word can be affected by a particular rule. For example, a rule that
requires a 1-1 [-2 can apply in a word w ith a ]-1 [-2 bracket but it cannot apply in a
word w ith a 1-2 [-1 bracket. Also, when considering rules th at are ranked or
delineated for particular brackets, both brackets of the juncture must be compared
w ith both brackets of the words being affected. A rule that is delineated for ]-2 (-1
can apply at ]-2 [-1 but it cannot apply to any form where the right facing bracket
]0 [0 brackets are not listed above because these never occur word
internally. The principles of X-bar theory m ake this impossible since it is not
possible to have a word made up of tw o categories th at are not the same as the
category of the entire word. Thus the ]0 [0 brackets are a natural representative of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
112
word terminus. The notion of 'external bracket' is refers to word initial and word
th at the 'X ' variable is being assumed. The brackets in (7) indicate that a rule may
refer to a leftm ost elem ent, rightmost elem ent, or internal bracket (as in SPE and
Stanley 1 9 7 3 ). The facing boundaries indicate that there are seven possible types
bracket strengths w ith the possibility of three possible directions; left, right and
internal and each juncture requires tw o brackets. The left-facing ([) and right-facing
final position. W hether or not rules need to refer to morpheme initial and
morpheme final position is an open question for any language. The distinction
betw een morpheme initial and morpheme final seems to be unimportant in English,
how ever, Stanley (1 9 6 9 ) demonstrates that the Navajo language requires at least
six boundaries and th a t some of these refer to morpheme initial or morpheme final
they occupy in the word and it is not possible to give a one-to-one correspondence
of his boundaries w ith those of this theory. This theory provides a means to refer
in Stanley (1 9 6 9 )).
The underlined sections in (9) indicate that this method of bracketing might
actually allow rules to refer to sequences of three or four brackets. In these strings
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
113
of rules.
(9)
[N° fA° faf'1 un la f'1 fA '1 IN 0 gram m at ]N° [af'1 ical la f'1 1A'1 IA° [af'1 ity ]af_1 ]N°
facing left and right brackets (][). Thus in this theory any rule th at refers to a
accordance w ith the principles of X-bar theory. The possibility that a rule might
refer to more than one bracket on either side of the adjacent ][ pair, needs more
research. In general it is not necessary for a rule to refer to more than the pair of
facing brackets at any one tim e. H ow ever, there are tw o possibilities for citing
junctures in the environment of a rule: 1) a rule can have a pair of brackets at the
end of the preceding environm ent, or 2) a rule can have a pair of brackets at the
This theory assumes th at brackets to the left and right of the tw o brackets
specifically mentioned. Since there are some rules th at do require more than the
means of determining w hen a rule can refer to more than one bracket.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
114
Examples where a rule m ay involve more than one bracket can be motivated
Im m ediate Dominance Condition. For exam ple, in rules like trisyllabic laxing, stress
the distinction betw een Class I and Class II affixes. The only property that
distinguishes these tw o affixes is the fact th at un- needs to use the Im m ediate
Dominance Condition for its attachm ent while in- does not. These rules are
described in the last section of this chapter but w e mention them here to point out
more than one bracket. Particularly, it is possible to predict the application of rules
th at distinguish between Class I and Class II affixes by allowing those affixes that
( 10 )
[ [ un ] n believe ] [ able ] ] ]
determ ine w here it can attach. In the rules th at distinguish betw een Class I and
Class II affixes it is possible to refer to this distinction by allowing the attachm ent
of such an affix to link the tw o brackets it requires for its attach m ent, and then
allow the rules to refer to this distinction. Thus where the juncture of an ordinary
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
115
The inclusion of the category 'F' also needs some discussion. The category
'F ' refers to 'fragm ents' as defined in Chapter T w o . These fragm ents are excluded
from ordinary X-bar labeling because of the interaction of X-bar theory and the
W B H . Given the W B H , fragm ents do not receive their own listing in the lexicon.
As stated in Chapter T w o , words w ith fragments are entered into the lexicon by
truncation. For instance, the word tolerable can be formed from the word tolerate
by the operation of the WFR th at attaches -able followed by the truncation of -ate.
Since the fragm ents do not have a category or a listing they cannot be assigned a
category label. Thus they are given the 'F' label to indicate that a speaker is aw are
their exact nature in gram matical term s cannot be determined. Incidentally, the
m anner of bracket assignment of this theory predicts that truncation removes the
brackets as well as the phonemes of the item truncated, if it also deletes the node.
w hether or not the rule deletes the node. In the case of truncation it is likely that
the affix node is deleted thus the brackets must go as well since the only function
of the brackets is to encode categorial information and the rule is w ritten to operate
would be reasonable to assume that the brackets could be maintained in the string
to represent the fact th at the category node w as not deleted. The facts of English
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
116
and M aiayalam discussed in this thesis do not refer to any such phenomena but the
W FRs and affix subcategorizations are presented first and the ability of the X-bar
morphology also improves the generality and precision of the W FRs. For example,
it is possible to improve the accuracy of the WFRs by adding the X-bar levels to the
(10) affix A — [X [X
becomes ...
(1 0) describes an affix th a t can attach to a tree like (11) but not like (1 2 ).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
One practical exam ple of this is th at it is possible to improve the description
of the distribution of the affix un- The facts of the distribution of un- discussed in
this section are taken from W alinska de Hackbeil (1 9 8 3 ). She points out th at un-,
( 13)
*unyoungish
‘ uncleanish
*unslowish
*unoldish
The interesting point is th at un- can appear on adjectives that have been
(14)
ungirlish
unmannish
unmonkish
unstylish
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
118
possible to account for both facts: th at un- cannot attach to adjectives th at have
been formed from adjectives w ith -ish, and that un- can attach to an adjective
The WFR for un- as presently stated cannot describe these facts. This WFR
is repeated in (1 5 ).
un- has a stipulation using the Im mediate Dominance Condition th at states that un-
which are made up of a fragm ent (a noncategory) and an affix. The tree for
F'2 af"1
r *
toler able
Th at is, the current stipulation on the affix un- is that it cannot attach to an
adjective if the 'A ' node of the affix to be affixed im m ediately dominates a
noncategory. Formulated like this it is not possible to capture the facts illustrated
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
119
affix un- can attach only if the 'A ' node of the adjective to which it attaches
dom inates an 'X 0' it is possible to do so. The trees in (1 7 ) and (1 3 ) illustrate this.
(1 7 )
young ish
(18)
un monk ish
Thus changing the WFR that attaches un- as in (19) below captures these
facts.
