Académique Documents
Professionnel Documents
Culture Documents
VOLUME 29
Managing Editors
Editorial Board
The titles published in this series are listed at the end of this volume.
ABDELKADER FASSI FERRI
Department of Arabic Linguistics and Literature,
Mohamed V University, Rabat, Morocco
ISBN 978-90-481-4228-6
PREFACE XI
TRANSCRIPTION Xlll
CONCLUSION 281
BIBLIOGRAPHY 284
~ r (glottal plosive)
,_, b
.::..,
,_, (unvoiced interdental)
t
t J (as in "joke")
<: I) (pharyngeal)
t X (velar fricative)
;;;, d
~ g (voiced interdental)
J r
J z
(.)" s
(.)" s (like sh in "english")
<..r' ~ (emphatic s)
<..r' <;I (emphatic voiced dental)
1. (emphatic unvoiced dental)
.b q (emphatic voiced interdental)
t c (pharyngeal fricative)
t g (voiced velar fricative)
._j
f
..
<.;! q (velar glottalized plosive)
.!.)
k
0 n
h
J w
~ y
Vowels:
central open: a
front closed:
back closed (rounded): u
long vowels: aa, ii, uu.
X
PREFACE
This study has a twofold goal. First, it investigates the internal structure
of words and clauses in Standard Arabic (SA), in the light of recent
developments of Government and Binding Theory (GB). Second, it argues
for a specific theory of typology, and proposes a particular view of how
parametrization can be construed and executed.
SA is a language used throughout the Arab world, in contrast to specific
local dialects which are limited to a particular area. The language has a
number of features which make it particularly suitable for cross-linguistic
comparative morpho-syntax, as well as research in different modules of
the theory of grammar. SA morphology is essentially non-concatenative,
though a rich analytic affixation system makes word formation hierarchical.
Word order in SA is basically VSO, but the language has alternative SVO
structures as well. Sentences can be 'nominal' (i.e. with no verb or copula
realized at surface structure), or verbal. Arguments can be left syntacti-
cally unexpressed (i.e. SA is a null argument language). SA is an agreement
language, with a rich and complex agreement system interacting with word
order, pronominal incorporation, and expletive structures. It also has a
productive morphological case system. Tense, Aspect, Modal, and Negation
properties interact in intriguing ways. Finally, SA's DP system exhibits
interesting complementary distributions between overt determiners, genitive
complements, and possessive markers. It also uses different licensing
strategies for Genitive Case marking.
These properties are shared only to some extent by those languages which
have been extensively studied in the GB literature (in particular Romance
and Germanic). It would thus be an important contribution to linguistics
to describe the salient properties of a language which seems to have no
real analogue among languages which have been extensively described:
by doing so one could confront the results reached with the views
established in the field. As it turns out, our treatment of the problems of
SA on 'its own terms' provides new insights for the investigation of similar
or parallel phenomena in other languages, including typologically different
languages like English. Since SA has been poorly studied, the method
adopted here is to focus on its properties as a point of departure for more
general and theoretical issues. This does not preclude cross-linguistic
comparison, nor the examination of theoretical consequences.
Two additional points are in order here. First, Arabic traditional gram-
matical thought is rich in informal ideas about SA, although it is coercive
and has had damaging consequences for the development of modern
XI
XII PREFACE
1. UG AND PARAMETRIZATION
Since its inception, the generative programme has been concerned with
issues centering around the knowledge of language, hence contributing to
the development of what has been termed the cognitive revolution. Among
the significant issues addressed are the following:
(a) what is the nature of linguistic knowledge?
(b) what is the origin (or source) of this knowledge?
(c) how is this language used?
Humans, unlike other creatures, can attain knowledge of languages like
English, Japanese, or Arabic. It is reasonable to take this knowledge to
represent a real state of the mind/brain, or a mental 'organ' with a specific
structure and properties. Linguistic theory shares with other cognitive
sciences the view that a number of characteristics of the mind/brain can
be approached as computational systems, with rules forming and modi-
fying representations, and used to execute actions or interpretations. By
studying natural languages, generative grammarians have contributed to
clarifying the nature of the systems of knowledge, belief, understanding,
interpretation, etc., and helped to develop elaborate formal and computa-
tional systems (see Chomsky, 1986a, and 1989b and c).
The knowledge of language, or its development/learning, can be thought
of in terms of transitory mental states. The initial state (which is part of
the biological endowment), often called language faculty, evolves under
normal conditions of interaction with (limited) social experience to a steady
state at a maturational stage, via intermediate states. These states repre-
sent the forms of the language acquired, or the knowledge of a particular
language.
The knowledge of a language is represented by a particular generative
grammar, or a theory of the state of the mind/brain at a given stage. As
for Universal Grammar (UG), this is construed as a theory of the initial
state, or a language acquisition device which interacts with a limited
linguistic experience to attain the knowledge of a particular language.
What generative grammar means is a grammar which explicitly enumer-
ates the forms and expressions of the language. The language, in this view,
is a particular generative procedure which associates with every expres-
sion of a language La representation of form and meaning in L. Generative
grammars are more concerned, however, with intensional aspects of
language (i.e. with what Chomsky (1986a) calls the !-language) than with
2 CHAPTER I
the external product or the different aspects of behaviour, or even the set
of expressions in the language (i.e. the E-language in Chomsky's termi-
nology; see Chomsky, 1989b and c)
It is by now, I hope, clear that a study of a particular language cannot
be carried out while ignoring the issues mentioned, and without taking
account of the results of comparative, general, and psychological linguis-
tics. The description of a particular language is a description of the
judgments of the speaker-hearer about the expressions which he can produce
or interpret. In the 'learning' process, the external experience plays only
a minor role, due to the poverty of stimulus, or to what Chomsky calls
Plato's problem. 1 The speaker basically 'knows' the language he is
'learning', being innately equiped with a system of linguistic principles
and constraints which limits the set of possible languages. He then fixes
the parameters on the basis of limited empirical evidence. 2
On this view, a language is a set of specifications for parameters (or
variants) in an invariant system of principles of UG. The constructions to
be studied in a particular language (or cross-linguistically) can be regarded
as largely epiphenomenal, but more interestingly, as taxonomic instantia-
tions of the interaction of variant and invariant principles of grammar
(cf. Chomsky, l989a). Consequently, any approach to a particular language
is simultaneously concerned with the following two problems:
(d) manifestations of the general language faculty.
(e) systematic differences among languages, i.e. instantiations of the
particular language faculty acquired.
Two views of grammatical parametrization have produced fruitful results
in recent years. First, a parameter has been viewed as essentially a corre-
lation of grammatical properties. As a corollary, fixing the value of a
parameter one way or another would trigger the existence or absence of
all the clustered properties. The pro drop parameter, for example, has been
thought of in this way (work by Rizzi, 1982, Kayne, 1983, and Chomsky,
1981 essentially illustrates this line of research; see also White, 1985 for
applications to language acquisition). Second, parameters of UG can be
thought of as being typically lexical. Thus " ... each parameter refers to
properties of specific lexical items of the lexicon, or to categories of lexical
items ... ". Consequently, " ... there is only one human language, apart
from the lexicon, and language acquisition is, in essence, a matter of
determining lexical idiosyncracies" (Chomsky, 1989a). Moreover, the
functional component, as mentioned earlier, plays an important role in
determining how language learning proceeds. The latter view is adopted and
elaborated throughout this work in what I have called a Multi-Valued
Functional Parametrization system (see in particular Chapters 2 and 3 for
details).
The Principles-and-Parameters model proposed by Chomsky (1981 and
passim) is an advanced form of the generative programme which started
PRINCIPLES, PARAMETERS, AND MODULES 3
in the 'fifties. In addition to its content and subtheories and modules, the
model can be characterized by its philosophical/conceptual foundations
and methodological aims. It is expected that metalinguistic scientific
discourses would vary according to these foundations, and to the tradition
of epistemology and linguistic thought in one cultural area or another. These
variations would be a source of enrichment for conceptual, methodolog-
ical, and empirical inquiry. Thus it is enriching for modern Arabic linguistic
research to integrate a number of informal ideas and generalizations found
in traditional Arabic thought. However, a caveat is in order here: although
it might be reasonable to think that the history of thought and science has
to be linear, continuous, and cumulative, modern linguistic research is
concerned with traditional questions only to a minor extent. When tradi-
tional questions are asked, they are not posed in the same form. Hopefully,
we have made enough progress in linguistics to know that we are not
concerned with objects, questions, or methodologies of traditional thought.
Consequently, a 'compromise' with traditionalists can only be a weak one. 3
Traditionalists tend to think that almost every idea or analysis can be found
in texts written by our ancestors of the eighth or tenth centuries. Indeed,
an 'epistemological cut' with this kind of discourse is necessary, other-
wise we will contribute to the aggravation of what Chomsky calls Orwell's
problem. 4
2. THE MODEL
3. X' THEORY
Predicates differ with respect to how many and which arguments they select,
depending on whether they are transitive or intransitive, whether they take
NP, PP, or S as complements, etc. The list of arguments and their internal
organization has been termed either predicate argument structure (PAS;
see e.g. Bresnan, 1982c), or thematic grid (see Stowell, 1981). This has
led to terminological confusion in the literature: some linguists call
arguments what others call theta roles, and vice versa. Conceptually,
argument structure and thematic structure are distinct objects. For example,
verbs like darab 'beat' and caraf 'know' both have a PAS with two
arguments, but the subject is an 'agent' with the former, and an 'experi-
encer' with the latter. There are various proposals concerning the nature
of roles and their origins, how they are linked to arguments, etc. I limit
myself to the discussion of a single problem here, namely theta
assignment.
The PAS of a lexical item representing its arguments may take the form
of a configurational tree projected from a more remote lexical conceptual
structure in which arguments are represented by variables (along the lines
suggested by Hale and Keyser, passim). The Theta Criterion proposed
by Chomsky (1981) requires arguments to be realized by appropriate
constituents (NP, S, etc.) in the syntax. The mapping beween arguments and
constituents is bijective in the argument domain of a head. The Theta
Criterion has different formulations, among which is the following (original)
one:
(6) Each argument bears one and only one theta role, and each theta
role is assigned to one and only one argument.
The mechanism of theta assignment establishes a relation between the
constituents occupying positions in the syntax and variables found in
the PAS of the lexical item. For example, the verb assigns a role to its
complement, the preposition mediates the assignment of a role, the VP
PRINCIPLES, PARAMETERS, AND MODULES 7
assigns a role to the subject, etc. There is reason to think, however, that
the Theta Criterion is true of chains (a chain consists of a number of
positions), and not of single positions (see the formulation of the Chain
Condition proposed by Chomsky, 1986a in Section 5, below).
In Williams' (1980, 1981) theory of argument structure, VP assigns a
role to the NP subject only via predication. In this view, the latter is
generated outside the VP projection: it is an external argument. Other
arguments are inside VP, hence their designation as internal arguments.
GB linguists sometimes use the expressions 'internal' and 'external' roles
and 'internal' and 'external' arguments interchangeably.
Positions in which NPs receive theta roles are called theta positions,
others are non-theta or theta-bar positions. All arguments are supposed
to occur in theta positions at D-structure, if the latter is a 'pure' represen-
tation of thematically relevant grammatical relations (as claimed by
Chomsky, 1986a). This view is further sharpened by the Uniformity of Theta
Assignment Hypothesis (UTAH), proposed by Baker (1988):
(7) UTAH. Identical thematic relationships between items are
represented by identical structural relationships between those
items at the level of D-structure.
The set of theta markers might be limited to lexical categories (Vs, As,
and Ps in particular). Chomsky (1986a) assumes that N s are theta assigners
just like verbs, but Emonds (1985) questions such a view. According to
Emonds, Ns theta assign only indirectly (usually via prepositions). On the
other hand, there are reasons to think that functional categories play also
a role in theta assignment. Theta theory is then revised along the lines
proposed by Higginbotham (1985 and 1986). The mapping between
arguments and roles can be achieved through three different mechanisms
of theta assignment: theta marking, theta identification, and theta binding.
These mechanisms are discussed at length in Chapter 4.
Case is assigned to NPs (and other categories like APs and adverbial
phrases) by a governor. Case is abstract in the sense that it may be
morphologically realized or not. The Case Filter, given in (8), plays an
essential role in the Case module because it forces NPs to receive Case:
(8) *NP, if NP is lexical (has phonetic content) and has no Case.
This filter applies only to NPs in Chomsky's formulation. It has to be
generalized, however, to APs and AdvPs, as the Arabic facts suggest. 6
Case assigners in Chomsky (1981) are essentially [-N] categories, i.e.
V and P, and also I. Chomsky (1986a) adds N and A to the list, but
distinguishes two types of cases: structural Case (which is assigned by virtue
8 CHAPTER I
7. BINDING
8. PREDICATION
The Move-a rule applies freely, with no constraints on its application, but
its outputs are constrained by various principles belonging to different
modules and subtheories, including bounding theory, the theory of empty
categories, Case theory, and Theta theory. At first glance, there appear to
be constraints on movement, but a closer examination reveals that they
can be derived from independently motivated principles.
Consider first the type of movement which is a substitution. Chomsky
(1986b) provides the following general properties of substitution:
(33) a. There is no movement to the complement position.
b. Only xo can move to the head position.
c. Move-a applies to xo or to xmax.
12 CHAPTER I
(33a) can be derived from the Theta Criterion, since the latter limits
movement only to non-theta positions. The complement position being a
theta position, a movement to it would yield a chain with two theta
positions, an option barred by the Theta Criterion. As for (33b) and (33c),
Chomsky proposes deriving them from the Structure Preserving Constraint
of Emonds (1976). 7
There are a number of movements which can count as substitutions.
Among these is NP movement in passives, from object to subject position
(as in standard accounts), or movement of the thematic subject from Spec
V to Spec T, and Spec AGR, to become an inflectional subject, or even
movement of the wh-phrase to Spec C (see Chapter 2). Considerations of
Case, scope, and morphology motivate or force these movements.
Let us turn now to adjunction. The movement of V to T, and to AGR,
is a case of adjunction. It is motivated by morphology, since the V is a
root which has to attach to a vocalic (inflectional) support. The movement
is head-to-head, which is constrained by the Head Movement Constraint
(HMC) proposed by Travis (1984) (see also Chomsky, 1986b). Baker (1988)
formulates HMC as follows:
(34) xo can only move to yo which properly governs it.
Chomsky (1986b) proposes the derivation of this constraint from ECP.
In sum, the Move-a rule applies freely, but its outputs observe general
well-formedness principles at play in different modules and components
of the grammar (syntax, lexicon, PF, LF, etc.).
10. MORPHOLOGY
(35) a. Words have heads, in the same way that syntactic phrases have
heads. The features of the head are inherited by the complex
form, and become properties of the whole.
b. Affixes (and especially suffixes) head words.
c. Affixes (like content words) are lexical categories. They are
categorially labelled as N, V, A, etc.
d. Affixes have predicate argument structures.
e. Affixes have whatever properties stems or words have, except
that they are bound morphemes.
f. There is no (principled) distinction between derivational and
inflectional morphology. Derivational morphology can be inte-
grated into inflectional morphology.
Selkirk (1982) has proposed that word formation rules take the form
of phrase structure rules, which operate concatenation of morphological
classes (of components). Applied to Arabic, this apparatus yields the
following rules:
(36) a. root ~ root root
b. root ~ af root
c. root ~ root af
d. stem ~ root af
e. word ~ stem af
(af = affix).
Observe that Arabic has no productive rule like (37), which concatenates
two words:
(37) word ~ word word
This rule forms compounds in French like coffre-fort, chou-fleur, rouge
gorge, etc. It is also productive in English (as in heart broken, living room,
school teacher, etc.). The process is not regular in Arabic, however. What
is found is so-called naht 'shaping' by traditional grammarians, that is,
the formation of a root from two (or more) roots. As for compounding, it
is very marginal, and is only found in proper names like baC[a-bakk or
hadra-mawt (two names of Arabic cities), etc. It is also found in some
particles like qalla-maa 'rarely-that' ka)fa-maa literally 'how-that', meaning
'however, whatever', etc.