The WFR in (19) indicates that un- requires a bar level of [0 [0 in order to be
attached. In the cases described in (13) and (14) the structures have a -1 where
the WFR requires a 0 and thus un- cannot attach. Given the hierarchical order of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12 0
brackets presented in the first section of this chapter this rule is also not applicable
in the presence of the -2 in a word like [° ['2 toler ]'2 I'1 able ]'1 ]°.
account for the facts of English stress. While the facts of stress are actually quite
the stress rules. In this discussion w e merely accept the stress rules as proposed in
SPE and dem onstrate how the theory of juncture proposed here can be used to
account for the junctural facts expressed in these rules. The rules th a t account for
stress in SPE are: 1) the main stress rule, 2) the alternating stress rule, and 3) the
4 .3 .2 .1 . The Main Stress Rule. The main stress rule assigns primary stress
to nouns, verbs and adjectives and is modified by the application of the other tw o
stress rules. The main stress rule places primary stress on unanalyzable words (see
definition of analyzable and unanalyzable in Chapter Tw o). The rule is very complex
of the word and on the syllable structure. SPE (p. 70) gives a prose statem ent of
(i) the pen ultim ate vow el if the last v o w e l in the string under consideration is nontense
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 2 1
(2 1 ) SPE {p.6 9 -7 1 )
The main stress rule needs to mention the category labels of phonetic strings
and it also needs to stipulate th at there are no internal occurrences of '# ' in the
string. The main stress rule is repeated as (2 2 ). In this rule and in rules (25) and
(26) the boundaries and boundary features are indicated by double underlining.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
122
(22)
-tense I avocIv
V — ► [latresa] / £x_ yatreaal CJ0 ( aeons I>
V J -ant _J
r ^ r
-stress
(v
/ —
/ \ (fjJc)At
( < )< r
< l+ c > ‘
n
-tense
-cons _
[■fconsjgj V
Y
' [[ + D ]C °
I £&B>^Pc° r-stress ic ° < 2v°c °>2 1
conditiona: 0
YL 2
X contains no internal
\ -back I+voc) / C_
\ +high r +cor "y
j -cons [-voc] / C \
The category labels such as N , A , and V are treated in exactly the same
prevents the compound blackboard from being accented like the NP black board and
(This is typical of words affixed w ith Class II affixes and gives them the name
'stress neutral').
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
123
(23 )
Ni
/
A0
\
N '1
I I
black board
(24)
soliloquiz ing
contains no internal occurrence of ]-1 [-1 or stronger. This allows the rule to apply
across a fragm ent but blocks it in all other cases. Of course, the double 0 brackets
th a t are linked by the Im m ediate Dominance Condition block the application of this
rule as w ell.
mention the possibility that ' = ' could optionally be present and it needs to mention
the category of the string. This rule is m eant to apply after the main stress rule to
assign stress to the antepenultim ate vowei in words of three or more syllables.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
124
The only boundary mentioned in this rule is the ' = ' boundary and this can be
optionally present. This will allow this rule to apply in the case of the words w ith
possible to capture this fact by replacing the optional ' = ' boundary w ith an optional
]-2 [-2 bracket. The ]-2 [-2 bracket only occurs in a word when a noncategory
assigning affix is attached to a fragm ent. This actually predicts the forms are
headless but this is not problematic since these forms are entered into the language
Three and is an autom atic consequence of the word based hypothesis. This ]-2 [-2
bracket predicts the fact that these forms could be affected as a group but unlike
the unmotivated ' = ' of SPE these brackets are assigned by the same methods that
determine all brackets. Following SPE this thesis accepts that there is WFR rule
adjustm ent rule needs to refer to brackets and category labels. This rule is
repeated as (2 6).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
125
This rule needs to mention terminus of the word on either side of the
environm ent and it needs to specify th at if there are no internal '# # ' boundaries the
the end of the rule. This is m eant to ensure that the rule can distinguish
(27) NP (28) N
A N A
/ N\
' i 1
black board black board
The theory of X-bar brackets can w rite this rule w ith a slight variation. W ith
word terminus and replace the stipulation 'Z = ... # # ... ' with 'Z = ... ]0 [0 ...'
W ith X-bar brackets a word terminus alw ays has a sequence of tw o adjacent Os as,
]0 0[. This should not be confused w ith a pair of Os which face the same direction
as [0 [-1 un ]-1 (OA [QV believ ] ]0 V [af'1 able la f'1 10A. In the case of compounds
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
126
bar brackets. The rules th at are illustrated here are taken from SPE and Lexical
analysis from the other tw o theories is not discussed, but the page references are
given. Rules from SPE and M ohanan (1 9 8 6 ) th a t are not relevant to juncture are
(2 9 ) ]0 [0 W ord terminus
] 0 or [ 0
]-1 or [-1
1-2 or [-2
category assigning and noncategory assigning affixes) as reflected in the X-bar rules
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
127
provides evidence th at this distinction actually predicts facts about the application
of morphologically sensitive rules that other theories cannot. This remains the main
distinction by which this theory might be falsified. If it can be dem onstrated that
the bar levels of changed category versus nonchanged category can not predict the
describes the variations betw een /t / and /s / th at occur before the noun forming affix
(*chocolacy) which are derived by adjective forming -y illustrate that the rule must
(3 0 )
In the case of noun forming -y, the rule of 'Assibilation of /t/' is applicable
but in the case of adjective forming -y it is blocked. In terms of this theory, the rule
of 'assibilation of it /' requires a ]-1 ]-2 and it applies correctly in words taking the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
128
(31)):
(3 1).
\
af'2
president y
[° [° president y I 2 1°
(32) A0
N °/ a f'1
C
chocolate I
[° [° chocolate J^J'1 y I'1 ]°
This rule requires the ]-1 [-2 bracket and is blocked by the ]0 [-1 of words
like chocalatey. The /i/ surrounded by [-2 1-2 brackets indicates the -y affix.
This rule is meant to apply at -1 brackets and is blocked by the '0 ' bracket
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
129
The other rule of /g / deletion deletes /g / before certain boundaries as in forms like
log ~ logger. One deletes /g/s before nasals th at precede certain junctures as in
(3 4 ) through (40) provide some examples w ith the trees th at determine their bar
levels.
(34) A0 (35) A0
A° a f1
Iorj lor^g er
(36) A0
A '1 a f'1
lorjg est
Given these facts it is possible to w rite the rule of Ig l deletion for a ]0 [-1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
130
The rule in (3 7 ) can apply in (34) but the ]-1 bracket of (35) and (36) does
not m atch the required ]0 bracket and hence does not apply. Examples (38)
through (4 0 ) illustrate the application of this rule w ith the verb to long which
counterexam ple. In this exam ple the inflectional affix -ing creates a ]-1 bracket like
the com parative and superlative in (35) and (3 6 ) which predicts th a t /g / deletion
(3 7 ) V° (38)
V° a f’1
I I
lorj log er
1° log ]° [° [° log ]° C1 er ] ]0
(39 ) /N (40 )
V°
I
/ v. af'
I I
A
V '1 af°
I
lorj ig lot) •a
it actually raises an interesting point about the distinction betw een inflectional and
derivational endings. Inflectional affixes like genitive case marking and plural seem
to act just like noncategory assigning affixes and thus should be assigned a -2 as
their basic level. Tense how ever poses a different problem in that it is a category in
its ow n right and is introduced by the syntax. Therefore tense cannot be given a -
2. Tense must be given 0 level as its maximal level in the lexicon because it is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
131
assigned by the syntax. Further, since tense does not play a role in the assignment
of the category to the verb to which it is attached, it cannot be the head and thus
the -1 level would also be inaccurate. Based on the fact that tense is not a head
and also cannot be given a -2 w e assume these affixes m ust remain a 0 . Also
since the process of encliticization rather than a W FR creates this rule it is not
possible to say th a t sing contributes to the creation of the category and thus the
morpheme sing is also exem pt from Succession. This actually w orks to the
(4 1 )
V 1 af°
I
I
sig '3
inflectional endings have on junctures, it is clear this distinction should be m ade and
that it has an a ffe c t on the junctures. Inflectional endings for plural and genitive
m ust also be attached at a level higher than other noncategory assigning affixes as
indicated in (4 2 ).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
132
H ow ever, since they do not play an independent role in the syntactic string
and since they do not assign categories the best analysis for these kinds of
subcategorization th at allows them to attach higher than 0. This predicts that the
/z / ending of the third person present indicative of regular verbs should also be
assigned a 0 , while the /z / of the plural and the genitive should be assigned a -2.
(43) (44)
NP
/ \ VP
I I
N .V N
I / \ /\
V af N af
1 I I I i
he long s song s
The rule of g deletion as w ritten indicates that it applies before the verb
marker /z /, but not before the plural or genitive. However, the /g / is deleted in all of
these cases, thus it seems th a t the /g / is deleted by some general phonetic process
such a sequence.
4 .3 .3 .3 . Rule # 3 : /g / Deletion II. In this rule /g/'s are deleted before nasals
which precede a ]° [0 juncture. The examples in (45) illustrate that the rule of /g/
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
133
inapplicable before the w eaker ]a [-1. Also words that end in inflectional ending -
ing appear as counterexam ples if they are attached by WFRs at the 0 level,
how ever if they are treated as independent lexical items (as is discussed above) the
juncture cannot be a -1 but m ust be a 0 and thus the rule should and does apply as
in rule # 2 above. As per the above the discussion the base is also not affected by
in these form s because they are separate elem ents at the morphological level. A t
the syntactic level the principle of Succession does not apply because syntactic
(45)
signal
signify
(46) (47)
N°
N'1 af'2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
134
(48) (49)
,V ° V°
v
/1
1 \ a f -l
a f1 v A
f> a f o
1 I
1 1 I I
sign ify sign ing
(50 )
/N °
/ \
V° a f’
r I
assign ation
The exam ples above illustrate th a t all the cases of g deletion can be
explained in term s of the ]0 bracket. In (49) the word signing does not allow the
application of g deletion because the -ing is a tense marker and receives a 0 level
g —- / f / / r j / l a [0
works in the same manner as the rule of g deletion II. It applies at a ]“ [0 bracket
and it
accounts for variations such as hymn — hym nal and condemn - condem nation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
135
n — 6 / ___[ + nasal] ]° (0
(54) (55)
A0
A0 a f'1
I I
solemn solemn ity
(56) (57)
V°
\
I V°
I
a f'1
1
damn damn ation
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
136
(58)
V°
A af°
r i
damn ing
[0 [0 damn ]0 [0 ing 10 10
(5 9 )
V°
/ N\
V° af'1
I i
hymn hymn al
(60)
I 1
hymn ing
[0 [0 hymn 10 [0 ing ]0 10
grounds that hym nal is derived from the noun hymn rather than the verb hymn.
The examples in (61) through (65) are accounted for w ith the same rule.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
137
(61)
N°
I
column
[0 column ]0
(62)
N° a f1
1
cofumn
<
ar
(63) (64)
V° N°
V af
I l
condemn condemn ation
(65)
I I
column s
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
138
prevent th at.
(66) (67)
Adv°
A \
A0 a f1
i I
simple simpl ly
(68)
A a f'1
simpl ify
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
139
(69) (70)
V° Adv°
/ \
V° a f1
I I
double doubl ly
(71)
N
/ \
N'1 V1
[
doubl et
(72) (73)
V° No
V' V I
[
burgle burgl ar
(74) (75)
V° A0
/ \
V° a f1
I I
wrinkle wrinkl y
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
140
twinkling (V) w here the I optionally is syllabified in verbs case but not in forms
derived from verbs. This alternation illustrates a distinction between -ing used as a
tense marker or -ing used as a category assigning affix. In the former the affix gets
a 0 as explained above in the discussion of rule # 2 . In the latter since the affix is
being used as a category changing affix it should not receive the 0 bracket of a
tense marker but should be treated the same as other category assigning affixes,
by assigning boundaries in this manner it is possible to predict the facts of the rule
of sylabification of sonorants.
(76) (77)
a f'1
(78)
/ \
af°
tw inkl ing
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
141
[ + son] — [ + syl] / V [0
consist. In the terminology of this proposal these variations occur only w hen the
rule of intervocalic is/ voicing occurs at junctures which begin a fragm ent and are
the rule is w ritten for a ]-2 [-2 the rule will apply correctly.
phonetically identical free form as illustrated in the pair reserve (meaning 'to serve
again') and reserve (meaning 'to holof for someone'Jthe ]0 bracket of the reserve
(80)a
(80)b
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
142
(81) (82)
A
a f'2 F2 af'2
A V '1
re serve re serve
/s / - * ( Voice) / V ]-2 1 -2 V
These examples are accounted for simply be writing the rule of s voicing to
be sensitive to 1-2. This rule is ranked for ]-2 [F and is blocked by all stronger
brackets.
converts /g / to /j/ and /k / to /s / before high front vowels as in words like critic -
criticism and analogous - analogy. W hile this rule seems w ell m otivated in the /k /
representation; th at is, SPE argues th at the underlying forms for pairs like
recognize the excessively abstract nature of some of SPEs analyses are quite
questionable, the rule of velar softening is included here w ith this analysis because
analysis the facts of the Ig l to l]l sub rule are quite unpredictable. W ith the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
143
abstract representation it is possible to state that the /g / becomes l]l before [-low]
(84)
analogous analogy
reciprocity reciprocal
allege allegation
rigid rigor
regicide regal
k - s / _ ]a [-1 V
[£ * J
g—i /_ -1 v
; bck]
(8 6 ) (87)
.0
N°
N'1 a f1
critic
t
critic
r
ism
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
144
(88)
a f1
critic
softening should be applicable to the affix -al as before the affix -ine as in (8 9 ).
H ow ever, it does not apply in m edical and medicate. There seems to be no clear
H ow ever, of all the rules considered here the rule of velar softening is the least well
m otivated. The theory of X-bar brackets is no more nor no less able to arrive at a
(89) (90)
t
medic ine medic al
[0 [-2 medic ]-2 [-1 ine ]-1 ]0 [0 [-2 m edic ]-2 [-1 al ]-1 ]°
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
145
rule because it does not operate on the nasal in words w ith un- as in unpassable or
unclear.
un reliable ir respective
As (93 ) indicates the trees for the words w ith un- and in- are identical and
(93)
un believe able
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
146
(9 4 )
af'1 V° a f'1
in pass able
In order to w rite this rule it is necessary to find a means to allow the rule to
be sensitive to the difference betw een un- and in-. It is necessary for the rule of
nasal assimilation to be sensitive to the distinction betw een Class I and Class II
affixes. This is best described in term s of the property which distinguishes the tw o
affixes; the ability to see limited amounts of internal structure due to the Immediate
Dominance Condition. In order to reflect this in the rule w e simply allow the
brackets together to reflect the fa ct that both are relevant to the affix. The rules
are then w ritten to be either triggered or blocked by this double bracket. This
linking reflects the fac t th at the Im m ediate Dominance Condition involves tw o levels
un believe able
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
147
(96)
toler
The facts of the distribution of un- and in- are discussed at length in Chapter
T w o , briefly restated, un- has a restriction on its attachm ent th at states that it
illustrated by the dotted line in (9 6 ). This information can also be indicated on the
phonetic string by indicating th at un- is sensitive to both the first and the second
bracket of the word. This is indicated by the underlining in (9 6 ). Since the facts of
the distribution of affixes like un- can be indicated in the phonetic string it is also
possible to allow the rules to refer to this distinction simply by allowing the rules to
use the double bracket in the place of other brackets. This distinction then makes
distinction betw een Class I and Class II affixes such as: nasal assimilation, stress
retraction, and trisyllabic laxing. This is the only property that can be found that
characterizes the difference betw een these affixes and also the means by which to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
148
This rule requires the ]-1 [0 bracket and is blocked by the ]-1 [0 [0 of words
derived in un-.
stronger bracket. Thus nasal assimilation can apply in [0 [-1 in ]-1 [0 (0 pass ]0 [-1
ible ]-1 ]0 but not in [0 1-1 un ]-1 [0 [0 believe ]0 [-1 able ]-1 ] (see trees in (95)
and (96)). It does not apply in passable because this word has not been affected
by the Im m ediate Dominance Condition and therefore the brackets do not get
linked.
causes stress to retract one syllable to the right in the presence of a Class I affix as
in (98).