DSW argue that identifying morphological heads is different from
identifying syntactic heads. The latter are determined by their categorial
label and their type (or the number of primes or bars). Thus it is easy to
say that V or V' is a head of V" in (38), for example, but it is difficult to
do so in (39):
14 CHAPTER I
(38) V"
~
N" V'
~
V N"
(39) a. N b. v
~ ~
N N V N
NOTES
Plato's problem is the following: 'why do we know so much, given that we have such
limited evidence?' A similar question has also been raised by Bertrand Russell. The poverty
of stimulus (that is, the fact that the language learner faces in his environment only very
limited linguistic experience, yet he manages to 'learn' all the properties of his language)
is behind most generative apporaches to UG.
2 Some psychologists, by contrast, defend the view that humans acquire language by using
'generalized learning' mechanisms or 'overlearning'. Language is conceived as a system of
habitual dispositions, which is covered by available experience. Producing or interpreting new
forms is effected by analogy. The problem of the poverty of stimulus, however, shows that
overlearning cannot be the correct approach. See Chomsky (1986a and 1989b).
3 The position of Chomsky ( 1966) with respect to traditional thought has often been
misinterpreted as acknowledging that our ancestors are precursors of modern scientific issues,
and that the medieval issues are directly relevant to modern linguistic thought. I interpret
that position as a weak compromise, aiming mainly to rehabilitate (globally) previous research
in the field, against the attacks of structuralists. This cannot mean, however, confusing the
conceptual, methodological, and empirical aims of different traditions and schools.
4 Chomsky characterizes Orwell's problem as follows: 'why do we know so little given
that we have so much evidence?' The pure traditionalist position, which is strong in the
Arab world, is often one of the greatest obstacles to knowledge. According to this view,
only ancestors 'know' in the first place, and only they have the right answers.
5 This view contrasts with lexicalist views, which require lexical entries to be fully expanded,
and hence to contain redundant information. Cf. Halle (1973), Jackendoff (1975), and Bresnan
(1982a) on the matter and, for a criticism, see Chapter 2.
6 Chomsky (1986a), following a suggestion by Aoun, proposes deriving the Case Filter from
the Visibilty Condition given in (i):
(i) An NP position which is the head of a chain (i.e. the last position of a moved category)
can only bear a theta-index if it receives Case.
An SNLLT reviewer has pointed out that the conditions in (33) apply equally to movement
as adjunction (in some cases vacuously). This observation appears to be true, although the
derivation of these constraints differs depending on which type of movement is involved.
For example, a substitution which involves the Camp position violates the Theta Criterion,
but an adjunction to the same position does not (as far as I can tell).
CHAPTER 2
In this chapter I address three important and related issues in the theory
of Arabic clause structure: (a) word order, (b) verb agreement patterns,
and (c) Cases of subjects. I establish that Arabic is essentially a VSO
language, although it also seems to belong to a 'mixed' VSO/SVO type.
This typology correlates with properties of inflectional categories heading
clauses, in particular the properties of AGR(eement) morphemes. I claim
that AGR in VSO languages is both poor (in that it is specified at most
for Gender), and weak (in that it does not play a role in Nominative Case
assignment and/or checking). But although some AGR morphemes in these
languages are rich (in that they exhibit Number specification (NUM), in
addition to Gender (GEN)), as is the case in Arabic, AGR remains weak.
It is unable to protect subjects in SVO structures from external governors,
and it does not play a (significant) role in Nominative assignment and/or
checking. It is argued that this weakness of the inflection in VSO lan-
guages (even of the mixed type) is traceable to a parametric categorial
property of AGR: AGR is nominal in some languages, and non-nominal
in others.
As regards syntactic rules involved in constituency characterization, I
assume that the grammar of Arabic instantiates a canonical phrase struc-
ture like (I):
(1) IP
~
I'
~
I VP
~~ V NP
16
WORD ORDER, AGREEMENT, AND CASE 17
rich AGR on the verb. Poor agreement prevents such an operation from
taking place, thus resulting in a VSO structure. Part of this chapter is devoted
to investigating how and why this correlation between word order and
agreement types holds. Moreover, cross-linguistic distribution of rich and
poor AGR is analyzed, together with the distribution of thematic subjects
in preverbal or postverbal positions. I propose a principle regulating the
positioning of subjects with respect to the AGR inflection, as well as their
feature sharing relationship, which I name the AGR Criterion.
The two raising operations postulated have fairly standard properties, and
pose no particular problem for movement theory. 1 The base-generation of
the subject inside VP has gained quite wide acceptance, although some
GB grammarians still assume that D-subjects are generated in Spec of IP. 2
Let us suppose that Chomsky's ( 1986a) Principle of Full Interpretation is
to generalize to D-structure, and that the occurrence of NPs at D-struc-
ture positions is licensed only by virtue of a thematic relationship, then a
thematic NP subject would occur in Spec of VP at D-structure, not in
Spec of IP. If that is true, then the analysis based on the configuration in
(1) and the raising operations postulated are conceptually preferable. 3
The conceptual merits of this canonical approach are not shared by a
number of recent analyses proposed for VSO languages (and/or struc-
tures). Here are some claims and hypotheses made in the literature: 4
A. The subject is in Spec of IP at D-structure and
(a) lowers to adjoin to V at S-structure or
(b) lowers to adjoin to VP.
B. The subject is in Spec of VP at D-structure, and V raises to adjoin to
IP.
These analyses are illustrated in (2), (3), and (4), respectively:
(2) IP
~
(NP) I' (NP)
~
I VP
~
v
~
v
18 CHAPTER 2
(3) IP
~
(NP) I' (NP)
(~
~ VP
~
V NP
(4) IP
~
V IP
~
I VP
~
NP V'
~
V NP
Adjunctions in (2) and (4) are not canonical. In Chomsky (1986b ), adjunc-
tion of xo is limited to X 0 , and adjunction of XP to XP. In other words,
the two constituents of an adjunction structure created by movement have
to be of the same bar level. As Chomsky points out, this requirement can
be interpreted as an instance of the Structure Preserving Constraint proposed
by Emonds (1976). The adjunction in (3) is canonical but costly, assuming
Chomsky's Principle for the Economy of Derivations. 5 The raising and
canonical approach (as described) thus appears to be conceptually favored,
compared to other proposals made in the literature. Furthermore, it is also
preferable on empirical grounds, as I will argue.
Note that one feature of the structure proposed in (1) is that the inflec-
tion has a Spec position. The existence of this position will be motivated
against the widely spread view that VSO languages (especially Celtic)
generate no Spec of IP positions. 6
This chapter is organized as follows. In Section 1 I first examine the
properties of VSO order, and argue that it derives from a configurational
IP and VP structure in which V raises to I. I then argue for the existence
of SVO structures (as distinct from left dislocated structures), and motivate
the rule of Subject raising. In Section 2 I analyze how subjects can (or
cannot) undergo raising, depending on the properties of AGR, and in con-
formity with what I call the AGR Criterion. I then turn to the problem of
licensing poor agreement, and argue against the view that it is licensed under
direct Spec-head agreement with an expletive. I also examine the various
distributions of expletive forms cross-linguistically, and show that their
WORD ORDER, AGREEMENT, AND CASE 19
features depend on the properties of AGR, although the reverse is not true.
In Section 3 I discuss the variety of Cases assigned to Arabic subjects,
and how they are licensed and checked. I investigate why subjects are
uniformly Nominative in postverbal positions, and why they are (normally)
non-nominative in SVO contexts. In Section 4 I analyze some consequences
of the analyses adopted for the Extended Projection Principle, expletive
licensing, VS and SV order typology, extraction, and Case assignment. I
also discuss further motivation for V-to-I raising.
In this section I argue that two raising operations may take place to derive
the S-structure of Arabic clauses: (a) an xo raising of V to I (to derive
VSO structures), and (b) an XP raising of the NP subject from Spec of
VP to Spec of IP (to derive SVO).
b. ?a"t;ay-ta-nii ?iyyaa-hu
gave-you-me it
You gave it to me.
If pronominal affixes are pronouns incorporated onto the verbal governor
(as we will see in Chapter 3), then these facts can be captured if we assume
that the incorporation process preserves basic order.
There are contexts in which pronominalization forces the shift from basic
order, hence providing a clue for recovering it. Consider the following
examples:
( 14) a. sakaa 1-walad-a ?abuu-hu
complained the-boy-ace father.nom-his
(There) has complained about the boy his father.
His father has complained about the boy.
In (19a) and (19c), the reflexive is bound within Sand NP, respectively. The
ungrammaticality of ( 19b) can be explained if we assume that the reflexive
subject asymmetrically c-commands its antecedent. As a consequence, the
former cannot be bound by the latter, and the structure ( 19b) violates
principle A of BT. Structure (19b) also violates Principle C, since a name
is bound, while it should be free. Let us turn now to embedded contexts,
as in (20):
(20) a. qanna r-rajul-u nafs-a-hu gabiyy-an
thought the-man-nom self-ace-his silly-ace
The man thought himself to be silly.
Appendix 1 and Chapter 4 for details about the split INFL issue). The
verb raises to T and AGR (which are lower than the other inflectional
categories), to support these affixes morphologically. There is no reason,
however, to think that the inflected verb moves to Mod and Neg, and that
all these segments move further to C. Mod and Neg constituents are not
(generally) bound morphemes and, as far as I can tell, there is no reason
for them to move. The S-structure of the embedded sentence in (27) is
then presumably (28), where I contains T and AGR, and V has moved to I:
(28) CP
~
C ModP
I~
?an Mid ~
qad Neg IP
I~
laa I VP
I~
ya?tii NP V'
I
Zayd
This structure accounts in a straightforward manner for the distribution of
inflectional constituents, without postulating any (unmotivated) movement
of Neg and Mod. 17 There are further descriptive reasons to think that V
raising occurs even in Arabic SVO structures (preceded) by C, and that V
stops in I. These questions are dealt with in the next
subsection.
Having argued that Arabic is a VSO language, I turn now to the task of
showing that it also exhibits structures which are best characterized as
instantiating SVO order. Such constructions have often been analyzed as
cases of left dislocated structures, but we will see that this view cannot
be maintained. 18
Consider the following sentence:
(29) al-?awlaad-u jaa?-uu
the-children-nom came-3.m.pl.
The children, they came. The children came.
As the translation indicates, this sentence is ambiguous in a way that neither
(30) nor (31) are:
28 CHAPTER 2
C
~
AGRP
~
~
The next subsection is devoted to analyzing the properties of these two
structures as well as their distribution.
Preverbal subjects and topics (in the intended sense) have referential
properties which set them apart. For example, Arabic topics are neces-
sarily definite (as in (30)), whereas preverbal subjects can be indefinite
(as in (31)). Indefinite preverbal subjects cannot be non-specific, however.
They have to be specific, generic, or bound by a quantifier, as in the
following examples: 19
(33) a. laa jund-a y-astatiic-uu-na duxuul-a 1-macrakat-i
no soldiers-ace 3-can-pl.m.-indic entering-ace the-battle-gen
No soldiers can enter the battle.
Having established that preverbal NPs in (31) are subjects, not topics, it
is reasonable to think that they have raised from their base position (in Spec
of VP) to a Spec of IP position at S-structure (or Spec of AGR, as dia-
grammed in (32b) above). Facts about agreement and Case marking of
subjects support this view.
There is evidence based on agreement facts which indicates that subjects
may move to Spec of AGR at S-structure. Subject verb agreement is limited
to GEN when the subject is postverbal (call it poor AGR), but must involve
NUM (and PERS(on)) when the subject is preverbal (call it rich AGR). Thus
the following contrasts obtain: 23
32 CHAPTER 2
correlates with agreement and case properties. These two issues will be dealt
with in the next two sections.
2. AGREEMENT
The first clause of (55) accounts for the ill-formedness of (54a) since
AGR there is not licensed. The second clause accounts for the ungram-
maticality of (54b) since the thematic NP there is not licensed in Spec of
AGR.
The principle (55), which accounts for the Arabic facts, has a wider
cross-linguistic application than it appears at first sight. Consider, e.g.,
the following constructions in French:
(56) a. II est venu trois enfants.
b.* lis sont venus trois enfants
c.* Est venu trois enfants.
d.* Sont venus trois enfants.
(57) a. Trois enfants sont venus.
b.* Trois enfants est venu.
Leaving aside the problem that French (unlike Arabic) is not a null subject
language (hence the ungrammaticality of (56c) and (56d)), the opposition
(56a)-(56b) is parallel to that in (53a)-(53b), and the opposition in (57)
is parallel to that in (54). The two clauses of the AGR Criterion will then
account for the French judgments in the same way. 26
North Italian dialects also exhibit similar contrasts. According to Brandi
and Cordin ( 1989), Trentino (T) and Fiorentino (F), like standard Italian,
allow so-called subject inversion with any type of verb (ergative or not),
and with any type of subject (definite or not). Unlike Italian, however, verbs
do not agree with postverbal subjects in NUM, while they do with preverbal
subjects. The following examples show that only poor agreement is possible
with postverbal subjects in T and F:
(58) a. Gli e venuto delle ragazze (F)
b. E' vegm1 qualche putela (T)
is come some girls
Some girls have come.
It seems then that the use of poor or null agreement with postverbal subjects
is widely spread among languages, and that VSO languages make essen-
tial (or exclusive) use of this option. I return to licensing conditions of
poor agreement in the next subsection. 28
In addition to poor agreement, languages may also make use of rich
agreement morphology, its distribution being regulated by the AGR
Criterion. There are languages, however, in which agreement patterns on
predicates do not vary depending on whether subjects are placed before
or after predicates. This situation seems to pose a serious problem for the
above version of the AGR Criterion, since, in both contexts, AGR is rich
(involving NUM specification). Standard Italian is one instance of those
languages, as we have seen earlier, and Moroccan Arabic (MA) is another
instance. In MA, the verb must agree in NUM with the thematic subject,
be it preverbal, as in (69), or postverbal, as in (70):
(69) 1-ulad ja-w
the-boys came-pl.
The boys came.
AGR feature specifications. They also indicate that a revision of the AGR
Criterion (which would extend to those cases) is necessary.
In order to account for these systematic differences, we first need to allow
rich AGR to occur with postverbal subjects. Let us then revise the first
clause of the AGR Criterion as follows:
(75) Rich AGR is licensed by
(a) an argumental NP in its Spec, or
(b) by a chain of which one member is an argumental NP.
As revised, this clause makes rich AGR licensing possible either through
argumental NPs, or through expletive chains (of which one member is
argumental). It follows that constructions like (70), (73), and (74) in
languages like MA (and presumably Italian) are regulated. The following
question, however, remains to be answered: why can null or poor agreement
not be generated with postverbal subjects in those languages, i.e. why are
cases like (71) and (72) ruled out? This question is dealt with in the next
subsection.
There are essential questions that arise with respect to regulating null or
poor agreement patterns:
(a) how are these patterns licensed?
(b) why are they possible (or required) with postverbal argumental subjects
in Arabic, French, and North Italian dialects, but impossible in MA
or Standard Italian?
(c) why is GEN specification allowed as an instance of poor agreement
in some languages (e.g. Arabic and Celtic), but not in others (e.g. French
or North Italian)?
An answer to (a) which has been frequently provided in the literature
is that poor agreement arises as a result of a Spec-head agreement with
an expletive subject. Since the latter is specified (by default) as singular
masculine, then AGR on the verb must have the same default values. This
proposal, which will be referred to as the expletive hypothesis, has been
advanced by e.g. Mohammad ( 1990) and Fassi (1988c) for Arabic, and
Pollock (1983, 1985) for French. There are serious reasons to think that this
hypothesis cannot be maintained.
3. CASES OF SUBJECTS
This section deals with the variety of Cases assigned to Arabic subjects.
While subjects of VSO and VOS structures are marked with Nominative,
under government by internal I, Cases of subjects found in SVO struc-
tures indicate that the latter are accessible to external governors, which
assign them non-nominative Case. In this respect, they behave like subjects
of verbless (or so-called nominal) sentences. These properties correlate with
the occurrence of rich AGR in these constructions. The latter is shown to
be nominal in character. It is also weak, in the sense that it does not act
as a protector for the subject, in contrast to what happens in languages
like French or English, in which AGR is strong and non-nominal, as I
will explain.
(96) CP
~
C IP
I~
?inna I' NP
~ ~
I PP rajul-an
~
e P'
~
fii d-daar-i
One way to prevent Accusative case from being assigned in VOS (and VSO)
is to appeal to the Minimality Condition proposed by Chomsky (1986b).