(98)
Since this rule is triggered in words w ith Class I affixes and blocked by the
application of this rule in terms of the affix classes. As w ith nasal assimilation and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
149
trisyllabic laxing this rule can be correctly formulated in terms of the Im m ediate
Dominance Condition and a double bracket. (99) states the rule and it is blocked by
This rule can apply across the bracket of [N ° [A0 divine ]A° [afN'1 ity ]afA'1 ]N°
but not in [N° [ N 1 maid IN '1 (af0-1 en l a f 1 IN '1 [af'2 hood ]af'2 ]N°.
presence and absence of Class I and Class II affixes and is applied in a manner
variations such as profane ~ profanity and serene ~ serenity (SPE:54). This rule
( 100 )
divine divinity
sane sanity
This rule can be w ritten as (1 0 1 ) and like nasal assimilation and stress
retraction this rule is blocked by the doubled bracket that is linked by the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
150
(1 0 1 )
Thus this rule is w ritten for the ]0 bracket and is blocked by the double
th e Class II affixes.
th a t changes the Latin root vert to vers ' ... when followed by the suffixes + ive,
dem onstrates th at this change only occurs w ith this fragm ent (as in insertion) and
Given the theory of X-bar brackets, this one exam ple of an allomorphy rule
can be correctly predicted. If the rule of it /— ►Iz l is w ritten w ith an adjacent ]-2
-ory
M — >/z/ / ___ 1-2 ]0 1-1 -ive
-ion
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
151
(1 0 3 )
/•-
af'2 r\ >2
in vert ion
(1 0 4 )
V° af
af'2 F'2
I I ‘
con vert ion
past theories because the theory of X-bar brackets reduces the am ount of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER FIVE
5.1. INTRODUCTION
This chapter evaluates the merits of the three different theories th at have
been proposed to account for the facts of morphological juncture: the boundary
theory of SPE, the strata theory of Mohanan (1 9 8 6 ) and the theory of X-bar
brackets presented in this thesis. This chapter is necessary to dem onstrate that
these theories are not merely notational variants of one another and that there are
clear reasons to prefer the theory of X-bar brackets over the others. This
comparison is organized into three sections: the first describes the ability of these
theories to account for the data, the second compares them on theoretical grounds,
and the third investigates the different predictions made by these three theories.
Particularly, this third section investigates the ability of the basic mechanisms of
these theories to account for the data. In Mohanan (1 9 8 6 ) the main mechanism is
the division of the lexicon into strata and stratal ordering. In X-bar brackets the
main mechanism is the X-bar theory which is affected by the morphological notions
of category assigning and noncategory assigning. The third section discusses the
examples that are counterexamples for Mohanan (1 9 8 6 ) but are predicted by this
theory.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
153
their ability to account for the data. Number (1) illustrates in a general sense the
(1)
SPE LP X-bars
a. Class I + Strata 1 ]F ]0
or]-1 ]0
b. Class II # Strata 2 ]0 ]0
or ]-1 [-1
(2) illustrates the different w ays these theories refer to junctures in rules,
a. P Q / + ___
b. P —►
Q / [strata one]
c. P - - * Q / ] . , __ 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
154
5 .2 .1 . SPE. This discussion begins w ith the boundary theory of SPE. All
three theories claim to account for all the data. The SPE theory accounts for all of
the data w ith o u t counterexamples; how ever, the theory itself falls to criticisms of
being far too pow erful, ad hoc, and unfalsifiable. This is discussed more completely
The SPE theory; 1) assigns the ' + ' boundary to form atives, 2) it assigns
on either side of every lexical category in the syntactic string, and 3) it assigns '# # '
proposing t h a t ' + ' and be inserted in the lexicon w ith the appropriate affixes.
Siegel (1 9 7 4 ) dem onstrates th a t the ' = ' boundary can be eliminated by reanalyzing
certain aspects of the stress rules (as discussed in Chapter One). In general the
SPE fram ew ork correctly generates the sorts of boundaries th a t are required for the
application of the relevant rules. Any problems or paradoxes th at arise are handled
by the special boundary ' = ' or by readjustment rules (as discussed in Chapter One).
d ata w ith the exception that Mohanan (1 9 8 6 ) does adm it to some serious
lexicon must be divided into a series of ordered strata to account for the order and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
155
discussed in Chapter One). In M ohanan's theory the rules that apply to Class I
affixes apply have strata one as their domain of application, the same strata in
which the Class I affixes are attached. Rules th a t affect Class II affixes are applied
at the level in w hich these affixes are attached, strata tw o . Rules th a t a ffe c t
compounds are applied at strata three, the compound strata. Finally, the rules th at
most of the data but there are tw o types of counterexamples that get in the way;
the mixed strata affixes like 'un-' and derived forms like ungram m aticality and
reorganization. The mixed strata affixes w eaken the stratal uniqueness hypothesis
The subdivision of the lexicon into ordered strata m akes some rather strong
th at this theory m akes is th a t the strata are necessarily sequential. W ithout this
affix. Stratal ordering requires th at all Class I affixation precede all Class II
counterexam ples to the theory of ordered strata in the lexicon begin in Aronoff
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
156
these examples the Class II affixes re- and un- must have been attached before the
Class I affixes -ation and -ity because the Class I affixes are not subcategorized to
attach to the earlier forms. T hat is, re- attaches only to verbs and thus it must
have been attached to organize not organization. Similarly, the un- affix of
ungram m aticality cannot attach to gram m aticality because un- attaches only to
Aronoff and Sridhar (1 9 8 3 ) point out that all potential words of the form
explain aw ay a small subset of the counterexamples but these forms are actually
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
157
5 .2 .3 . X-bar Brackets. The theory of X-bar brackets accounts for the data
criticisms of those theories. Particularly, the theory of X-bar brackets avoids the
mechanisms used in this theory (X-bar theory, labeled brackets, the locality
lexicon. It also predicts the correct attachm ent of the affixes in the
manner in which X-bar brackets predict the correct application of rules that refer to
juncture. Chapter Six gives further evidence of the ability of X-bar brackets to
account for the relevant data. To date there are no clear counterexamples to the
5 .3 . THEORETICAL CONSIDERATIONS
theories on theoretic grounds. As stated earlier, SPE theory requires the theoretical
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
158
(5)
(6)
1. A subdivided lexicon.
2. Rules need to be marked for their domain (the strata in which they
apply).
X-bar brackets uses the mechanisms in (7) (all of which are independently
(7)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
159
The notions of the SPE theory in (5) are not independently m otivated and are
ad hoc. The nature of the boundaries them selves are not discussed and the rules
th at insert them are unm otivated. The readjustm ent rules th at form a crucial part of
the theory are extrem ely pow erful and m ake the theory unfalsifiable. The
theory th at did not need to posit such extra phonemic elem ents would be preferred.
Thus SPE can only account for all the data because of its excessive power.
As for Mohanan (1 9 8 6 ) the division of the lexicon into ordered strata is the
m ost controversial aspect of this system . M any linguists have found this to be
describing both the facts of affix distributions and the facts of morphological
juncture. The motivation for the strata comes from the distributions of Class I and
Class II affixes, and from the distributions of rules that refer to cocompounds and
subcompounds in M alayalam . As stated earlier the Class I affixes are those which
attach to both fragm ents and words and the Class II affixes are those th a t attach
Recall th a t the original m o tiv a tio n for stra tu m ordering in Lexical Phonology w a s the
be a tta c h e d to strings containing class 1 d erivation, but not vice versa) correlated w ith
p ro ce s s e s ....