In essence, Minimality is a locality condition on government, preventing
a governor from governing inside the domain of another governor. The
contrast between subjects of verbless sentences and those of VSO and
VOS indicates that I (or more precisely T in I) counts as a head governor
for Minimality, provided it is lexically supported. In verbal sentences, it
48 CHAPTER 2
( 101) Minimality:
X governs Y only if there is no Z, such that
(a) Z is a potential governor for Y.
(b) Z c-commands Y and does not c-command X.
The lexical requirement on head governors can be embedded m the
definition of potential head governors as follows:
3.3.1. Nominative
In (112), the prefix on the auxilliary verb marks PERS and GEN, the suffix
marks Mood, and the internal vowels Tense or Aspect; in the thematic
verb, the suffix marks AGR, and the internal vocalic pattern Tense and
Aspect. The construction (113) shows that Negation can be placed inside
the functional domain in which the thematic verb is included, suggesting
that the latter is tensed. 46
It appears, then, that a sequence of two finite verbs is possible in Standard
Arabic. In this language, there are no bare verbs, no verbal (non-finite)
participles, and no infinitives (see Chapter 4 for details). The construc-
tions given here thus provide a serious base for the existence of a structure
with two finite verbs. The question then is: why are these structures found
in Arabic, but not cross-linguistically? For example, the English or French
counterparts of these constructions are ungrammatical:
(114)a.*John had ate.
b. John had eaten.
(115)a.*Jean avait mangeait.
b. Jean avait mange.
A traditional answer to this question has been something like the state-
ment in (116):
(116) An auxiliary verb selects a VP.
Gueron and Hoekstra (1988), for example, assume the following
functional definition of auxiliaries:
(117) An auxiliary verb governs a verbal projection [a VP; FF]; a
non-auxiliary verb governs either a nominal projection or a
verbal projection.
While statements like these might be correct for English or French, they
are not general enough to carry over to Arabic. 47
Anticipating somewhat the argument to be presented in Chapter 4, I
will assume that auxiliaries select inflected VPs or IPs, not bare VPs.
Moreover, an auxiliary licenses its IP complement by T(ense)-marking it
(as in Zagona, 1988; Fabb, 1984; and Gueron and Hoekstra, 1988, among
others). This is the case in English or French, where VPs can have non-
finite morphology. This possibility is not available in Arabic, however.
The closest counterpart to the English (114a) or the French (115a) is ungram-
matical in Arabic, as the following construction shows:
(118) *kaan-a ?akal-a r-rajul-u
was ate the-man-nom
The man had eaten.
How is such a structure to be excluded? A quite straightforward solution
WORD ORDER, AGREEMENT, AND CASE 53
The ungrammaticality of sentences like (108), (109), and (118) can now
be seen as a direct consequence of KRP. More precisely, tensed or non-
nominal IP would resist T-marking, and the IP there would not be licensed.
The only possibility left in such cases is then to 'nominalize' IP via rich
AGR, and to license the whole IP as a nominal argument of the auxiliary
(see Chapter 4 for details).
Summarizing, I have shown that, unlike French or English subjects, which
are uniformly Nominative in finite clauses, Arabic subjects may be non-
Nominative, as well as Nominative, depending on the properties of AGR
in I. When the latter is non-nominal (and hence strong), subjects are marked
as Nominative. When AGR is nominal (and hence weak), subjects are
non-Nominative. Differences of behaviour in ECM and auxiliary struc-
tures are also traced back to AGR nominality.
It is then clear that the position in which the expletive occurs is not
theta-marked. To see this, consider the D-structure of (132), given in (137):
(137) IP
~
I VP
~
Spec V'
I~
u.
-tu V IP
-b~
I VP
~
?axuuka V
I
jaa?a
In this structure, both the lower and higher verbs move to I, to support T
and AGR in I. The expletive is generated in Spec of AGR, and raises from
there to incorporate onto the higher verb.
Given this picture, it is clear that the expletive does not occur in a
theta-marked position, nor is it assigned a theta-role. It might be suggested
that the latter, though not receiving a theta-role, does belong to a chain which
receives one. If the expletive and the postverbal subject were coindexed and
form a chain, then the former would belong to a theta-marked chain, in
which the 'foot' of the chain (?axuuka). is theta-marked by V'. Note,
however, that this chain cannot be a well-formed chain, assuming Chomsky's
( 1986a) General Condition on Chains to be valid. The latter states the
following:
(138) If C = (a 1, • • • , a") is a maximal Chain, then an occupies its
unique theta-position, and a 1 its unique Case-marked position.
By this condition, only one member of the chain is able to receive Case.
In the S-structure of (132), however, the expletive is Case-marked by the
matrix V as Accusative, and the thematic embedded subject receives
Nominative (presumably under government by T). This indicates that the
expletive and the thematic subject do not belong to the same chain, and
that the former does not belong to a theta-marked chain.
further evidence that French and English have preverbal (AGR) subjects,
but Arabic does not.
Consider first a problem that arises with respect to subject extraction
in Arabic, but not in French or English. As explained earlier, when a subject
is questioned or topicalized, rich agreement on predicates is required. Poor
agreement leads to ungrammaticality, as the following contructions show:
(139) * ?ayy-u rijaal-in jaa?-a
which-nom men-gen came-J.s.m.
Intended to mean: Which men came?
(140) *r-rijaal-u jaar-a
the-men-nom came-J.s.m.
Intended to mean: The men came.
In these contexts the subject presumably moves to Spec C. Agreement
here looks like a case of long distance agreement in the sense of Kayne
( 1987). The latter is broken up into strictly local configurations, yielding
the desired results. Regarding these constructions, the following questions
arise: why can't an (empty) expletive appear there (e.g. in Spec of TP or
AGRP), and hence license the poor agreement on the verb? Why can't
Arabic have constructions exactly like French ones, in which the thematic
subject is extracted over an expletive subject, as in (141):
(141) Combien de filles est -il venu?
Consider first the problem of locality in agreement. Kayne has argued that
in cases of 'long' parti.ciple agreement like (142), the chain of movement
contains a number of empty categories:
(142) Je me demande combien de tables Paul a
I wonder how many of tables Paul has
repeint-e-s
repainted-/-pi
I wonder how many tables Paul has repainted.
If that is true, then there is a more strongly local agreement relation than
it first appears, as is represented in (143):
(143) [... ] combien de tables, Paul a [el AGR, repeint-e-s [el
Kayne assumes that the participle is headed by AGR, and that the trace
which is adjoined to AGR makes agreement local (see his paper for details).
Adopting Kayne's idea of locality, we can say that the NP moving from
subject position has to move either to Spec of AGR, or to a position adjoined
to AGRP, hence ensuring that its trace would c-command AGR in a local
configuration. Note, however, that this does not ensure that a poor AGR
WORD ORDER, AGREEMENT, AND CASE 61
would not appear there, yielding the ungrammatical (139) and (140). In
order to avoid this we have to ensure that subjects would never be 'directly'
extracted from postverbal position. The movement to Spec of C must
occur through Spec of AGR, triggering rich AGR, in conformity with the
AGR Criterion. If the movement is not through Spec, but by adjunction
to AGR, then nothing forces AGR to be rich. When objects are extracted,
they are presumably adjoined to AGR, but verbs do not agree with them.
As far as I can tell, there is no reason to think that rules of Adjunct Head
agreement exist, along with Spec Head agreement rules. On the contrary,
there is reason to think that when AGR is rich (and hence nominal), adjunc-
tion to it is not allowed, as we will see. Given these considerations, we
might want to limit configurational agreement to NPs located in Spec or
Camp positions, in line with recent work by Chomsky (1989) and Rizzi
(1990 and 1991). 52
The analysis of more complex cases of extraction also supports the
strict locality requirement on AGR, built in the AGR Criterion. Consider
the following examples:
(144)a. ?ayy-u rijaal-in t-aqunnu ?anna-hum jaa?-uu
which-nom men-gen you-think that-them came-3.pl.m.
Which men do you think that (they) came?
b.*?ayy-u rijaal-in t-aqunnu ?anna-hu jaa?-uu
which men you-think that-it/him came-3.pl.m.
c.* ?ayy-u rijaal-in t-aqunnu ?anna-hu jaa? -a
which men you-think that-it/him came-3.s.m.
The fact that (144b) is ruled out suggests that the wh-word cannot bind
the agreement marker (and/or pronominal) on the verb because it is not
'close' enough, and the pronoun on the complementizer does not agree with
the wh-word. It also shows that if the embedded topic there is an exple-
tive, then it would prevent the binding from being local. The grammaticality
of (144a) shows that the topic pronoun makes the binding local. As for
(144c), it is ruled out because there is no agreement between the wh phrase
and the verb, suggesting that movement has not operated in stages through
specifiers, or that expletive agreement has taken place.
Let us turn now to the fact that preverbal subject expletives cannot
occur with questioned or topicalized Arabic subjects, unlike the situation
in French, for instance. In this language expletives are licensed by
Nominative, and thematic subjects have some inherent Case which they pre-
sumably carry when they move. The situation in Arabic is different.
Expletives are not independently Case marked by AGR. As such, they are
not licensed in preverbal position. Moreover, the movement of the subject
preverbally forces the activation of AGR features (through local agree-
ment links), in conformity with the AGR Criterion.
62 CHAPTER 2
4.3. Extraction
Let us turn now to extraction over complementizers. The two Arabic Cs ?an
and ?anna behave differently with respect to the traditional [C e] filter.
While ?an can be adjacent to a trace of a subject, ?anna cannot:
(157)a. ?ayy-u rijaal-in J:!asib-ta ?an jaa?-uu
which-nom men-gen thought-you that came-3.pl.m.
Which men do you think that came?
Note that even if the wh-word surfaces in the accusative (i.e. the case
assigned by ?anna), the construction is still ungrammatical.
This contrast indicates that the complementizer Filter is descriptively
incorrect. There are different ways to derive the difference in grammati-
cality between the two constructions. We can explore a difference between
the two Cs in terms of Case assigning properties: ?anna is a Case assigner,
and ?an is not. The Case that ?anna discharges seems to require that its
bearer be governed by ?anna at S-structure, in order for it to be checked.
This is not true of ?an. Note that the same analysis can be extended to
exclude (152), given that ?inna there is a variant (in root clauses) of the
complementizer ?anna. But the construction is also excluded because AGRP
WORD ORDER, AGREEMENT, AND CASE 67
b. ?a zayd-un jaa?-a
Q Zayd-nom came
Is (it) Zayd (that) came?
There are differences in distribution between the two particles that do not
concern us here (for details, see Fassi, 1981). Arabic can also form
yes-no questions by making use of intonation only, as in (159):
(159) t-ul)ibbu-haa?
you-like-her
You like her?
On the other hand, Arabic not only has simple wh questions, as seen above,
but also multiple wh questions, as in (160):
(160) man <;!arab-a man bi-maagaa?
who beat who with-what
Who beat whom with what?
Arabic has also so-called 'echo' questions, where the wh word may remain
in situ, as in (161):
(161) jaa?-a man
came who
68 CHAPTER 2
b. *<;larab-hu-tu
beat-him-/
*I beat him.
(164b) is ungrammatical because the object affix precedes the subject. Now,
when the object is affixal, and the subject is not, the object must precede
the subject:
In this structure the object affix adjoins to the right of the complex xo
adjunction, which includes the subject, but the result is (surprisingly)
ungrammatical.
It might be suggested that a distinction can be made between affix-like
elements (which are phonologically dependent) and clitic-like elements
(which are relatively independent). But even if this distinction can be
supported, it is easy to show that Arabic subjects do not behave like adjoined
clitics.
70 CHAPTER 2
The fact that the light PP cannot be separated from the verb (even by a
parenthetical) suggests that it may have been adjoined to V and forms an
xo constituent with it. The parenthetical will then have to occur outside
this X 0 , hence the only possibility is (170c). In contrast, the subject being
outside the xo in which the verb is located can be separated from the verb
by both the light PP and the parenthetical, as (170a) and (170c) show.
that V-to-1 has taken place, leaving the adverb in its original position (i.e.
adjoined to VP):
( 178) a. 1-fuqahaa?-u y-aquul-uu-na caadat-an maa laa
the-scholars-nom 3-say-pl.m.-indic usually what not
y-afcal-uu-na
3-do-pl.m.-indic
Religious scientists usually say what they do not do.
b. *1-fuqahaa?-u caadat-an y-aquul-uu-na maa laa
the-scholars-nom usually 3-say-pl.m. what not
y-afcal-uu-na
3 -do-pl. m.-indic
In (178a) the adverb precedes the (free relative) object, and follows the
inflected verb, while it precedes the verb (and the object) in the ungram-
matical ( 178b). These judgments parallel those found in French, as in the
following contrast:
(179) a. Jean mange sou vent des pommes.
John eats often apples
b.* Jean sou vent mange des pommes.
They indicate that the verb has raised over the adverb, although it has
originated lower than the latter at D-structure.
The distribution of floating quantifiers may also be suggestive. Arabic
quantifiers are heads of the NPs they quantify, as shown in Fassi (1981).
They behave in every respect like a nominal head. Thus like nominal heads,
they receive case, bear definite or indefinite articles, head genitive
constructions, as in the following examples:
(180)a. jaa?-a 1-kull-u
came the-all-nom
Everybody came.
b. caad-a kull-un ?ilaa manzil-i-hi
came.back every-nom to house-gen-his
Everybody came back to his house.
The analysis of clauses developed so far has relied on an implicit theory of Arabic word
formation. My purpose in this Appendix is to make explicit the general model which accounts
naturally for Arabic word properties while being compatible with the theory of clause struc-
ture adopted. Verb formation is taken to be illustrative of how words are internally organized,
and how their combinatorial properties parallel (to a large extent) those of clauses. 1 Whatever
'mirroring' effects and syntactic organization there are in internal word structure, it seems
unlikely that they can be captured naturally in a strictly lexicalist (or 'redundantist') view
of word formation processes. 2 The same is true of a theory in which words are only manip-
ulated in a completely autonomous morphological component. 3
In Section 1 I discuss and reject an apparent argument for a lexicalist approach to Arabic
word formation. In Section 2 I argue for internal bracketing in Arabic words. In Section 3
I examine the question whether Arabic syntactic xo categories are lexically filled by stems
or by roots. I then provide the form of a general model of Arabic word formation in Section
4. Finally, I analyze how inflectional categories (i-categories) are articulated once we assume
an INFL split theory.
In (la), the adjective is formed from a stem [petit], linearly followed by the discrete feminine
suffix [e], followed in turn by the segment for number, [s]. Similarly, the conjugated verb
is formed from the stem [chant], representing the bare verb, the affix [r] for tense, and the
affix [ons] for person and number. Arabic, by contrast, instantiates a nonlinear or non-
concatenative morphology in a substantial number of conjugations of words. Thus the plural
of daar 'house' is duur or diyaar. The vocalic information in the singular form is com-
pletely lost in the plural, and the derivation of the plural cannot be linear. 4 The same
phenomenon can be observed in inflected verbs. In Arabic, there is no bare (stem) form of
the verb with which T and other i-morphemes can concatenate to form a conjugated verb.
When a verbal stem is formed, it is already inflected for T, Aspect, as well as Voice. The
most simple stem daxal 'entered', for example, is a conjugated verb, expressing past T and
active voice; the prefixed form y-adxul '3-enters' (with a different vocalism) expresses the
non-past imperfective; dxul 'enter' stands for the imperative mood, etc. Word formation is
highly 'non-cumulative', especially with respect to vocalism. Derivations of words from
one another (if there are any) have to be root-based, because the vocalism of the stem is
dropped.
This state of affairs seems to have motivated an essentially 'redundantist' view of Arabic
word formation. McCarthy (1979), for example, takes the lexicon to be fully specified with
all forms, including inflections. Morphological rules are conceived of as lexical redundancy
rules, by means of which the cost of predictibilty is computed. Derivational relationships
between words cannot be determined by internal proper bracketing of forms, because, as
McCarthy argues, "every property of the source except its root is ignored in the form of
the derived binyan". Thus McCarthy denies the existence of any bracketing in 'allam 'to
teach', which is derived from 'alim 'to know', or kaqqab 'to take as a liar' from kaqib 'to
lie', or kaatab 'to correspond with' from katab 'to write', or ?ajlas·'to seat' fromjalas 'to
WORD ORDER, AGREEMENT, AND CASE 77
sit', or stawjab 'to necessitate' from wajab 'be necessary', etc. His essential argument is
that the vocalic pattern has no relevance for the derivation. 5
But despite the fact that vocalic information may be lost in the process of deriving
phonological forms of words, it seems unlikely that Arabic words (including those provided
by McCarthy) have no internal bracketing. The fact that the mapping from morphological
structure to phonological form is not one-to-one does not preclude internal bracketing, as
will become clear. Moreover, there are serious descriptive and theoretical problems with
redundantism (or lexicalism), which make it undesirable.
must have internal bracketing, and they must be subject to syntactic well-formedness
principles.
A question that comes to mind then is whether bracketing is necessary in every part of
the word, or only partially so, i.e. limited to the concatenative part, whereas the non-
concatenative part (i.e. the stem) has no internal bracketing. Considerations of how Arabic
stems are formed seem to suggest that bracketing does not depend on the concatenative vs
non-concatenative nature of morphemes, as will be shown in the next subsections.
consonantal root ktb or the (inflected) stem katab? Each choice has different consequences
for phrase structure, and for word structure as well. 8
Suppose it is the stem which is projected, then it should occur under a joint projection
of V and I in the syntax. It would be incorrect to insert katab under V only, because the
latter form expresses also the perfective T/Aspect and the active voice. The same observa-
tion applies to kutib, ktub, etc. Suppose, on the other hand, that only the consonantal root
is projected, then i-categories like T and Voice would be adjoined to it in the syntax, and
the stem would be formed (depending on this information) only in the phonology. General
syntactic considerations (throughout the book) as well as conceptual ones will lead us to
generate T, Asp, and Voice in the syntax, not in the lexicon. Among these considerations
is the fact that standard X' theory does not allow multiple heads. Moreover, these heads
have different properties (compare, for example, the properties of T with those of AGR or
those of passive Voice, etc.). Finally, if these i-categories are merged with the root in the
lexicon and projected as such, then it is difficult to separate their syntactic properties from
those of the lexical category they merge with.
(6)
Lexicon:
- roots, stems, affixes
- move-alpha (head-to-head)
~
D-str-uctur-e
transformational move-head-to-head
~
8-str-uctur-e
morphology:
- readjustment
- spelling
- derivation of binyanim
~
mor-pfw(m;ttcaL for-m
phonological rules
photwttc Jor-m
Details of the morphological and phonological components do not concern us here (see
Seghrouchni, 1989 on the matter), but the general picture is quite clear.
Consider a word like yustawzaruuna 'they are appointed as ministers'. The consonantal
root [stwzr] is complex. It is formed by raising the basic root [wzr) to the causative prefix
[st] in the lexicon (see Hale and Keyser, 1989 on the application of Move-a in the lexicon).
At D-structure, only the complex causative root is projected (with no further internal
structure). Thus the causative construction (7) has a D-structure like (8):
(7) st-awzara zayd-un r-rajul-a
made-minister Zayd-nom the-man-ace
Zayd has appointed the man as minister.
WORD ORDER, AGREEMENT, AND CASE 81
(8) VP
~
NP V'
I~
zayd V NP
I
stwzr
I
r-rajul-a
The structure in (8) is presumably the representation of the verb's thematic structure in the
lexicon. But it arises only as an output of the rule forming causative verbs in the lexicon,
from an underlying structure of the form given in (9):
(9) VP
~
V'
V
~VP
I~
st V'
~
V NP
I
wzr
I
r-rajul
In this structure, the intransitive verb wzr has no thematic subject, and the affix [st] is a
causative verb. The embedded verb raises to incorporate into the higher one, and unneces-
sary nodes are dispensed with (or pruned by convention), along the lines suggested for
conflation rules by Hale and Keyser (1989). As a result, the structure given in (8) is formed,
and subsequently projected in syntax.
In the syntactic component, the verbal root is inserted in a structure which contains various
inflectional constituents. These morphemes merge together at the phonological level, but
bracketing operates in syntax, as previously described. As a consequence, an internal struc-
ture like (10) is generated:
(10) Mood
~
AGR Mood
~I
T AGR [na]
T
~I
V [uu]
~
y-ustawzar
This structure is the input to morphology, on which readjustment and spelling rules operate,
followed by purely phonological rules. The rule which affixes the causative morpheme rep-
resents the general model of affixation to a root in the lexicon. The affixation of inchoatives,
reflexives, reciprocals, etc. follows the same pattern, along the lines proposed in Fassi ( 1986b ).
Not only does the rule Move-a apply in the lexicon to merge a root with an affix, it
also applies to merge a root with another root. The phenomenon, called na/Jt 'dressing' by
traditional grammarians, is a particular form of compounding which operates solely on
roots. It merges a triliteral root with another triliteral or biliteral root, as in basmal 'to utter
the invocation: in the name of God', formed from bism 'in the name of' and llah 'God'.
Then the extra consonants are 'cut' or lost in order to keep only a quadriliteral form, which
is the maximal canonical number of root consonants. The consonant reduction here recalls
82 CHAPTER 2
what happens with plurals of words containing more than four consonants. For example,
the plural of <andaliib 'nightingale' is <anaadil, the plural of safarjal 'quince' safaarij, etc.
Thus apart from quadriliteral roots in which the last consonant is reduplicating the third
one, as in jalbab 'to clothe in a garment', most of the complex roots are formed by the
rule Move-a, which adjoins a root to another root, or to an affix, in the lexicon. The rule
applies equally in the syntax. allowing further adjunctions to T, Asp, AGR, etc. A number
of morphemes are abstract, as we have mentioned, and they may or may not be realized
separately in the morphology. For example, the so-called third form CaaCaCa, which expresses
reciprocality, is usually formed from triliteral roots, and the abstract morpheme, which is
manifested by lengthening, is only spelled out correctly in the morphology. 10
happening, when occurring with perfect verbs (equivalent to 'indeed' in English or to the
emphatic modal 'do' in 'John did go'). It also expresses probability or uncertainty with imper-
fective verbs. On the other hand, sawfa (like sa-) is used to specify future meaning, but it
has also the modal connotation of certainty or effective happening. That the two particles
are in complementary distribution is illustrated by the ungrammaticality of (14):
(14) * qad sawfa y-al:l(Jur-u
may will 3-is.present-indic
*He may will come.
These modality particles occur after C, as shown in (12), and before Neg, as in (13). Neg
follows Mod, after C, but it precedes AGR and T, as in the above sentences. 11
C, Mod, and Neg, have a different status from other items of the list. They are not normally
incorporated onto the verb, though some of them might be bound to the verb (like the prefix
sa-). This class of i-categories contrasts significantly with the class of elements which are
merged within the stem of the word, or are even necessary to make it pronounceable. The
latter class is represented by the complex stem V.Vo.Asp.T-AGR-Mood (Vo = Voice, and
Asp = Aspect; I use dots to separate categories which form the stem, and dashes between
the different concrete segments of the word). Thus Vo, Asp and T are merged together in
one morpheme. AGR is another morpheme, as is Mood. It should be born in mind, however,
that a single morphological unit does not necessarily correspond to a single syntactic category.
T, Asp, and Vo might represent more than one syntactic category, as we will see in other
chapters, but all these categories are nevertheless merged into one morphological unit.
The distinction between Mod and T is based both on morphological and syntactic evidence.
Tense markers are either vowels merged in the stem of the verb, or prefixal or suffixal con-
sonants which are strictly internal to the word. Modals, in contrast, are usually non-bound
morphemes, and they precede (and govern) verbs. An exception to this appears to be
sa-, but even in this case there is no context of true 'fusion' in the sense of Bybee (1985),
and the modal has to precede the prefix tense marker on the verb. For example, the internal
(vocalic) morphology of passives extends to the prefix tense marker, but not the modal, as
shown in (15):
(15) sa-y-u9rab-u
will-3 -beat. pass. -indic
He will be beaten.
If the sa- was a Tense marker, then as far as I can tell, nothing will stop the passive vocalic
morphology from 'spreading' to it. The result will then be the ill-formed *su-y-u{irab-u. There
are other differences between Tense markers and modals which keep them distinct. For
example, T is a Case assigner of Nominative, whereas Mod may assign Mood or be neutral
with respect to this process (see Chapter 4).
We can adopt a maximalist general strategy of decomposing the traditional constituent I
into different categories, and assume that Asp and Mood, like T and Mod, head their own
inflectional projections.
Asp, where grammaticalized in verbal morphology (as in Greek and Slavic perfective or
imperfective) can be easily distinguished from T. But Asp is not encoded by a specific
morphology on the Arabic verb, at least if the perfect/imperfect distinction is not seen as
the basic interpretation of this morphology (but see the controversy on this problem in Chapter
4; see also the references cited there). Likewise, Mood can be separated from Modality,
although this distinction is more difficult to support than the former one. In Arabic, the
distinction between the two is grammatically clear, despite the fact that semantically and
conceptually, there might be some confusion and overlapping. Mood is an inflectional element
which is suffixed to the verb, and merges with its stem, whereas Mod is usually expressed
by particles, which are not part of the verb. The perfect form of the verb is almost exclu-
84 CHAPTER 2
sively used for the indicative Mood (presumably by default), while the imperfect form has
different endings, depending on whether it expresses the jussive as in (16), the subjunctive,
as in (17), the conditional as in (18), or the energetic, as in (19):
(16) li-t-axruj
let-you-go
You go!
(17) ? -uriid-u ? an t-ar tiy-a
1-want-indic that you-come-sub}
I want you to come.
(18) (in t-adxul 1-qaa'ata
if you-enter the room
(19) la-t-adxul-u-nna
you-enter-2.pl. -indeed
You will enter indeed.
The attachment of the Mood suffix precedes the attachment of pronominal affixes on the verb.
Modal morphemes, by contrast, do not have the same nature, nor the same distribution.
They are usually non-bound particles or verbs. Their semantics includes 'probability', 'pos-
sibility', 'necessity', etc. Here are some examples:
(20) qad ? -a'uud-u rilaa faas
probably 1-retum-indic to Fez
I will probably return to Fez.
(21) sawfa ?-a'uud-u ?ilaa faas
will 1-retum-indic to Fez
I will return to Fez.
(22) la'alla-hu gaa? ib-un
maybe-him absent-nom
He may be absent.
(23) yajibu (an t-aquul-a haagaa
is.necessary that you-say-sub} this
It is necessary that you say this.
It is true that there is an overlap in meanings expressed by Mod and Mood, which may suggest
that the two inflections realize only one semantic category, but it is not that obvious. The
question merits further research. 12
Given the order discussed, then if these constituents are X0 heads of their syntactic
projections. their order might be a reflection of either deep or surface constituent relations
within the tree. Since T is apophonic, i.e. it forms (part of) the internal vocalic structure of
the stem, prior to any possible prefixation or suffixation, it is natural to think of it as the
lowest element in the tree adjoined to the root. On the other hand, AGR is higher in the
tree structure than T, since it is suffixal or prefixal, and does not belong to the stem. As
for Mood, it follows AGR. Taking these observations into account, I propose that the sentence
in (20) has the following D-structure:
WORD ORDER, AGREEMENT, AND CASE 85
(24) ModP
Mod
~MoodP
I~
qad Mood AGRP
I~
[u] AGR TP
~
T VP
~
Spec V' V'
I
Pro
I
V
I
?a
I
cuud
As mentioned earlier, Move-a applies at S-structure, raising V toT, then V-T to AGR,
then V-T-AGR to Mood, thus deriving the right internal order of morphemes within the word,
before phonological rules apply. 13
NOTES TO APPENDIX l
1 Baker (1985a, 1988) defends a view of morphology based on the Mirror Principle.
2 The redundancy view has originally been developed by Jackendoff (1975), on the bases
of observations made by Chomsky (1970) and Halle (1973). Redundancy rules are descrip-
tive in nature. They relate lexical entries in the following manner:
(i) [x ... ] H [y ... ]
where x and y are variables for integers numbering the entries, and the dots stand for the
phonological, syntactic, and semantic information. There are almost no limits or constraints
on these rules. Redundantists aknowledge the fact that there are syntactic expressions listed
in the lexicon (namely idioms) which are not words, and there are 'possible' words in the
language which are not listed in the lexicon (namely neologisms). But this paradoxical
situation has not managed to convince them to give up their view, as Di Sciullo and Williams
( 1987) observe.
3 For a recent defense of an autonomous morphological component, see Di Sciullo and
Williams (1987).
4 These observations apply to so-called broken plurals. Sound plural formation is essentially
concatenative (see McCarthy, 1981).
5 According to McCarthy (1979, p. 388), a lexical entry of a form w-referred to as L(w)-
is a directed graph whose root is w. "That is, a lexical entry is a rooted, n-ary branching
tree. For any b which is dominated by a in L(w), we say that b is derived from a. If a and
b are both dominated by some w in L(w), then we say that a is morphologically related to
b". The evaluation metric is conceived as follows: "any relationship of immediate domina-
tion in a lexical entry that can be predicted by any morphological rule is without cost. [The
optimal morphological system] ... will have only the value of the sum of the values of all
root nodes of lexical entries plus the sum of the values of all morphological rules." (See
p. 390.)
6 On strong lexicalism, see Lapointe (1981). A recent and elaborated version of the thesis
of atomicity of words, incompatible with syntacticism (though not lexicalism) can be found
in Di Sciullo and Williams (1987).
7 Di Sciullo and Williams (1987) argue that morphological rules are recursive, quoting Halle
86 CHAPTER 2
(1973) for the series of words anti-missile missile, anti-anti-missile missile missile, and so
on " ... an infinite series of words, each with a determinate meaning different from that of
all the others" (p. 7). In their view, the lexicon is limited to the set of listemes, that is, the
list of memorized objects. Furthermore, listedness is no more intrisincally characteristic of
words than it is of phrases. On the other hand, the authors correctly reject the view that the
rules of morphology are essentially redundancy rules over a set of objects (i.e., words).
"To us, this makes no more sense than to say that VP ~ V NP is a redundancy rule over
the sets of VPs, most but of course not all of which have zero redundancy." (Seep. 4.) On
the other hand, productivity has been used as a criterion for separating what is in the syntax
(or the grammar) from what is in the lexicon (see Chomsky, 1970; Selkirk, 1982; and Fabb,
1984, among others). Di Sciullo and Williams correctly observe that productivity does not
determine the representational level. It merely determines the contextual restrictions for words.
8 It might be that the introduction of abstract citation forms of stems (like CaCaC) is
equivalent to the introduction of roots. The point is neutral with regard to the two options.
On the other hand, the question arises whether these consonantal roots are categorially
specified in the lexicon, or they inherit their categorial specification in the syntax (assuming
that categorially unspecified words project a default XP schema, which has to be licensed
by D-structure principles). In order to answer this question properly further research is needed
to determine how notional 'types' in the sense of Hale and Keyser (1991 ), such as events,
entities, and properties are associated with categories such as V, N, and A. If the associa-
tion occurs universally in the lexicon, then the roots will be categorially specified. If the
association is a matter of variation (occurring in the syntax or the lexicon, depending on
languages), then no categorial specification is necessary. For ease of exposition, I will assume
the specification view. Thanks are due to an NLL T reviewer for helping me clarify this
point.
" An alternative to bracketing is to claim that i-elements are features (which are realized
on roots), following a proposal made by Anderson (1988). Spelling rules would then have
to take care of ordering the different i-elements. This option appears to be stipulative. however.
For criticism, see Halle (1989).
10 I think that nominals like kitaab 'book', ?asad 'lion', rajul 'man', etc. are listed as
such in the lexicon, and projected as stems, at least when they are ordinary nouns. Roots
of these forms, however, do exit in the lexicon to form broken plurals, causative verbs, etc.
No derivation of these words from stems is possible. Plurals of these nouns (kutub 'books',
?usud 'lions', rijaa/ 'men') do not contain any of the vowels of the singular noun. The
same is true of verbs which have the same roots (?aktab 'to make write', ista?sad 'to
become a lion', raj}at 'to change to a man', etc).
11 It is worth pointing out, however, that medals and Neg particles are in complementary
distribution when the latter are Mood assigners, as the ungrammaticality of the following indi-
cates:
(i) *?sawfa ian y-ai)qur-a
will Neg.fut. 3-is.present-subj
He will not be present.
See Chapter 4 for an analysis of this complementarity.
12 For some informal suggestions, see Lyons (1977) and Palmer ( 1985), among others;
see also Chapter 4.