The main objections to this theory are these; 1) the separation of the rules
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
160
the loop th at allows rules to violate the principle of stratal uniqueness causes the
theory to be too pow erful, and unfalsifiable, 3) the W ord Based Hypothesis (Aronoff
1 9 7 6 ) cannot be maintained, 4) the facts of the distribution and order of the Class I
and Class II affixes cannot be predicted and m ust be stipulated, 5) the comparative
and superlative affixes -est and -er w hich are generally classified as inflectional
endings m ust be listed a t stratum one w ith the Class I derivational affixes, because
affixes.
affixes and the application of morphophonemic rules; how ever, its theoretical cost
allow s for the replacem ent of word boundaries, but the problems w ith word
boundaries are replaced by the problems w ith strata (outlined here). Hargus (1 9 8 9 )
points out th at Mohanan does not justify these levels; he merely accepts this notion
word boundaries such as is found in SPE, but the division of the lexicon is still ad
hoc and counterintuitive. A theory th a t could account for the same facts w ithout
# 2 . The loop. W hile the notion of stratal ordering requires that all rules that
apply at later strata should not be able to affect forms that are created at lower
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
161
levels. T hat is, a rule that applies at stratum 2, for instance, should not be able to
apply to forms created at a later level. This is does not actually hold in all cases
and so Mohanan introduced the notion of a 'loop' to allow certain rules to loop back
to an earlier stratum in order to apply to earlier forms. For exam ple, compounds are
created at stratum 3 in English but they can be affixed w ith Class II affixes which
Class II affixes. Thus the word frogman is created at strata 3 and then it loops
back to strata 2 in order to receive the affix ex- to form exfrogman. Mohanan
The loop that Mohanan uses to allow compounds to be affixed w ith Class II
affixes violates the concept of stratal uniqueness. This raises the question of
w hether or not the theory is falsifiable. The loop is also used in the Mohanan
and for stress assignment (Mohanan 1 9 8 6 :1 2 0 -1 2 4 ). If the device of the loop can
be brought into play any tim e counterexamples are found then the theory is no
longer falsifiable. So far the loop has only been used in cases referring to
connecting levels, but Mohanan proposes no theoretical constraints on the loop and
he does not address the question of falsifiability. Even if there are w ays by which
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
162
it is possible to falsify the concept of stratal divisions, the devise of the loop is still
excessively powerful.
such as toler- of tolerable can not be listed in the lexicon w ithout an affix to make
them into a full word. A theory that subdivides the lexicon in order to describe the
facts of the Class I and Class II affixes must give all bound stems individual listings
in the lexicon in order to have them available at the first level for Class I affixation.
W ithout the W B H , fragm ents like toler- can be listed in the lexicon as lexical
morphology must maintain th a t stems are given individual listings in order to predict
the order of attachm ent of Class I affixes before Class II affixes. If the bound
stem s are not given individual lexical entries it is not possible to m otivate the
required first level for Class I affixation. Theories that do not adhere to the WBH
such as Mohanan (1 9 8 6 ) need to allow fragm ents to have lexical entries in order to
g et the correct distribution of the Class ! and Class II affixes. H ow ever, this has
the unfortunate consequence th at these theories must also mark all words in the
lexicon in some manner to make sure they take affixes at the appropriate level. If
all words are not individually marked as to which class of affixes they take then
those that take Class II affixes would erroneously be available for Class I affixation
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
163
as th ey pass through the different layers of the lexicon. In a theory that does not
divide the lexicon both the Class I and the Class II affixes are in com petition and
requires the listing of bound stem s is a w eaker theory and is counterintuitive. Most
th e distribution of the affixes. Theories w hich accept the WBH do not allow the
fragm ents to have their own listing, thus for them the level ordering hypothesis
cannot hold. None of the researchers w ho assume th at fragm ents should be given
lexical entries provide any arguments for it (See Scalise (1 9 8 6 ) for a discussion of
the more restrictive theory and it is in keeping w ith the subjective judgements of
speakers.
#4. The fa c t th a t English affixes separate into the Class I and Class II
stipulated. Chapter T w o demonstrated that the facts of the distribution and order
of attach m en t of the Class I and Class II affixes can be predicted using the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
164
the fa c ts of the application of the rule of /g/-deletion in words w ith the com parative
affix -e r and the superlative affix -est indicate th a t listing these affixes at strata four
w ith other inflectional affixes gives the wrong results. In English /g/s are ordinarily
deleted w hen a strong juncture is present ('# ' in SPE or strata 2 in Lexical
Phonology). The rule is not applicable in the presence of w eaker boundary such as
A N
/ \
A af af
The rule of g-deletion is assigned to strata tw o . If -er and -est are attached
a t strata four the rule is no longer availableso can not apply. This predicts that the
/g / should appear in the surface structure which is clearly false. Since -er and -est
are inflectional endings they should be attached at stratum 4 w ith all other
form s using X-bar brackets th at does not require such a reanalysis of the affixes.
com parative affixes w ith the Class I affixes indicates that the theory is not actually
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
165
able to reflect the facts of English affixation. The ordinary analysis of -er and -est
m ust be dropped in favor of a proposal designed simply to get them to fit into the
level of phonology, Ford and Singh (1 9 8 3 ) point out that many theorists such as
empirical support for the notion of the morphophonemic level. SPE says of such a
level.
another linguistically significant level of representation, interm ediate in 'a b s trac tn e s s '
b e tw e e n the lexical (phonological) and phonetic and m eeting the conditions placed on
existence of such a level has not been d em on strated and th a t there are strong reasons
m otivate the separation of the lexicon into separate strata. W ithout the subdivision
of the lexicon it is not clear w h y there should be a separation of rules into lexical
notations in that: 1) it lessens the amount of theoretic apparatus required for the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
166
gram m ar, 2) all the theoretical mechanisms it uses are independently m otivated, 3)
it can predict the counterexamples of other theories, 6) it does not require segm ent
like elem ents to be inserted into the phonological string, and 7) it does not require
particular, the X-bar theory is a w ell-know n component of syntax. Besides the fact
arrangem ents of lexical categories, it is logical that this theory should also account
brackets uses the labeled brackets of the phonological string as the markers of the
morphological information in the rules. The labeled brackets are required for the
syntax, morphology, and phonology and some version of a locality condition like the
in morphology and phonology are the result of the fact that word formation acts in
blocks in different levels of the lexicon. For the theory of X-bar brackets these
hierarchical differences are the result of the operation of the X-bar principles.
In m any cases, all these theories make exactly the same predictions.
H ow ever, there are cases w here the evidence illustrates that the predictions of one
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
167
of the theories are preferable to those of the others. T h at is, the fact th at the
theory proposed here refers to structural relationships within the word rather than
different lexical strata constitutes a different empirical claim. Thus, the theory of
attach m en t, not the organization of the lexicon. Further, the morphological function
versus noncategory assigning. This prediction is borne out in those cases w h ere X-
bar brackets account for the counterexam ples to stratal ordering such as
junctures in term s of the X-bar levels. The notions of category assigning and
noncategory assigning are the underlying notions that determine the bar levels and
the distinctions betw een inflectional and derivational affixation. The fact that this
theory predicts the correct application of the rules that generate the forms th a t are
account for the facts of juncture than the division of the lexicon.
will depend on the strata (as determ ined by their domain of application), the theory
of X-bar brackets predicts that these rules will apply according to w hether or not a
category has been changed as reflected in the bar levels. This is discussed
thoroughly in Chapter Five. The Chapter Five discussion dem onstrates that the
diferent predictions made by these theories are illustrated w ith exmples that are
counterevidence to the theory of Lexical Phonology, but support the theory of X-bar
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
168
the facts of the application of these rules in all cases. The notions of class
not been captured by the notions of class membership and the distinction between
derivational and inflectional afixation. These points are illustrated w ith th e relevant
rules in Chapters Four and Six. For instance. Chapter Four dem onstrates that the
This particular formulation of X-bar theory provides solutions for all the
extrinsic ordering statem ents for the order of affixes. Before demonstrating the
usefulness of this theory for the problems of morphological structure, how ever, the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER SIX
X-BAR BRACKETS A ND M A LA Y A LA M
6 .1 . MALAYALAM PHONOLOGY
This chapter illustrates the ability of the theory of X-bar brackets to account
for the junctural phenomena of a language other than English. The language chosen
and discusses the relative merits of the different theories, this chapter merely
illustrates the application of the theory of X-bar brackets w ith the facts of
H ow ever, in M ohanan's analysis Malayalam does require four strata: one each for
that the tw o different kinds of compounding that are required to explain the facts of
M alayalam support the overall theory cf Lexical Phonology. He claims that the
facts cannot be explained w ithout stratal divisions of the lexicon, because SPE
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
170
dem onstrates th at the theory of X-bar brackets, however, can account for this data
w ithout strata and thus this argument for stratal ordering can be dismissed.
of the lexical and postlexical levels, second the organization of the four strata of
(1 9 8 6 ). The rules and data presented in this chapter do not question Mohanan's
basic analysis and the reader is referred to this for a full discussion. The phonemes
typew riter keyboard except for the schwa (for a mid lax vow el) and /n / (for a velar
nasal) which are represented as 'E' and 'N ' respectively. The nasals In i and Ini
alveolar nasal.
phonological rules into tw o groups: lexical and pcs tlexical. Lexica! rules are all
those rules th a t are sensitive to morphological structure. These rules are applied in
the lexicon and are assigned to particular morphological strata. The second group
of rules is the postlexical level. This level involves the application of all the rules
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
171
boundaries th at do or do not intervene. These rules apply a fte r the application of all
the lexical rules. W hile this separation of rules into lexical and postlexical accounts
for the facts of the application of these rules it requires th a t there be still another
level of rule application after how ever many strata are required and it also requires a
level of basic phonemic representation that is part of the lexical entry, then there
are interm ediate levels of representation at each strata and finally there is the
postlexical level. Thus in the description of any language w ith Lexical Phonology
there are at least three levels of representation (assuming a minimum of one strata
for any language); the phonemic level, the stratal level and the phonetic level.
phonological rules, phonemic and phonetic. Phonemic rules are those th at are
independent of the morphology and phonetic rules apply w hether or not a juncture
processes. Mohanan does not have the option of explaining this phenomena in
term s of rule type because these phonetic rules must apply after all of the rules that
are ordered in the lexicon. This forces Mohanan to include yet another level of
representation in phonology.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
172
and the postlexical level for his description of Malayalam phonology. The strata are
(1)
stratum 1: derivation
stratum 4: inflection
affixation such as the Class I and Class II affixes, so one level for derivational
affixation is sufficient. M alayalam does how ever require tw o strata for the
compounds: one for cocompounds and one for subcompounds. This difference in
English insect predator (for an animal that preys on insects) and insect predator (for
an insect which preys on something) which motivated the use of ]0 bar level for
following section.
provide an interesting test of the ability of X-bar brackets to account for the data.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
173
(Mohanan 1 9 8 6 :8 3 ):
Subcom pounds have th e com m on head and m odifier structure, as in the English
com pound colourblind, w h ile cocom pounds (dw anda com pounds in Sanskrit) have the
structure head head + head .... as in th e English com pound d e a f-m u te .'
If the words compounded are used in a coordinate sense, the words are
this chapter it is demonstrated that the theory of X-bar brackets can predict these
different strata as illustrated in (2) and (3). The X-theory formulation presented in
as illustrated w ith the examples in (2) and (3). The cocompounds have an internal
]-1 [-1 juncture because both words of the compound contirbute to the category of
the entire word. In the subcompounds only the righthand member is demoted to -1
(2) Subcompounds
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
174
c. /N °
X '
N N '1
I I
tree horse
(3) Cocompounds
b. mahaa ausadham
• « mahausadham
•
c. N
N°
\
N'1
1 I
mahaa ausadham
*
is capable of making this distinction by giving each kind of compounding its own
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
175
strata. These rules and the other rules of Malayalam phonology pertinent to a
Malayalam .
6 .3 . X-BAR BRACKETS IN M A L A Y A L A M
The X-bar theory assigns bar levels to the different sorts of derived words as
illustrated in (4) - (8). Throughout this illustration of the use X-bar brackets for
/ \ / \
X° X '1 X '1 X '1
I I I I
[0 [0 !0 [-1 1-1 10 [0 [-1 ]-1 [-1 1-1 !0
(6) Inflection1
/ XKX°
°
f'
X-1
/ \af o *0
I '
[0 [-1 ]-1 [-2 ]-2 10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
176
X° X°
/\
X° af*'
/\
X '2 a f'1
to throughout this discussion as w ell, they are repeated in (9) for convenience.
(9) The hierarchical order of the brackets arranges them from 'strongest' to
'w e a k e s t' w here the 0 is strongest, -1 is interm ediate and -2 is the w eakest.
A . A rule 'ranked' for ]-1 can apply ]-1 and at 1-2 but is blocked by ]0.
B. A rule 'requiring' ]-1 can apply only when this bracket is present.
C. A rule 'delineated' for ]-1 the rule can apply at ]-1 and 10 but not at 1-2.
and delineating become a little more complicated in these instances. For exam ple a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
177
rule th a t is delineated for ]-2 [-1 will apply a t ]-2 [-1, ]-1 [-1, and ]0 [-1. This rule is
only blocked in those cases w here the right hand bracket is lower than -1 as in ]X [-
delineating and requiring a double bracket. If delineated for a particular juncture the
rule m ust consider the left hand bracket as the low est possible left bracket and the
right hand bracket as the low est possible right hand bracket. A clear case of this
occurs in M alayalam w hen a rule applies at strata 1, 2, and 3. That is a rule that
To state this in X-bar bracket terms, it is necessary to delineate the rule for a 1-2 [-1
bracket. It will not apply in inflections because the left hand bracket of an
strata 2 ]0 [-1
3 T h e X - b a r b r a c k e t v e rs io n o f s tr a ta 1 h e re is f o r n o n c a te g o r y a s s ig n in g p r e fix e s . These
f o r m t h e m a jo r ity o f t h e e x a m p le s in t h e M a la y a la m d a t a . A c a te g o r y a s s ig n in g a ff ix w o u ld
c r e a t e a ]-1 [0 b r a c k e t f o r p re fix e s a n d ] 0 - 1 [ f o r a u ffix e s . T h e t h e o r y o f X - b a r b ra c k e ts
p r e d ic ts t h a t th e s e c a s e s w o u ld n o t p a ra lle l t h e n o n c a te g o y r a s s ig n in g a ff ix e s . T h is is an
e n tir e ly e m p iric a l m a t t e r b u t t h e d a ta in M o h a n a n ( 1 9 8 6 ) d o e s n o t p ro v id e e x a m p le s t h a t w ill
d e m o n s tr a te th is .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
178
creates a ]-2 [-1 (in prefixation) ]-1 [-2 (in suffixation). Since category assigning
affixes create a 10 [-1 internal bracket, this predicts that category assigning affixes
brackets or w ith derivations (as predicted by Lexical Phonology). All the examples
if there are category assigning affixes in Malayalam the theory of X-bar brackets
ranking, delineating and blocking are less than obvious in some cases. Particularly,
must rank, delimit or require both brackets of a juncture. For exam ple, of a rule
requires a 1-1 [-1 bracket no other bracket will trigger the application of th a t rule. If
a rule is delimited for a 1-1 [-1 bracket it will apply at 1-1 [-1, 1-1 (0, 1-1, 1-1 [0 or ]0
[0. If a rule is ranked for a 1-1 [-1 it will apply at 1-1 [-1, ]-2 [-1 , 1-1 [-2 or 1-2 [-2.