13 An SNLL T reviewer has objected to the decision for generating T lower than AGR on
a number of grounds. He pointed out, for example, that if ?an in constructions like (23) is
treated as a Mood marker or a T element equivalent to English infinitival marker to (rather
than C), then Mood or T would be higher than Neg, and Neg would be higher than AGR
carried by the verb. That Neg is higher than AGR seems also to be supported by the fact
that in Neg clauses, Tense is realized on Neg particles, while agreement is realized on the
WORD ORDER, AGREEMENT, AND CASE 87
So-called nominal sentences seem, at first glance, to resist an analysis in which they are
assigned a fully expanded inflectional structure. Unlike sentences headed by a verb to which
T and AGR are attached, this type of sentences contain no (visible) verb and hence, no (visible)
T and AGR. But there is evidence that these sentences are headed by essentially the same
i-components which head verbal sentences, including T and AGR, contrary to what a small
clause analysis of these constructions would predict. 1
1. Tense
In Fassi (1981 ), it is proposed that the D-structure of a nominal sentence like (I) is iden-
tical to that of a copular sentence like (2):
(I) r-rajul-u marii9-un
the-man-nom sick-nom
The man is sick.
(2) kaana r-rajul-u marii9-an
was the-man-nom sick-ace
The man was sick.
The copular verb, which is present at D-structure in both cases, fails to surface when it
bears an unmarked T feature, namely [-past]. By contrast, the [+past] T forces the copula
to be present at S-structure. One claimed advantage of this analysis is that it unifies the under-
lying structure of superficially verbal and non-verbal sentences. In Fassi (1987a), I have
adopted an analysis in which I heads nominal sentences, though no copular verb is present
at D-structure. The essential motivation of both analyses is that nominal sentences do contain
T.l
There is empirical evidence supporting this view. Temporal adverbs locate the event in
time, and depending on their temporal specification, they may or may not be compatible
with the tense morphology encoded on the verb. The contrast in grammaticality between
(3a) and (3b) is due to this compatibility requirement:
(3) a. yaakulu r-rajul-u 1-?aan-a
eats the-man-nom the-now-ace
The man eats now.
b. * yaakulu r-rajul-u ?amsi
eats the-man-nom yesterday
88 CHAPTER 2
Similar contrasts obtain in nominal sentences, pointing clearly to the fact that their inflec-
tional structure contains a non-past T:
(3) a. r-rajul-u mariic;l-un 1-? aana
the-man-nom sick-nom now
The man is sick now.
2. AGR
Verbless sentences like (1) seem at first glance to contain no sentential (or tensed) AGR.
The only visible agreement there is phrasal AGR, which shows up on the adjective. There
is evidence, however, that these sentences do contain sentential AGR. Thus, although
sentential AGR is not usually realized in nominal sentences, it does show up in some negative
contexts, as will be explained. Consider the following sentence:
(5) lays-at hind-un mariic;lat-an
Neg-.s.f Hind-nom sick-s.f-acc
Hind is not sick.
There is reason to think that the position of the Neg morpheme laysa in (5) is not its D-
structure position. Like a copular verb (and unlike other Neg morphemes), laysa assigns a
(morphological) accusative to the adjective. This case is presumably assigned (under
adjacency) by the negative. The latter is generated lower than AGR and T. Then laysa head
moves first to T. then to AGR. The D-structure for (5) can be diagrammed as follows:
(6) AGRP
~
AGR TP
~
Spec T T'
~
T ~
Neg AP
I~
laysa NP A
I
hind
I
marii~at
It is interesting to note that the form of AGR that /aysa supports is identical to the form which
appears on verbs. This is why traditional grammarians treat laysa as a verb. The occur-
rence of the right form of agreement on lays a can be ensured if we assume that the structure
of the sentence contains a sentential or a tensed AGR (which is different from phrasal AGR
that occurs on adjectives). Then laysa incorporates T and AGR, ensuring that the right form
of agreement is attached to it.
WORD ORDER, AGREEMENT, AND CASE 89
NOTES TO APPENDIX 2
1 Mouchaweh (1986) has adopted a small clause analysis for Arabic nominal sentences,
based on a proposal in the same vein made by Kayne (1983). See also Rapaport (1987) for
a similar view applied to Hebrew.
2 Further research is needed to determine under what (morphological) conditions the copula
is visible, keeping in mind the initial idea of Fassi (1981) that the copula is 'there', but it
may or may not surface. In languages like English and French, the verbal copula is realized
whatever its tense and aspect. This is not so in Arabic, where it is not realized with unmarked
tense and aspect. On the other hand, a predicate nominal or an adjective receives morpho-
logical accusative Case only when the copula surfaces, indicating that lexicalization of the
copula is a necessary requirement for Case assignment (see Chapter 4 for a more extensive
discussion).
NOTES
(i) s
~
I NP VP
~
V NP (Sproat, 1985a)
90 CHAPTER 2
(ii) s
~
I S
~
NP VP
~
V NP (McCloskey, 1990)
(iii) s
~
I VP
~V'
NP (Rouveret, 1990)
7 On the formal properties of head movement rules, including the Head Movement
Constraint, see in particular Travis (1984), Chomsky (1986a), and Baker (1988).
8 Alternatively, we might adopt a logicosyntactic motivation for V raising a Ia Pollock
(1989a), in which V raises to I, to enable the latter to bind the E variable in V. See note
61.
9 For additional arguments, see in particular Fassi (1981) and ( 1987), and Bakir ( 1980).
Abdo (1983) argues that SVO is the neutral order.
10 This fact can be derived, in turn, if the subject is taken to have moved to an A-position
(through specifiers), since movement to an A-bar position does not preserve basic order.
As will be shown, the subject there has to move to Spec of AGR (and then eventually to
Spec of C), as in Rizzi (1991a).
11 Joan Maling pointed out to me that the exact counterpart to (21) is perfectly grammat-
ical in languages with possessive reflexives (e.g. in Scandinavian). A crosslinguistic
comparison of bound anaphora constraints and parameters is beyond the scope of this work.
12 The expression 'neutral coindexation' is used in the sense of Reinhart (1983).
13 Barss and Lasnik (1986), among others, have proposed that precedence plays a role in
the grammar of binding. See also Evans (1980) for pronominal binding. Reinhart (1983)
suggests that precedence belongs to the theory of 'pragmatics', rather than to the grammar
proper.
14 Among the arguments usually provided to establish subject/object asymmetries is idiomatic
interpretation. Objects are claimed to form idioms with V 'easier' than subjects do. See
Chomsky (1981) and Marantz (1984) for the form of the argument, and Bresnan's (1982)
criticism of it. See also Mouchaweh (1986) on similar argumentation for Arabic. Note that
Chung ( 1983) and McCloskey (1984) have claimed that there is no subject/object
asymmetry in terms of government in Chamorro and Irish, respectively.
15 Note that additional problems arise in the context of precedence with regard to Crossover.
Arabic observes both Weak and Strong Crossover constraints on pronominal binding, as shown
by the ungrammaticality of (i) and (ii) on one hand, and that of (iii) on the other hand:
(i) *man ?al).abb-at zawjat-u-hu
who loved-! wife-nom-him
*Who did his wife love?
(ii) * haa<jaa huwa r-rajul-u lla<.!ii ntaqad-at zawjat-u-hu
this he the-man-nom who criticized-f. wife-nom-him
*This is the man who his wife criticized.
(iii) * maktab-a zayd-in daxa1-a
office-ace Zayd-gen entered-3.s.m.
* Zayd' s office, he entered.
WORD ORDER, AGREEMENT, AND CASE 91
All these examples are ruled out under the coreference reading. Assuming that Weak Crossover
effects are observed only when scrambling is to an A-bar position, as proposed by Mahajan
(1990), and that the same is true of Strong Crossover in (iii), this suggests that the condi-
tion (18) is not sufficient to regulate the distribution of A-bar pronouns.
16 This is conditioned by the non-existence of an overt complementizer in C, so that the
basic motivation might be viewed as being the complementarity of V2 and overt Cs. This
analysis is originally due to Hans den Besten (1977), later published as (1983). For an
elaborated version of V movement in Germanic, see Platzack and Holmberg (1990).
17 The V -to-C-analysis incorrectly establishes a similarity between the inflectional struc-
ture of Arabic and that of verb second Germanic languages, which exhibit a number of
different and intriguing properties. For a critical review of these properties, see Weerman
(1989). See also note 16 and references cited there.
18 The Arabic dominant grammatical tradition of Basrah (Iraq) has denied the subject
interpretation. According to this tradition, a preverbal NP is always interpreted as a 'topic',
i.e. a constituent in an A-bar position (see e.g. Sibawayhi, 8th century). Previous genera-
tive work has adopted (incorrectly) the Basrian's view (see e.g. Bakir, 1980; Fassi, 1981;
and Ayoub, 1982 among others). The Kuufa's tradition admits the existence of preverbal
subjects, but without convincing arguments (see e.g. Ibn Madaa?). It is then an important
contribution to Arabic linguistics to show that this view, which was dominant for centuries,
is inaccurate. Furthermore, it is equally of great interest to theoretical linguistics to assess
the existence of mixed type languages, as described here. Note also that Arabic is crucially
not analyzed as an SYO language (as in Abdo, 1983 or Mohammad, 1990 among others),
but essentially as VSO.
19 Non-Arabists might wonder why 'jund-a' is Accusative here, although it is a subject.
In fact, the so-called generic negation marker 'laa' (by traditional grammarians) is assigning
morphological Accusative to the noun in this case. That assignment takes place inside the
subject NP. On Case assigning properties of negative markers, see Chapter 4.
°
2 Chafe (1975) claims also that topics must be definite, and he attempts an explanation
of this requirement from a funtional perspective.
21 See Cinque (1984), who proposes a similar condition.
22 Howard Lasnik (personal communication) informed me that quite the same judgments are
valid in English. On the other hand, the requirement that raised subjects must (at least) be
weakly referential has to follow from the referential properties of AGR. It is rich AGR
which seems to force (weak) referentiality, as the facts suggest. On the correlation between
specificity, AGR, and Case, see En~ (1990), Mahajan (1990), and Moltman (1990), among
others. On referential properties of AGR, see Gueron (1989). See also Chapter 3.
23 As will be shown in Chapter 3, PERS is not necessarily specified in Spec-Head agree-
ment, but NUM is. Moreover, licensing conditions for this feature might be different (see
Rouveret, 1991 and Chapter 4). Matters of specification are more complex than the taxonomy
rich/poor might suggest. In affirmative imperatives, for example, the suffix marker on the
verb encodes GEN and NUM, as illustrated by (i):
(i) dxul-na 1-l)arb-a
enter-pl.f the-war-ace
Enter (you women) the war 1
When the imperative is negated, the prefix encoding PERS must be introduced:
(ii) laa t-adxul-na 1-l)arb-a
not 2-enter-p/.f the-war-ace
Do not enter (you women) the war!
This seems to suggest that prefixes on imperfective verbs carry PERS feature, and that poor
AGR (at least with imperfectives) carries PERS as well as GEN. If this is true, this would
92 CHAPTER 2
mean that Spec-Head agreement triggers only agreement in NUM. The facts are complex,
however, and will be elaborated in Chapters 3 and 4.
24 It is worth observing that Accusative 'spreads' in (Sib), in which there is a govern-
ment by a verb. but not in the counterpart of (50). when governed by C, as in (i):
(i) ? inna r-rajul-a mariic.J-un
that the-man-ace sick-nom
The man is sick.
In this case, only the subject is accusative. but the predicate adjective must be nominative.
The ungrammaticality of (ii) indicates that case spreading is not possible in this case:
(ii) * i' inna r-rajul-a mariic.J-an
that the-man-ace sick-ace
The man is sick.
In fact, all four logically possible combinations (of nominative and accusative) are avail-
able in Arabic data. See Fassi ( 1987a) for analysis of these distributions. On case spreading,
as well as an analysis of case predicate APs agreement, see Yip, Jackendoff, and Maling
(1987). Thanks are due to Joan Maling for having helped me clarify these data.
25 Compare in particular with Rizzi's (l99lb) Wh Criterion.
10 As will be explained, however, French exhibits only two patterns of agreement: null
(or unspecified) and rich (or specified). It has no poor AGR, which is specified, but poor.
See below.
n For similar data in Breton. see Anderson ( 1982). and in Irish, see McCloskey and Hale
(1984). and Hale (1987a).
28 Ferguson ( 1984) points out that some Arabic SVO dialects exibit instances of null or poor
agreement when the subject is postverbal. The following examples. taken from Ferguson
and Barlow (1988), belong to Syrian Arabic:
(i) i'ijaa-ni makatiib
came.3.m.s.-me letters
I got some letters.
(ii) i'ijit-ni makatiib
came.3.fs.-me letters
I got some letters.
In (i). there is null agreement with the thematic subject, and in (ii). agreement in GEN only.
On the other hand, a SNLLT reviewer has observed that what I have termed rich agree-
ment is also found in VS contexts in languages usually classified as VSO, such as Chamorro
(see e.g. Chung. 1990), or Berber (as in Ouhalla, 1988). These facts (if correct) have to
wait for more exhaustive and detailed studies of these differences.
:;q Despite the dominant literature on the matter. Mohammad ( 1990) assumes, without
argument, that Arabic is SVO. not VSO. Moreover. he establishes no correlation in his
work between the agreement patterns found in VS and expletive contexts. For example, he
overlooks the fact that expletives can be feminine singular and that this is also true of AGR
in VS constructions, as pointed out in Fassi (1988b). On the contrary, he treats the crucial
GEN feature as being something he can do without, when he states: " ... I have nothing to
say about gender assignment in VS orders; so I will ignore it throughout the paper and
concentrate on the number feature" (p. 98). But although he promises to deal with the
number feature, he does not deal with the problem of why a VS sentence with a plural number
agreement is ungrammatical in Standard Arabic, while it is grammatical in e.g. MA or Standard
Italian.
WORD ORDER, AGREEMENT, AND CASE 93
30 Basically, Mora claims that There raises from a small clause in which it functions as a
predicate. Sentences like (79) and (80) would then be 'inverse' copular sentences in Mora's
sense, and not 'canonical' ones. See his paper for details.
31 Note that it is possible to say:
(i) hiya 1-junuud-u
she the-soldiers-nom
It's soldiers.
This is due to the specific rule which treats broken plurals as feminine singular (see Fassi,
1984). For more data instantiating the same phenomenon, see Roman (1990).
32 A similar conclusion can be equally reached for French, although through a more complex
route. If we follow Kayne (1989) and Rizzi and Roberts (1989) in postulating the existence
of subject clitic expletives in complex subject inversion in French, then it cannot be true
that the expletive list is limited to singular masculine forms. The fact, then, that the exple-
tive is limited to the singular masculine form in the right context has to follow from the
fact that AGR there cannot be specified, otherwise its features would not be licensed.
33 For these data, see Wright (1974), Blachere and Demombynes (1952), Fassi (1984),
and Roman (1990), among others.
34 Constructions like (82), (89) and (90) indicate that expletives can be licensed as topics,
and not necessarily as subjects. Expletive topics have also been documented in Icelandic
(Zaenen, 1980), and Germanic (see Platzack, 1983 and Maling, 1988, among others).
35 See Pollock (1983) for some suggestions to account for French data.
36 Null agreement is used in MA only in contexts where there is no thematic subject, as
in (i):
(i) <.Jhar li belli layla ja-t
appeared me that Lay/a came-f
It appeared to me that Layla has come.
In this example, the matrix carries no specified form of agreement.
37 R-chains establish referential dependence between members of the chain, of which one
member is an R-NP, and another member is an expletive. R-chains are counterparts of CHAINs
in Chomsky's (1986a) system, although they do not seem to obey Chomsky's General
Condition on chains (formulated in (138) below)
38 This assumption is based on earlier work by Rizzi ( 1982).
39 I have assumed throughout that there is no counterpart of the AGR Criterion applying
to non-specified AGR, or expletives (with no specified features). See the observations made
in subsection 4.1.
40 It might be suggested that the extraposed subject inherits Case by transmission from an
empty expletive located in Spec of I. This claim cannot be supported, however. When an
expletive surfaces there, the only available Case for the extraposed subject is Nominative,
even though the expletive is marked with accusative, as the following example shows:
(i) ? inna-hu fii d-daar-i rajul-un (*rajul-an)
that-it in the-house-gen man-nom (*man-ace)
There is a man in the house (literally: there is in the house a man).