This is further complicated when a rule cites tw o different brackets in its structural
description. For exam ple, if a rule is delimited for a 1-2 [-1 it will be delimited for -2
and higher on the left side of the bracket and -1 on the right side. Thus it will apply
in 1-2 [-1 but it will not apply in a 1-1 [-2. This is because the rule is delimited for [-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
179
1 on the right side and the 1-2 is too w eak for this designation. This w as not too
accomplished by distinguishing between a bracket 1-2 [-1 bracket and ]-1 [-2
whether or not the affix nir- is a category assigning affix or not since tw o of the
translations look like a category change while tw o others do not. These are
(1 1 )
a. gunam nirggurjam
b. aasa Diraasa
'hope 'disappointment'
c. bhayam nirbbhayam
'fear' fearless
d. laija nirllajiam
'shame' 'shameless'
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
18 0
e. / N°
x \
a f2 N'1
I
nir aasa
(a) and (b) above indicate th a t nir- attaches to nouns form ed from nouns,
w hile (c) and (d) indicate th a t a word derived from nir- is an adjective. A check
w ith a native speaker of M alayalam confirms th a t the item s derived from nir- are
nouns. The discussion w ith this inform ant asked originally if a word derived by nir-
( 12 )
W e therefore argue th at all the examples w ith nir- are nonclass changing
necessary to find a form th a t starts as one category and the w ord formed by the
in M ohanan (1 9 8 6 ) derive nouns from nouns and thus it must be assumed that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
181
these affixes are not category assigning but are like English -hood as in parenthood
(13)
sa- w ith
an- negative
cem - red
sam- together
anu- together
prati- against
upa- sub-
6 .4 . THE RULES OF M A LA YA LA M
postlexical level. These are 'the voicing of stops' and 'the lenition of stops'. As
stated above, in the theory of X-bar brackets it is possible to account for such rules
by making a stipulation that some rules apply irrespective of the boundaries that
m ight be in the string. These rules are also not presented here because the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
182
(glide form ation and schw a insertion). These rules are not discussed because they
Som e of the rules of M alayalam also need to be sensitive to the origin of the
N ext ha discusses five rules for nasals. These are; 1). 'the assimilation of
nasal to place of following stop', 2). 'the assimilation of post nasal stop to manner
of articulation', 3). 'spreading of nasality', 4 ). 'the n - - > n rule, and 5). 'the ng ~ >
postlexical and are therefore not included in the discussion. The last three rules are
discussed in order.
stratum 1 rule this rule can not apply in either kind of compound nor in inflection4.
4 T h is ru le a n d s e v e ra l o th e rs t h a t f o llo w a re w r it t e n in t h e m a n n e r o f a u to s e g m e n ta l
ru le s . T h is s h o u ld p o s e n o p ro b le m s f o r th e d is c u s s io n h e re a n d it s h o u ld b e in te rp re te d as is
c u s to m a r y f o r t h a t f r a m e w o r k .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
183
I+ r son 1
r -son "I (domain: stratum 1)
-cont
] + voice]
but M ohanan provides no exam ples of this. M ohanan does not provide an example
of a subcompound for this rule. Examples (5) and (6) above provide representative
exam ples of such form s. The rule of nasal assimilation applies to m ake the /m / of
sam giitam
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
184
r
r
miin buddhi
This rule is ranked for the 1-2 [-1 juncture of noncategory assigning
spreading 'requires' this particular juncture marker and thus can not apply in other
forms.
(Mohanan 1 9 8 6 :7 1 -3 ). These rules are also assigned to strata 1 and they account
for the variation betw een /n / - In i in words like naayakan 'hero' and upanaayakan
'secondary hero' or the variation betw een ny and n in words of Dravidian origin.
Mohanan claims that the Dravidian subset of vocabulary has the ny as underlying
segm ents but rule # 3 changes them to n which accounts for the fact that Dravidian
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
185
vocabulary has no /ny/s. Thus the word for 'Sunday' is nyaayar underlyingly but
after the application of rule #3 it is naayar. Giving this rule a domain of strata 1
(1 9 8 6 ) provides no examples to illustrate this but the solution described for rule #1
is the same. Thus like rule #1 these rules are ranked for a J-2 [-1 bracket and are
derivational and inflectional affixation are superficially the same in th at both contain
a -1 and a -2 bracket, how ever these brackets are in opposite order in the different
kinds of examples provided by Mohanan. Inflectional affixes are alw ays suffixes
and derivational affixes are always prefixes. Thus in the case of derivations (with
prefixes) the internal bracket that is created is a 1-2 [-1 while the internal bracket
com pletely against the predictions of Mohanan (1 9 8 6 ). Rules that are w ritten for a
applies across words to insert a schw a after word final consonants other than /m /
below in (1 8) and (19 ). (18) gives examples of cocompounds and (19) gives
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
186
Mohanan does not explicitly state th a t the rule applies in citation form s as
w ell but Mohanan includes the inserted schwa (the 'E') in the underived word as
w ell as in the derived word as in (18)a waa/E. W hile this is a bit confusing w e
(18) Cocompounds.
b. waa[alak'k'aparkakalka
c.
b. aanEpennEwyatyaasam
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
aanE pennE w yatyaasam
(19) Subcompounds.
b. wirakE kolli
•*
wirakEkolli
• •
/ N° \
N° N '1
I 1
kaatE puucca
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
188
T he theory of X-bar brackets can account for these facts by delineating the
rule for a ]-1 [-2 bracket as in (2 1 ). By delineating this rule for the ]-1 [-2 bracket,
this rule is not applicable a t the ]-2 [-1 juncture of derivational prefixation but it can
apply at the ]-1 [-2 juncture of inflectional suffixation and the junctures of
cocompounding and subcompounding. Since this rule is delineated for ]-1 [-2 it can
apply at any juncture th a t is the same or higher than both. Thus the ]-2 [-1 of
derivational prefixation is low er than the bracket cited in the rule. Examples (18)
and (19) illustrate this rules application in sbucompounds and cocompounds, (22)
(2 1 ) Stem -final schwa insertion. This rule is delineated for its bracket.
This rule cannot apply in derivations because the 1-2 [-1 bracket cf a
the rule.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
189
(22)
/N \
N'2 N'1
[ I
nir bhayam
support the theory of Lexical Phonology. In subcompounds when the first word of
a subcompound ends in a nongeminate sonorant the rule does not apply. Mohanan
claims th a t this is because the rule of onset fusion fuses stem-final sonorants w ith
the onset of the following syllable thus blocking the application of schw a insertion
as in (2 3 ).
(23)
a. tala wallE
• •
w aalttala *w aalE tta
N°
A N '1
1 I
wallE
•»
tala
.M ohanan makes the prediction that w ith respect to onset fusion, derivational
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
190
affixation will pattern like cocompounding and inflectional affixation will pattern like
subcompounding based on the order of the strata. This is borne out in the data as
(24) Inflections
a. awalE awalEkkE
♦
b. awarE awarEkkE
awalE kE
(25) Derivations
a. akal- akalcca
to recede distance
b. pakar- pakarca
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
191
pakar ca
The theory of X-bar brackets can account for this data as w ell, if the rule of
schw a insertion is w ritten for a ]-1 [-2 juncture the rule will apply correctly in the
the juncture formed in derivation. By writing the rule of onset fusion to require the
predict that onset fusion bleeds schwa insertion in subcompounding the theory of
X-bar brackets orders the rule of onset fusion before the rule of schw a insertion.