In general, the Case transmission hypothesis seems to have no motivation in the grammar
of the language. On the contrary, it makes the wrong predictions (see note 41).
41 Note that if we were to hypothesize an empty expletive in Spec of I, together with Case
transmission, nothing would stop the ungrammatical construction. As with C, whenever an
expletive surfaces there, the subject must be marked with Nominative case:
94 CHAPTER 2
55 Note that this alternation was at the heart of Emonds' proposal (1980) to derive VSO
from an underlying SVO, via V fronting.
56 According to Chung (1990a), Chamorro's alternative order is VOS. As we have seen,
Arabic exhibits also VOS order.
57 This is in line with Chomsky's (1986b) theory of adjunction. Chomsky assumes that
adjunction to a maximal projection is possible only if the latter is non-argumental. Adjunction
to VP is allowed, but not to NP or CP (when they are argumental). Keeping the essential
insight, I propose that adjunction is limited to [-N] categories.
58 Recall that Chomsky (1986b) defines L-marking as follows:
(i) Where a is a lexical category, a L-marks b iff b agrees with the head of y
that is theta-governed by a.
As for theta-government, it is defined as follows:
(ii) a theta-governs b iff a is a zero-level category that theta-marks b. and a, b
are sisters.
59 The non-agreeing C in (157b) will prevent the trace of the Wh word in Spec of CP
from acting as antecedent governor for the trace in Spec of AGRP. In contrast, antecedent
government is not blocked in (157a). See Rizzi (1990) for proposals in the same spirit, and
Kenstowicz (1989) for discussion of similar facts in some Arabic dialects.
60 Joan Maling pointed out that it is not strictly true that lowering rules violate the Economy
principle. Because of ECP, they lead to derivations which do so.
61 Pollock ( 1989a) has provided another syntactic reason for V raising. Basically, he takes
T to be an operator (located in I in some languages) which has to bind a syntactic variable.
V to I is treated as an A-bar movement which leaves such a variable. Pollock then proposes
to implement this idea within Higginbotham's (1985) theory of saturation, making use in
particuclar of the Davidsonian E position in the theta-grid of verbs (which functions as a
variable).
62 For arguments that Neg may be generated as a head selecting an IP, see Laka (1990)
and Zanuttini (1989).
One might wonder why there are so many negative forms in Arabic, as illustrated in the
various examples, and what are their distributional properties. Answers to these questions
are proposed in Chapter 4.
63 See Sportiche (1988) for some recent proposal along these lines for French and English,
based on the seminal work of Kayne (1975).
64 These observations indicate that Arabic Qs should be treated like normal N heads.
65 In SV sentences, like in nominal ones, I assume that Nominative is licensed only by
default.
CHAPTER 3
96
PRONOUNS, INCORPORATION, AND FEATURE SPECIFICATION 97
cases. In other cases, there are alternative uses of the same forms as
pronouns or inflections. Syntactic context becomes then important for
solving the problem of specification, as we will see. 3
the pronoun must be affixed to its governor e.g. in (4), but not in (5), etc.
Given general principles banning unsupported bound forms, in particular
Lasnik's (1981) Filter, the descriptive content of this statement amounts
to claiming that bound pronouns (contrary to appearances) instantiate the
regular case, whereas independent forms are either regular or 'last resort'
instances. This view is corroborated by a number of further observations,
as we will see.
An examination of the lists of non-nominative forms provides straight-
forward support for this view. It is striking that independent non-nominative
forms consist of bound forms plus the ?iyyaa support. The latter, which
has no (synchronic) meaning, combines with the bound pronoun just in case
the pronoun finds no governor to support it. Compare the forms given in
Tables 1 and 2 here:
I -11 -naa
2 Masc -ka -kumaa -kum(uu)
Fern -ki -kumaa -kunna
3 Masc -hu -humaa -hunna
Fern -haa -humaa -hunna
?iyyaa-ya ?iyyaa-naa
2 Masc ?iyyaa-ka ?iyyaa-kumaa ?iyyaa-kum(uu)
Fern ?iyyaa-ki ?iyyaa-kumaa ?iyyaa-kunna
3 Masc ?iyyaa-hu ?iyyaa-humaa ?iyyaa-hum(uu)
Fern ?iyyaa-haa ?iyyaa-humaa ?iyyaa-hunna
b. ?actay-ta-nii-hi
gave-you-me-it
You gave it to me.
Pronominal affixes can also attach to prepositions:
( 12) iltaqay-tu bi-hi
met-/ with-him
I met him.
They can incorporate into a noun in a genitive complement construction,
as in the following example:
(13) intaqad-tu mu?allif-a-hu
criticized-! author-ace-his
I criticized its author.
(like those listed in Tables 2 and 3), but also bound forms (those of Table
1). In the next subsection, I turn to the question of how structures containing
the latter are derived. I will argue that incorporation is the appropriate
mechanism involved in this derivation.
1.2. INCA
If the list of pronouns includes the tables given, then all forms contained
there can be hypothesized to originate as pronouns in argument positions.
When they are bound, they must move, in order to be morphologically
supported. Following Hale (1988), I assume that pronouns originate as
bareD (determiner) heads of noun phrases, taken to be DPs (as in Abney,
1987). 6 Bound pronouns must then head move to lexical governors, to which
they adjoin (see Travis, 1984 and Chomsky, 1986a on the formal proper-
ties of head-movement rules, as well as the illuminating work by Baker,
1988 and Hale, 1988 on incorporation processes). To take a simple example,
consider the incorporation of the pronoun to the preposition in (12). The
D-structure of the PP is (!Sa), and the output of pronominal incorpora-
tion is (15b):
(15) a. pp b. PP
~ ~
P DP P DP
I I ~I
bi D pD I
I
-hi
I
bi
I
-hi
e
I
'
(The dash points to the bound morphological nature of the pronoun.) D
incorporates into P, leaving a trace there.
INCA straightforwardly accounts for two essential facts in Arabic. First,
it explains why the occurrence of a syntactic NP argument is incompat-
ible with that of a bound pronominal on the governor. This incompatibility
is instantiated by the ungrammaticality of ( 16): 7
(16) *<;iarabtu-hu 1-walad-a
beat-1-him the-boy-ace
*I beat him the boy.
This construction cannot be generated if the bound pronoun here were to
originate as a head of a DP. This would amount to the generation of two
DPs in the same argument position. Consequently, the construction (16)
would have no viable source.
Second, INCA describes in a straightforward manner the distribution
of (non-nominative) pronominal affixes on prepositions, nouns, verbs, etc.,
as in ( 1)-(9) above. A natural assumption is that the bound forms are
PRONOUNS, INCORPORATION, AND FEATURE SPECIFICATION 103
pronouns in these contexts. The alternative- that is, taking them as agree-
ment markers - has no motivation, whether theoretical or empirical, as
far as I can see. From a 'concrete' perspective, there is no reason to think
that Arabic instantiates morphological agreement with nouns, prepositions,
or even verbal objects. 8
Third, the importance of INCA also stems from its predictions with
respect to the forms of pronominals. Recall that according to INCA, a
pronoun is predicted to be bound to a governor, unless other requirements
(and/or principles) ban the incorporation process. Among the latter are
the Head Movement Constraint and the Minimality Condition. I will
show that depending on contexts, INCA can be either (a) obligatory or
(b) impossible. I will also provide cases of optional incorporation. These
alternations provide further support for the analysis of forms as pronouns.
Consider contexts of obligatory incorporation. As mentioned above,
canonical government is a necessary requirement for INCA. The obliga-
tory nature of INCA is illustrated by the following pair:
(17) a.* ra?ay-tu ?iyyaa-ka
saw-/ you-ace
I saw you.
b. ra?ay-tu-ka
saw-1-you
I saw you.
In these constructions, the verb canonically governs the object. The use
of the free form of the pronoun results in ungrammaticality, as the ill-
formedness of (17a) indicates.
However, although canonical government is necessary, it is not sufficient.
Thus there are situations in which canonical government is satisfied, but
incorporation is impossible. Consider the following examples:
(18) a. maa ra?ay-tu ?illaa ?iyyaa-ka
not saw-/ except you-ace
I did not see but you.
b.* maa ra?ay-tu-ka ?illaa
not saw-1-you except
In example (a) there is an exclusive particle which prevents the pronoun
from incorporating into the verb (although governed in the right direc-
tion). Assuming this particle to be a governing head, then Minimality can
account for the ill-formedness of example (b).
Another context in which incorporation is prohibited is conjunct struc-
tures. When a pronoun is conjoined to another NP, its incorporation leads
to ungrammaticality, as the following contrast shows:
104 CHAPTER 3
In these examples, the PERS of the first affix is lower in the hierarchy
than that of the second one, which leads to a violation of the Person
Constraint. In contrast, the following examples do not violate the latter,
and the results are grammatical:
(22) a. ?actay-tu-ka-hu
gave-l-you-him
I gave you him.
PRONOUNS, INCORPORATION, AND FEATURE SPECIFICATION 105
b. ?actaa-nii-ka
gave-me-you
He gave me you.
c. ?actay-ta-nii-hi
gave-you-me-him
You gave me him.
When affixation does not observe the Person Constraint, the recourse to
the independent strategy becomes necessary. The grammatical counter-
parts of (21) are the following: 10
(23) a. ?actaa-hu 1- rustaag-u riyyaa-ka
gave-him the-teacher-nom you-ac
The teacher gave him you.
The teacher gave you to him.
r r
b. ?actaa-ka iyyaa-ya 1- ustaag-u
gave-you me the-teacher-nom
The teacher gave me to you.
c. ?actaa-hu ?iyyaa-ya 1- ?ustaag-u
gave-him me the-teacher-nom
The teacher gave me to him.
The Person Constraint not only applies to object affixes, as in the above
examples, but also to non-nominative subjects of deverba1 nouns (so-called
masdar's by the Arabic tradition) as in (24):
(24) * <,iarb-u-hu-ka ?aqlaq-a-nii
beating-nom-him-you annoyed-me
His beating you annoyed me.
The grammatical counterpart of the latter must also use the independent
pronoun strategy: 11
(25) <,larb-u-hu ? iyyaa-ka ?aqlaq-a-nii
beating-nom-him you-ace annoyed-me
His beating you annoyed me.
In addition to contexts where the independent strategy is obligatory, there
are contexts in which pronominal affixation is optional, but not prefer-
able. One case is instantiated by affixation of double objects. Both
independent and bound forms are allowed, as in (26), although Modern
Standard Arabic (unlike Classical) tends to use the free form for the second
non-nominative pronoun more often:
106 CHAPTER 3
(26) a. ?ac~ay-tu-ka-hu
gave-l-you-him
I gave you him.
b. ?actay-tu-ka ?iyyaa-hu
gave-l-you him
I gave you him.
Likewise, the second non-nominative pronoun used with transitive masdar's
is usually expressed as a free form, as in (27b), although (27a) is also
grammatical:
1 -tu -naa
2 Masc -ta -tumaa -tum(uu)
Fern -ti -tumaa -tunna
3 Masc -a -aa -uu
Fern -at -ataa -na
?- n-
2 Masc t- t-aa t-uu
Fern t-ii t-aa t-na
3 Masc y- y-aa y-uu
Fern t- t-aa t-na
Recall that non-nominative forms are invariably suffixes, and they are
neither sensitive to Aspect/Tense distinctions, nor to the category of the
governor.
108 CHAPTER 3
1.4. INFLA
b. ha mangiato (Giovanni)
(He) ate (Giovanni ate).
A similar situation obtains in MA, were the bound form encoding NUM,
PERS, and GEN features can co-occur with the syntactic subject, as in (35a),
suggesting that the subject in (35b) is an empty pro:
b. ja-w
came-3.pl.
They came.
The rationale behind INFLA is that the treatment of the inflection has to
be uniform. It is assumed that the difference between (35a) and (35b) is
not in terms of the nature of the inflection (in both cases it is identified
as an agreement marker), but rather in terms of the phonetic content of
the argument. In (35a), the argument is phonetically realized, but in (35b)
it is empty (pro).
Chomsky (1982) characterizes the Pro drop parameter in terms of Case.
More specifically, AGR is postulated to have Case in pro drop languages
(like Italian), and to lack Case in non pro drop languages (like English).
Furthermore, a strict feature matching requirement between AGR and pro
must be met, banning pro in non pro drop languages, where AGR is claimed
to have no Case and hence does not match pro features (in terms of Case). 16
Rizzi's (1986) theory of pro (based on Rizzi, 1982) treats licensing and
recoverability conditions separately. It consists essentially of the licensing
principle (36), with a built-in parametrization of licensing heads, and the
convention (37) for the recovery of the content of pro:
(36) Pro is governed by X~
co-occur in postverbal positions, despite the fact that rich AGR and non-
pronominal syntactic arguments are truly incompatible. In other words,
we can try to pursue the idea that (30) and (31) are not ruled out for the
same reason, and that verbs agree differently with (postverbal) subjects,
depending on whether they are pronominal or non-pronominal. This problem
is dealt with in the next subsection.
As has already been observed, when subjects are postverbal, verbs agree
with them only in GEN. It might be suggested, however, that this limita-
tion is true only of non-pronominal subjects. It does not carry over to
pronominal ones. With the latter, verbs agree in all phi-features, hence the
ungrammaticality of (42):
(42) * jaa?a-thunna
came-f they.f
(Intended to mean: They (f.) came).
If this true, why then is (31), in which the verb agrees with the pronoun,
excluded?
Suppose that nothing is wrong with agreement in (31), and that the
ungrammaticality of the latter is due to something else. Assume, for
example, that an independent pronoun is licensed only when it is focused,
and that non-focused pronouns cannot be phonetically realized, in confor-
mity with Chomsky's (1981) 'Avoid pronoun' principle. Indeed, when
pronouns in postverbal position are focused, their grammaticality is not
disputed. This is illustrated by (43):
(43) jaa?-uu hum Iaa xuddaam-u-hum
came-3.pl.m. they.m. not servants-nom-their
They came, not their servants.
If that is true, then the complementarity in distribution is only apparent, and
INFLA (or pro drop) cannot be rejected as quickly as it might seem. 18
Upon closer examination, however, the claim that verbs agree differ-
ently with pronouns than they do with nouns is not supported. First, as
pointed out above, when the subject pronoun cannot be incorporated, the
verb does not agree with it in all phi-features, contrary to what INFLA
predicts. This is exemplified by (44):
(44) lam y-a?ti ?illaa hum
not.past 3-come except they.m.
There came noone but them.
Here the agreement is limited to the poor option (in the simple case).
114 CHAPTER 3
b. huwa jaa?-a
he came-J.s.m.
He came.
In those contexts, pronouns are not (necessarily) focused or strong. They
contrast with strong pronouns in examples (45) to (47) above, or more
directly, with the stressed pronoun in topic position found in (49):
(49) hunna laa y-ubaalii ?-ai)ad-un bi-hinna
they.f not 3-care one-nom about-them.f
As for them (f.), nobody cares about them.
Here the pronoun is a focused dislocated topic anaphorically related to
the incorporated resumptive pronoun on the preposition. Thus, despite a
formal resemblance, pronouns in (48) on the one hand, and (49) on the other,
do not have the same feature content, nor the same behaviour. The differ-
ence can be tested through the ability of incorporating into a governor. When
embedded in a context where they are governed by a lexical governor, weak
(subject) pronouns must incorporate. Recall examples (3) and (4), repeated
here as (50a) and (SOb):
(50) a. ?inna-ka marii<;l-un
that-you.acc sick-nom
You are (indeed) sick.
b. i)asib-tu-hu jaa? -a
thought-1-him came-3.s.m.
Literally: I thought him came.
I thought he came.
But a strong pronoun does not incorporate. It can only double a pronom-
inal affix, as the following examples show:
(51) ?inna-ka ?anta 1-mas?uul-u
that-you you the-responsible
You are the responsible indeed.
(52) ?inna-hu huwa ?a-maata wa ?a-i)yaa
that-him he make-die.past and make.alive-past
It is He (= God) who kills or who makes (people) alive.
venors in shaping their form, subjects turn out to have free and bound
weak forms, depending on contexts. Moreover, strong forms that occur in
postverbal positions in conjunction with a rich inflection on verbs are hardly
analyzable as subjects. The subject is the inflection, seen as an incorpo-
rated pronoun, but the strong pronoun is doubling the subject. If this line
of analysis is correct, then it has a number of interesting consequences.
In particular, verbal inflectional markers now appear to be morphologi-
cally ambiguous, since they can function as 'pure' inflections or as
pronouns. We turn to this problem in the next section.
2.1.2. Expletives
H forms are also used as expletives. In this use, they can be specified
for no feature. But they can also be specified for GEN, or NUM and GEN,
but not PERS.
Like English it (or French ce), the H form can appear as a surface subject
of an identificational sentence, as in the following examples:
(57) huwa 1-kasal-u
he the-laziness
It is laziness.
In order to solve this problem, I assume that H in (57) and (58) originates
in predicate position, and move to subject at S-structure (see Chapter 2).
On the other hand, predicative H agrees with the thematic subject in NUM
and GEN, as illustrated by the following contrast:
(61) a. humu 1-falaasifat-u
they the-philosophers-nom
It is philosophers.
Consider again the forms presented in the tables given in Subsection 1.1.
A quick comparison reveals the existence of a transparent morphological
(and presumably etymological) relationship between the non-nominative
forms in Tables I and 2, on the one hand, or between nominative forms
in Tables 3-5, on the other. Apart from a few idiosyncracies, free/bound
pairs are almost identical in form, with the sole difference that there is a
formative which serves as a support for the pronominal morpheme in the
independent form (?iyyaa for accusative forms, ?an for some nominative
forms, depending on the person involved, etc.). Moreover, there are close
similarities between nominative and non-nominative forms, though case
differences sometimes make them less directly transparent. To illustrate,
consider the pronominal forms in the following pairs:
(67) a. ji?-tumaa
came-you. dual
You two came.
b. ? antumaa mariic;l-aani
you.dual sick-dual.nom
You two are sick.
b. ?iyyaa-hum qa~ad-tu
them meant-!
It is them who I meant.
PRONOUNS, INCORPORATION, AND FEATURE SPECIFICATION 121
(75) D
~
D AGR
I~
[+<lef] N AGR
I
?an
I
tumaa
[+pro]
(76) D
~
D AGR
[+def] ~
N AGR
I
[+pro]
I
tumaa
124 CHAPTER 3
In preceding subsections I have shown that both bound and free forms
can be used ambiguously, depending on how their feature values are
specified. Specification of forms can be null, partial (e.g. limited to GEN),
or complete, depending on lexical entries and syntactic contexts. In this
subsection I examine how specification is divided between lexicon and
syntax. I also show that specification observes a feature hierarchy in which
a specification for a feature implies the specification of all features lower
than it in the hierarchy.
b. ?anaa gakiyy-at-un
I clever-f-nom
I am clever (f.).
These facts suggest that specification of features is hierarchically ordered.
A phi feature hierarchy can be stated as follows:
are there pure object AGR types or ambiguous object forms in Arabic;
or, more generally, are there non-nominative ambiguous forms? The question
is important given Chomsky's ( 1989a) conjecture that there is an AGRO,
wich parallels AGRS(ubject), and also work on Celtic proposing that
there are agreement conjugations of non-verbs (in particular prepositions;
see in this regard McCloskey and Hale, 1984, and Rouveret, 1991, and
references cited there).
It is striking that there is no object (or non-nominative) AGR. Clearly,
there is no case of poor agreement with objects (or non-subjects). For
example, when a verb or a preposition governs a syntactic NP, no agree-
ment form is attached to the governor as a mark of the object. This is
illustrated by the following constructions:
(90) a. ntaqad-tu r-rajul-a
criticized-/ the-man-ace
I criticized the man.
see Fassi, 1981); but this is irrelevant to our problem. Note also that similar
judgments are found with prepositions or nouns. These facts suggest that
there is no descriptive reason to postulate the existence of an AGR object
in Arabic.
If this reasoning is correct, then bound non-subject forms which occur
on verbs, prepositions, nouns, etc., are best analyzed as incorporated
pronouns (not inflections), along the lines discussed earlier. The following
examples (Nos. (11)-(13) above), illustrating instances of incorporation, are
repeated here for convenience:
(93) intaqad-tu-hu
criticized-1-him
I criticized him.
(94) ltaqay-tu bi-hi
met-/ with-him
I met him.
(95) ntaqad-tu mu? allif-a-hu
criticized-/ author-ace-his
I criticized its author.
As has already been mentioned, the internal structure of these pronominal
affixes is identical to that of their independent counterparts (listed in Table
4), except that they lack an overt pronominal support. Recall, too, that
one feature of non-nominative pronominal incorporation is that it is not
sensitive to the nature of the governor. As instantiated by the examples here,
the same form [-hu] functions as object of a tensed verb, complement of
a preposition, or of a possessor. This state of affairs contrasts significantly
with that of nominative AGR forms. The latter vary depending on whether
the governor is an adjective or an inflected verb, and depending on
Tense/Aspect carried by the inflected verb.
These features translating the morphological autonomy of non-nomina-
tive bound affixes indicate that these forms are not incorporated into
(clausal) inflection, unlike nominative affixes. The latter are best thought
of as incorporated into I, as we have seen (see also below, Subsection
3.2). If this is true, then this suggests that the sole reason why non-nomi-
native pronominal forms have to incorporate relates to their bound
morphological nature. As a corollary, we expect that when these forms
are prevented from incorporating into their governor, no AGR form would
appear on it, unlike what happens with nominative forms. This expecta-
tion is borne out, as has been amply demonstrated above. The fact that
non-nominative affixes do not incorporate into inflection strongly supports
their non-inflectional nature.
130 CHAPTER 3
(97) *jaa?-athunna
came-! they.f
They (f.) came
postulating the existence of AGR, then the Arabic data provide evidence
for the existence of AGRSs in Chomsky's (1989a) sense, but not for
AGROs. The question then is: why? Why do we have agreement mani-
festations of subjects, but not of non-subjects? How are AGR markers
licensed?
A possible answer is that the licensing mechanism for inflectional AGR
is predication. 32 Since the subject is a member of a predicate relationship,
but the object is not, inflectional AGR is available with subjects, but not
objects. Let us then suppose that a correlation holds between inflectional
AGR and predication, in conformity with the following principle:
Moroccan Arabic (MA), both of which are SVO languages, reveals that
incorporation is possible in MA, but not in English.
English is an SVO language which does not instantiate incorporation
of subjects nor that of objects. As postulated earlier, English simply has
no pronominal AGR, by virtue of having fixed the AGR pronominality
parameter that way. Moreover, there is no evidence that pronouns are
of affixal nature in this language. These two factors converge to bar
incorporation in English. Turning now to MA, we observe that pronom-
inal objects are (normally) affixed onto their governors, as illustrated by
(101) and (102):
(101) r-rajl kla-h
the-man ate-him/it
The man ate it.
(102) r-rajl ja mca-h
the-man came with-him
The man came with him.
In (101) the object of the verb must be a bound pronominal form. The
same is true of the object of the preposition in (102). Note that the
occurrence of an incorporated form is incompatible with that of a syntac-
tically realized DP, as the ungrammaticality of (103b) shows:
(103) a. kla 1-xubz
ate the-bread
He ate the bread.
b.* kla-h 1-xubz
ate-him/it the-bread
In (103a) the syntactic DP object is expressed, while it is incorporated in
(101). Co-occurrence of both forms is excluded, as (103b) shows.
MA has null subject constructions, as (104) indicates:
(104) ja-w
came-3.pl.
They came.
In Subsection 1.4 above I have hypothesized that the subject in these con-
structions is a null pro, thus parallelling similar Italian cases analyzed by
Rizzi (1982). The idea behind this approach is that the AGR marker on
the verb is best treated as a pure inflection because it can co-occur with syn-
tactically expressed subjects, as in ( 105):
PRONOUNS, INCORPORATION, AND FEATURE SPECIFICATION 135
In this chapter I have provided reasons for integrating pronominal and AGR
forms into one general system, and I have proposed mechanisms to derive
their different alternations and uses. On the one hand, close morpholog-
ical (and etymological) relationships between independent or bound
pronominals and pure agreement markers have been established. Phi feature
forms have been decomposed into three components: a D component, a
pronominal support, and an AGR component. Some forms have been shown
to be lacking some of these components, although they all share the AGR
component. On the other hand, the AGR pronominality parameter accounts
for the existence of null subject constructions, as well as alternative uses
of subject affixes as pronouns. The functional ambiguity of forms has
also been accounted for through the elements of a theory of (under)-
specification of phi feature forms. Moreover, the latter has been shown to
observe a hierarchy. Free and bound forms have been generated (and
interpreted) in different functional positions: under DPs, or under IPs. Other
differences between nominative and non-nominative affixes, or between
languages have been analyzed.
136 CHAPTER 3
NOTES
* An earlier version of this chapter has circulated in a manuscript form as Fassi (1989a).
Part of this work was presented in a GLOW talk (Spring 1990, London), and in various lectures
at the University of Paris VIII, SOAS (London), the International Linguistic Institute of
LSM at the University of Rabat (Summer 1990), UQAM (Montreal) and MIT (Spring 1989).
Thanks to audiences for helpful comments.
1 On 'richness' of INFL as a trigger of null argumenthood, see Taraldsen ( 1978). On various
approaches to pro drop, see especially Chomsky (1981, 1982), Rizzi (1982, 1986), and Jaeggli
and Safir (1989).
2 The term 'pronominal' here is not to be confused with that used by Rizzi (1982) to
characterize the !NFL of null subject languages like Italian. See Section 2 for detailed
clarifications.
3 Note that our treatment of bound forms is neutral with respect to the issue whether the
grammar is derivational, as in e.g. Baker (1988), or representational as in e.g. Rizzi (1987).
4 On diachronic evidence that Piyyaa pronouns have been formed from bound roots via a
sort of 'excorporation' at a later stage in the history of the Arabic language. See Bravman
(1971 ).
5 See note 22, however.
6 On the structure of DPs, see Chapter 5. Postal (1966) has argued that English pronouns
should be treated as articles. See also Hale ( 1988), who proposes taking pronouns as Ds. In
Section 2 personal pronouns are analyzed as DPs, and expletives as AGRPs (with no D
projection).
7 The construction is grammatical as a right (accusative) dislocation, but this is irrelevant
here. See Fassi (1981) on the matter.
8 A highly abstract approach to the matter is possible, but I see no significant conse-
quences of it. See Subsection 3.1, below, for a discussion.
9 This construction is grammatical if the NP after wa is taken as an 'accompaniment' adjunct.
The relevant translation is then the following: 'I saw you with Zayd'.
10 Note that other readings are possible here. For example, (23a) can have also the following
interpretation: 'I gave him to you'. Thus the use of a mixture of free and bound pronouns
renders the order of object 1 and object 2 without importance, although the order is totally
strict when only affixes are used (see Chapter 2).
11 Note that the Person Constraint does not apply to (nominative) subject affixes. The
following example illustrates this fact:
(i) <;Iarab-uu-ka
beat-3.pl.m. -you
They beat you.
We will see, on the other hand, that nominative affixes, unlike non-nominative ones, incor-
porate into the clausal INFL. This observation suggests that the Person Constraint must
take this difference into account.
12 I will not discuss the properties of these forms exhaustively here. See Fassi ( 1984) for
detail.
ll This is in line with Hale's (1988) argumentation for incorporation in Irish.
14 Here, too, canonical government is necessary, though not sufficient. Consider the fol-
lowing pair of constructions:
(ii) ?a t:Iasan-un huwwa
Q nice-nom he
Is he nice?
PRONOUNS, INCORPORATION, AND FEATURE SPECIFICATION 137
*?a l).asan-u-hu
Q nice-nom-him
Is he nice?
In (i), the subject pronoun is independent, although governed by the adjective. The construction
(ii), in which the subject pronoun is incorporated into the adjective, is ungrammatical. Only
non-nominative pronouns can incorporate into adjectives, as in (14) above. Nominative
pronoun incorporation is limited to finite verbs. Since the latter carry Tense, and since
bound pronominal subjects are inherently marked for Nominative Case, we might assume
that in order for Case to be checked (after incorporation), the presence of Tense (on the
incorporating governor) is necessary. Because adjectives do not bear Tense, the incorpora-
tion of a subject on adjectives will not satisfy Case requirements, and Nominative checking
cannot be felicitous. The only option then left for the subject pronoun is not to incorporate,
and to have its Case checked under government by clausal T.
15 One potential problem for INCA pointed out by Radford (personal communication) is
that, given this analysis, one would normally expect bound pronouns to be 'phonetically
consistent', in the same way as eli tics are. The latter show only minimal morphophonolog-
ical variation in form, so that e.g. Italian gli becomes glie before lo, etc. This is not true of
Arabic incorporated pronouns, however. These forms are sensitive to perfect/imperfect
distinctions, and they exibit discontinuity. These observations assess the lack of a (rela-
tively) consistent homogeneous form, which appears in turn to be a mark of inflection rather
than that of an incorporated pronoun. I think that Radford's observations do not necessarily
undermine INCA, because Arabic is a highly fusional language at the morphological level,
and there is no necessary one-to-one mapping between syntax and morphology. INCA being
a syntactic process, it may or may not be motivated on syntactic grounds. As for (concrete)
morphology, it is clear that it does not mirror syntactic relationships (see Appendix 1 of
Chapter 2, and references cited there).
16 For a thorough discussion and criticism of this variant of the Pro drop analysis, see
Fassi (1984).
17 McCloskey and Hale ( 1984) have adopted a Pro drop approach of inflection in Irish,
but Hale (1987, 1988) argues convincingly that a variant of INCA is more convenient for
Celtic and other languages.
18 Note that I have assumed that the Avoid pronoun principle will rule out the cases
discussed, although, as formulated, it is not clear how it does so.
19 The strong/weak distinction is found in traditional grammar. It has been made use of in
recent literature. See e.g. Holmberg (1991) and Rouveret (1991), among others.
20 See Higginbotham (1985) for discussion.
21 Note that first and second PERS free pronouns are used only as arguments. Bound
forms of first and second PERS, however, are (morphologically) ambiguous in that they
can be interpreted as inflections, or as incorporated pronouns. These idiosyncracies have to
be stipulated in the lexicon.
22 The morphological relationship is more transparent with first and second PERS than
with third. With the latter, [hum], [hun] or [h] may function as a support with third PERS
nominative, although it is part of the bound form with the accusative (see Table 2). Some
third PERS forms might be historically related to adjectival agreement affixes (see Russell,
1984). It is clear, however, that the former are functionally different from adjectival affixes.
One important difference already mentioned is that third pronominal affixes license null
anaphora, but adjectival affixes do not. See note 28.
23 From this perspective, they share a common core of properties with Romance clitics,
and are presumably derived through the same mechanism.
24 In Arabists' terminology, 'sound' plurals designate plurals which are formed from the
stem of the singular by adding a suffix (e.g. -uun) concatenatively. In contrast, broken
plurals are not formed from the stem, but only from the consonantal root. Their morphology
138 CHAPTER 3
is non-concatenative, and the vocalic information of the singular is lost in the plural (see
Wright, 1974 and McCarthy and Prince, 1988, among other references).
25 The [n] ending on the adjective is a (abstract) mark of possession. See Chapter 5 on
this matter.
26
NUM and GEN license null arguments with imperatives. Consider the following examples:
(i) dxul-uu
enter-pl.m.
Enter (pl.m.)!
(ii) dxul-ii
enter-fs.
Enter (f.s.)!
In these verbal forms, the prefix element (which usually appears on imperfect verbs to mark
PERS) is missing. Only NUM and GEN are encoded, and second PERS is inferred from
the mood (imperative). I assume that the bound form here is an agreement marker, not a
pronoun, and that the necessary PERS feature to license null arguments is provided by the
context.
Likewise, constructions like (iii) are grammatical in appropriate discourse context:
(iii) ?a waaqif-uun?
Q standing-m.pl.nom.
Are you standing?
The contextual value of PERS here should also be 2.
27 DEF is phonologically realized in the guise of a definite article with demonstrative and
relative pronouns like lladii (where [I] is the definite article followed by a demonstrative
support).