Schw a insertion does not occur in derivations because derivations do not have the
right bracket. The fact th at schwa insertion does not occur in subcompounds is
(26)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
192
c. N°
N° N’1
kallE pratima
(2 7)
a. ara kallara
c.
N° N'1
1 I
amma pennE
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
193
C (domain: strata 2)
[ + son!
Since the rule is w ritten for strata 2 it must be blocked from applying in
brackets if the rule is w ritten for a ]0 [-1 bracket it will apply correctly.
bracket.
C — $ / ]0 1-1 C . C
[ + son]
(30)
—* E / O N O 0
i i 1
X (X ) X (X) X X (XI
[ + son] [ + son]
gemination '... applies to stops that are preceded by a single nonnasal sonorant
consonant, but it does not apply if the nonnasal sonorant is double...'(M ohanan
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
194
(31)
a. —
tii katta tiikkatta
i« —
/ N\
N° N°
I I
swarnnam pazam
•* ?
[0 [0 swarnnam ]0 [-1 pazam ]-1 ]0
C -» C C / X ] [
V
v ' I — If
+so n j Is -son “1
[ -nas > I+ D r a v j
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
195
(33 ) Subcompounds
N° N'1
I 1
petti pattaayaNNalE
(3 4 ) Cocompounds.
a. pettipattaayaNNalE
petti pattayaNNalE
For X-bar brackets it is possible to describe these facts by writing the rule to
require a 10 [-1 bracket. No other word formation process creates that particular
juncture.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
196
bracket.
10 [-1
+sortj r-s o n ‘j
fnas J [fD ravJ
The rule of stem -final gemination applies in exactly the same domain as stem
initial gem ination and can be accounted for w ith X-bar brackets exactly as for stem-
describes '... an alternation betw een geminate and nongeminate consonants that
(36)
a. gunam nirgunnam
b. bhayam rnrbbhayam
c. lajja nirllajjam
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
197
\ I
.nir bhayam
Mohanan w rites the rule as in (37) and assigns these to domain strata 1.
C — C C C V
i i i
i i i
[-cont] [-cont] r+ sonA
L -nas)
X-bar brackets w rites these rules for ]-2 (-1 bracket and these will apply
correctly. As stated before this assumes that the derivational affixation being
C -* C C C ]-2 [-1 V
i i i
i i i
____
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
198
nasals in stemfinal position preceding a noun. Some examples of this are given in
(3 9 ).
(39)
c.
malsyam canta
N O N
[ + nasal] —* ] N[
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
199
In the theory of X-bar brackets the rule needs to be delineated for ]-1 [-1
this rule applies in subcompounds and cocompounds but it does not apply in
Since the inflections discussed in the rule of nasal deletion are of the case
marking and plural sort these are assigned -2 and thus the rule of nasal deletion can
be ranked for ]-1 [-1 and the rule will apply in only subcompounding and
not serve as an elem ent in the syntax like the tense markers of English. This
predicts that the rule of nasal deletion will actually apply before inflections that are
more tense-like. This is an entirely empirical m atter but the data in Mohanan
NON
[ + n a s a l]—* I I ’ / ____]-1 [-1
stem as in (42).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
200
N —* N / ] [ (dom ain:2,3)
X
J /I X X [ + Sanskrit]
I V
This rule, like nasal deletion, applies in subcompounds and cocompounds but
not in derivational or inflectional affixation. In this rule the sort of inflection that is
discussed is of the same sort as in nasal deletion and thus it is possible to w rite the
sam e sort of bracketing. This rule is delineated for ]-1 -(-1 and will apply only in
(43) Subcompounds.
c. / N -X
N° N '1
1 I
rati deew i
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
201
(44) Cocompounds.
c. /N ° .
N-1 V1
I 1
bhaatya bharttaawE
N —* N / 1-1 [-1
'X X X _ [ + Sanskrit!
(6) and (8) above) is too w eak for the application of this rule.
describes the process which merges adjacent vowel nuclei into a single one in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
202
[ + Sanskrit] w ords.' This rule applies at strata 1-3: derivation, cocompounding, and
(46) - (4 8 ).
(4 6 ) Subcompounds.
mahaa indran
(4 7) Cocompounds.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
sukham asukham
(4 8 ) Inflections
a. taaTa taarayaal
b. tala talayil
c. N°
N N'2
I \
tala yil
(49) Mohanan (1 9 8 6 :1 0 1 )
a. if tw o nuclei in a sequence are separated by an em pty onset, d elete the first nucleus
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
204
N on th e right.
In X -bar bracket terminology the statem ent 'domain: strata 1 -3 ' is replaced
w ith the s ta tem en t 'this rule is delineated for a 1-2 [-1 bracket'. It will thus apply in
derivational affixation (]-2 [-1), in subcompounding (]0 [-1), and cocompounding (]-1
[-1) since all of these brackets are stronger than the others. The ]-1 [-2 bracket of
inlectional affixation (strata 4) is too w eak (on the right hand bracket) for the rule to
apply.
6 .5 . S U M M A R Y A N D CO NCLUSIO NS
This dem onstration illustrates th at the theory of X-bar brackets is not a one
language theory and predicts th a t X-bar brackets will be useful in predicting the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
REFERENCES
Aronoff, M ark. 1976. W ord form ation in generative gram m ar. Cambridge, M A: M IT
Press.
, and Morris Halle. 1968. The sound pattern o f English. N ew York, NY: Harper and
Row.
Di Sciullo, Anna Maria and Edwin W illiams. 1 9 8 On the definition o f the word.
Cambridge, M A: The M IT Press.
Dolby, J.L. and Resnikoff, H.L. 1 9 6 The English w ord speculum. Volume III, The reverse
w ord list. The Hague: Mouton.
Ford, A. and R. Singh. 1 9 8 3 . 'On the status of morphophonology.' In Papers from the
parasession on the interplay o f phonology, morphology, an d syntax, ed. by John F.
Richardson, Mitchell Marks, A m y Chukerman. Chicago, IL: Chicago Linguistic
Society.
, and K.P. Mohanan. 1985. 'Segm ental phonology of modern English'. Linguistic
Inquiry 5 .5 7 -1 1 6.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
206
Jensen, John T . 1 9 8 1 . 'X -bar Morphology' In, Proceedings o f the eleventh annual
m eeting o f the N e w England Linguistic Society.
Kornai, Andras, and Geoffrey Pullum. 1990. 'The X-bar theory of phrase structure.'
Language 6 6 :2 4 -5 0 .
Pyle, C. 1 9 7 2 . 'On eliminating B M 's'. Papers from the 8th regional m eeting o f the
Chicago Linguistic Society, ed. by Peranteau P., J. Levi,G.Phares. 5 1 6 -5 3 2 .
Chicago, IL: Chicago Linguistic Society.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
207
Publications.
Stockw ell, R., D .B ow en, and I.Silva-Fuenzelida. 1 9 5 6 . 'Spanish juncture and intonation'.
Language 3 2 :6 4 1 -6 5 reprinted in Joos 1 9 5 7
W illiam s, Edwin. 1 9 8 1 . 'On the notions "lexically related" and "head of a w ord”.
Linguistic Inquiry 2 .2 4 5 -7 4 .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.