28 We still have to explain how the ambiguity is allowed, and which one is allowed. Consider
again the system of forms given so far. There are ingredients of regularity which suggest
that the system is further decomposable. For example, in perfective affixes, [a], [aa], and
[uu] stand for singular, dual, and plural respectively (though the latter is necessarily
masculine). The last two forms are also found in other systems, including adjectival and
noun systems. [t] is for 2 PERS usually, although it is also used in 1 PERS of the perfect,
and as feminine of 3 PERS, etc. Clearly, however, the system cannot be decomposed in
any straightforward transparent fashion. It is highly idiosyncratic from the synchronic point
of view, and some forms allow no decomposition at all. For example, [?-] is [lPERS, S.,
IMPERF], and unspecified for GEN; [n-] has basically the same specification, except that
it is PL.; [t-] is ambiguous between 2 PERS and 3 PERS. If it is 2PERS, it is M. and S.
by default; if it 3 PERS, it is also F. and S. by default. [t] and [y] in the imperfect give
rise to different NUM and PERS, depending on the suffix combined with them. For example,
[y] occurs with the PL.F. suffix [-na], but not with the dual F. [t-aa].
Thus segmentation of the forms of affixes is not possible. Moreover, it has no impact
on the functioning of the system, as far as we can tell. The same is not true of the
decomposition of affixes into features, especially in comparison with other affixes. In
particular, (under)specification of a form may give rise to ambiguity, thus accounting for
the ambivalent behaviour of the affix. Consider, for instance, perfect affixes for 3 PERS
S., i.e. [-a] and [-at]. If we compare the internal structure of 3 PERS form suffixes with
that of non-3 PERS suffixes, we realize automatically that the first consonant which stands
for PERS ([t] or [n]) is missing here (except in PL.F.). That is, the form of the affix tells
us that there is no specification of PERS. Bare vowel [a] cannot be taken as a specification
of S., as opposed to long [aa] for the dual, or long [uu] for plural. This is so because when
GEN and NUM are separated in the affix, GEN comes first, followed by NUM, as in
PRONOUNS, INCORPORATION, AND FEATURE SPECIFICATION 139
[at-aa] which is [F.-DUAL]. If this is true, then [a] has no grammatical specification. Its value
is only negative. If 3M. S. are default specifications, then [a] expresses them, in the absence
of positive marks. As for [-at], it is only marked for GEN, as feminine. In the optimal case
then, a representation of a perfect affix, for example, is the following:
(i) [-- PERS GEN NUM] rertect
Dashes preceding the specification point to the suffixing character of the form. Note that
no affix instantiates this (ideal) decomposition by three discrete segments. Affix forms, if they
allow decomposition at all, have no more than two segments. Others merge all these
categories in a single non-divisible segment.
It is possible, hopefully, to undertand how ambiguity arises, especially in the least
specified forms of the pronominal/agreement system. A form like [a] has no specified values
of its features. The latter can receive default interpretation, and hence 'license' null anaphora,
as in (ii):
(ii) jaa? -a
came·3.s.m.
He came.
This form can also occur with syntactically expressed subjects, as we have seen. In this
case, only GEN is specified at best. See note 30.
29 There is obviously a distinction to be made between Italian, a null subject language,
and English, which is not. Our purpose here is mainly to characterize the difference between
incorporating and non-incorporating languages. As for Italian, if we maintain the spirit of
Chomsky's (1982) and Rizzi's (1982) analyses, then it has to be accounted for via a variant
of INFLA. The problem which arises then is to know whether both INCA and INFLA are
needed. It is possible that the latter can be dispensed with, if Italian is treated like Moroccan
Arabic. See below, Subsection 3.3.
30 In order to distinguish finite verb inflection from adjectival inflection, it is necessary to
assume that the feature matrix of the former, but not that of the latter, includes a PERS feature,
even though this feature may be unspecified. The occurrence of this feature will be licensed
by Tense (Mood, or Aspect), as we will see in Chapter 4.
31 Cross-linguistically, the descriptive situation might be complex. In Welsh, for example,
verbs and prepositions seem to behave alike in that both can carry rich inflection (when
subjects of verbs or objects of prepositions are pronouns). In order to account for this
behaviour, Rouveret (1991) proposes a variant of INCA, in which only NUM incorporates
in the case of subjects, but the whole pronoun does in prepositional cases. Note, however,
that there is no incorporation of object pronouns. In other words, there is no 'synthetic'
form of the verb for objects, as there is one for subjects. On the other hand, independent
pronouns may appear as objects of 'analytic' prepositions, suggesting that incorporation is
not necessary. The fact that incorporation of objects is either not permissible or optional.
but not obligatory (as in the subject case) is expected given (a) the non-affixal nature of object
pronouns and (b) the absence of the need of agreement checking in the case of objects.
32 On predication, see Williams (1980) and Rothstein (1983).
33 The question remains open whether coindexation in Chinese and Japanese- which lack
overt subject predicate agreement- involves AGR in addition to T. or just T. See Kitagawa
( 1986) for a discussion.
34 As a matter of fact, the cases treated by Chomky (1989a) as object AGRs (or AGRO's)
occur in participle phrases, but not within VPs. Participles have indeed an inflectional AGR
projection, though verbs and their objects do not. Agreement in participial phrases is exactly
parallel to Spec-Head agreement in adjectives, which is a form of AGRS, as shown in Fassi
(1988b) and Mahajan (1989). What would be surprising, though. is to find a verb which agrees
with its object, as represented in a configuration like (i):
140 CHAPTER 3
(i) AGRP
~
Spec AGR'
~
AGR VP
~
v 0
In that configuration, it is hard to see what structural relation can be established between AGR
and 0. The latter is not in the domain of AGR. as far as I can tell.
Verbs tend to agree with their objects (or more generally with their arguments) in so-called
non-configurational languages (see Hale, 1988 and Alexander, 1990). In these languages, it
may be hypothesized that the verb has incorporated all its arguments in the form of
(referential) AGR markers. As for syntactic NP constituents, they stand in an adjunction
relation to these incorporated AGRs. Thus these AGRs function as true arguments, and
syntactic NPs are in some sense 'doubling' them. In these languages, agreement markers
may be discharging theta-roles (see Alexander, 1990 for suggestions).
35 It might be objected that languages such as Italian provide motivation for INFLA. Note,
however, that the situation of this language does not appear to be different from that of
MA.
On the other hand, there are languages. like Standard French, which allow only SVO order,
but in which pronominal subject clitics presumably incorporate into the verbal complex.
This state of the matter might seem problematic, as pointed out by an SNLLT reviewer. It
is not clear, however, why it would be so. Observe that both subject clitics and syntactic
subjects co-occur in Subject-Aux inversion in Standard French, as in Jean a-t-il mange?
(see also non-standard French where they co-occur freely). This indicates either that two
subjects originate in different argument positions (and that the clitic incorporates from one
of them), or that the latter is base generated under !NFL. Whatever the solution, it is not
necessarily problematic (for a recent analysis of subject inversion in French, see Rizzi and
Roberts, 1989).
CHAPTER 4
141
142 CHAPTER 4
In this section I deal with the issue of how Arabic temporal morphology
is to be construed. After discussing some preliminary assumptions about
how Tense and Aspect are articulated in a theory of temporal relations, I
provide supporting evidence for the TMA view of verbal inflection.
1.1.1. Aspect
a stative sentence like Mary knows Fred can have the following
notation:
(5) (3 e) [Knowing (e) & Subject (e, Mary) & Object (e, Fred) &
Hold (e, now)].
This notation reads as follows: there is a knowing that has Mary as its
subject, has Fred as its object, and hols now. In contrast, an event sentence
like (1) will have the following logical form:
(6) (3 t) [t now & (3 e) [Eating (e) & Subject (e, John) & Object
(e, the pizza) & Cui (e, t)]].
1.1.2. Tense
(20) a. t-aktub-aa-ni
2-write-dual-indic
You (dual) write.
b. ?-uriid-u ? an t-aktub-aa-~
1-want-indic that 2-write-dual-subj
I want you two to write.
c. li-t-aktub-aa-~
let- 2- write-dual-juss
Let you two write.
Note that the Mood mark on the verb follows the AGR marker. 8
This brief description enables us to assess a 'moodal' opposition between
imperfect and perfect forms: the former are marked for Mood, while the
latter are not. 9 Moreover, imperfect forms stand in opposition to each other
with respect to Mood marking. Mood completes the three-dimensional
152 CHAPTER 4
Moreover, that interpretation does not force the phonetic visibility of the
copula, only Modality does, as I have explained.
Note that the interpretation of stative and processive finite verbs
corroborates also the view that the present reading is unmarked. As the
interpretation of (9) indicates, a processive present can have either
simultaneous or posterior readings. But a stative verb in the present is
compatible only with a simultaneous reading, as the following examples
show:
(27) a. ? -afham-u kalaam-a-ka (1-?aan-a, *gad-an)
/-understand saying-ace-your (now, *tomorrow)
I understand what you are saying.
b. hal t-a"rif-u 1-jawaab-a (1- ?aan-a, *gad-an)?
Q you-know the-answer-ace (now, *tomorrow)
Do you know the answer?
In order to obtain the future reading with stative verbs, a modal like sawfa
must be introduced, as is the case with stative adjectives (although no copula
is necessary here).
Observe that, unlike the future, the interpretation of the past (or anterior
relation) does not depend on the aspectual nature of predicates or
situation types described by the latter. Whether the predicate is a pure stative
as in (22) above, an inchoative stative as in (28a), or a processive, as in
(28b), the inflected form uniformly describes a situation (a state or a process)
which is anterior to the utterance time:
(28) a. mari<;i-a r-rajul-u
became.sick the-man-nom
The man became sick.
b. tajawwal-uu
walked-3.pl.m.
They walked.
This independence of Past vis-a-vis situation aspect reinforces its status
as the positive member of tense oppositions. In contrast, the existence of
present and future with positive tense values cannot be so established. On
the contrary, the distribution of the present form of the copula indicates
that its visibility is not regulated by tense characteristics, but only by
modal and aspectual considerations. This conclusion will be systematized
in the next subsection.
TEMPORAL, ASPECTUAL, AND MODAL CATEGORIES 155
As we have seen earlier, the visibility of the copula is forced by past tense
(with stative predicates). This is not true of future or present interpretations.
As observed, processive participles express future (or present) without
introduction of the copula. The presence of the latter is required only with
stative adjectives located in the future. Although participles carry no finite
(tense) morphology, they behave like finite (processive) verbs in allowing
both future and present interpretations. The future interpretation is not
available, however, to stative predicates, indicating that future is sensitive
to the type of states of affairs. In order for future interpretation to obtain
with stative verbs, a modal must be introduced. The latter, in turn, forces
the introduction of a copular verb (if no other verb is there), to satisfy its
selectional requirements.
There are other aspectuo-temporal and modal conditions which require
the appearance of the copula. For example, the copula occurs when a stative
adjective or locative conveys a general or habitual meaning, as in the
following examples:
(29) cindamaa *(y-akuun-u) r-rajul-u marii9-an
when is the-man-nom sick-ace
fa- ?inna-hu laa y-ubaalii
then-that-him not 3-cares
When the man is sick, he does not care.
(30) Q.iina ?-akuunu fii d-daar-i ?-akuun-u murtaal)-an
when l-am in the-house-gen l-am relaxed-ace
When I am at home I am relaxed.
It also occurs in complex tenses such as present or future perfect:
(31) t-akuun-u ?axta?-ta 1-hadaf-a
you-are missed-you the-goal-ace
You (will) have missed the goal.
Another context marked for Mood in which the copula must appear is imper-
ative or so-called prohibitive (negated imperative), as in the following
constructions:
(32) kun ?ustaag-an naajil)-an
be.m.s. professor-ace successful-ace
Be a successful professor!
laa t-akun gabiyy-an
not 2-be.m.s. silly-ace
Do not be silly!
156 CHAPTER 4
In this section I argue that Tense, Aspect, and Mood distributions are best
treated by associating verbal inflection with a matrix of TMA features,
paralleling AGR feature matrices, which may or may not be (under)-
specified. Correlatively, a hi-inflectional temporal structure is postulated
to account for auxiliary structures and modal or negative constructions.
Moreover, it is shown that a mono-inflectional structure in which TMA
are taken to be split heads of different syntactic projections within a unique
(local) domain is inappropriate to describe internal structures of those
constructions.
finite AGR (as argued in Chapter 2), and finite T, if the analysis proposed
in Subsection 1.2. is correct. In (37), for example, the prefix on the
auxilliary verb marks PERS and GEN, the suffix marks Mood, and the
internal vowels Tense or Aspect; in the thematic verb, the suffix marks AGR
features, and the internal vocalic pattern Tense and Aspect. Both (38) and
(39) show that Neg(ation) can be placed inside the functional domain in
which the thematic verb is included, whereas Neg in (40) is in the domain
in which the auxiliary is found. This distribution of Neg suggests that
both domains are tensed. 11
In languages like English or French, auxiliary verbs are usually taken
to select (or govern) VPs, not IPs. Consequently, constructions like the
following are excluded:
(41) *John had ate. (John had eaten.)
(42) *Jean avait mangeait. (Jean avait mange.)
2.1.2. Modals
Now consider modal structures: it has been pointed out by Emonds (1985),
among others, that English modal verbs are best analyzed as instances of
the general inflectional category AUX (commonly replaced by INFL in
recent literature). Emonds argues that modals are not verbs, but inflections. 15
Moreover, Emonds' analysis points to two properties of English modals.
First, they are in complementary distribution with tenses, as illustrated by
the following rules:
(45) AUX ~[±TENSE,± PAST]
(46) [- TENSE, + PAST] ~ {would, could, might, ... }
(47) [-TENSE,- PAST] ~ {will, can, may, ... }
Second, AUX is in construction with a bare VP, since in his system AUX
is a Spec of VP. Other recent studies have taken AUX (replaced by INFL)
to be heading the sentence, and selecting a bare VP (see e.g. Chomsky, 1981,
1986b).
Are these two properties true of Arabic modals? Arabic modal verbs
do not govern IPs nor VPs. They are necessarily followed by what appears
to be a CP, as in (48):
(48) y-ajib-u ?an ?-aktub-a
]-want-nom that 1-write-subj
I want to write.
In other words, modals do not appear to belong to the same inflectional
domain as that of the thematic verb. At any rate, they can hardly be seen
as the exclusive instantiater of INFL.
The closest categories to be compared with English modals are parti-
cles, not verbs. Like English modal verbs, modal particles select verbal
projections, not CPs. But Arabic modals, unlike those of English, govern
160 CHAPTER 4
verbal projections which are inflected for TMA (in addition to AGR).
Thus these projections are, in fact, IPs (or TPs), not VPs, despite the fact
that Arabic modals, like their English verbal counterparts, can be shown
to express both modal and temporal meanings.
Modal qad, which expresses 'certainty' or 'effective happening' when
used with past forms of the verb, is also used to convey 'near past' temporal,
as in (49): 16
(49) qad qaam-at ~-~alaat-u
in fact came-f the-prayer-nom
a. The (time of) prayer has in fact come.
b. The (time of) prayer has just come.
With non-past form, qad (which is said to be 'expectative'), according to
traditional grammar, implies future referenceY
(50) qad y-a?kul-u
may 3-eat-m.s.indic
He may eat.
Similarly, the modal sawfa (which means 'procrastination') forces future
reference with non-past verb forms, although the latter morphology is neutral
with respect to the future/present distinction:
(51) sawfa y-a?kul-u
will 3-eat-m.s.indic
He will eat.
These facts (and others) can be accounted for along the following lines.
Modal particles are generated as heads of a Mod(al) projection, and they
select a finite projection headed by TMA. Both Mod and TMA contribute
to temporal meaning, but there is no sense in which Mod and T can be
treated as complementary in distribution, as would be the case in English.
Moreover, there is no sense in which modals can be taken as selecting/
governing bare VPs. If this is true, then Arabic modal structures can be seen
as bi-temporal (or bi-inflectional) in the sense that both Mod and T, which
are viewed as temporal, project in the same functional domain.
and which has the same form as nominative Case in nominals; [a], which
stands for so-called subjunctive, and which has the same form as accusative
in nominals; [0] or absence of ending which stands for so-called jussive.
I will follow traditional grammarians in taking the first TCase to be assigned
by default (thus paralleling Nominative in the nominal system), whereas
other cases are assigned/checked under government. In (52), the embedded
verb is marked with [u], the indicative. In (53), the absence of ending marks
the jussive. In (55), the [a] ending expresses subjunctive:
(55) Ian y-a?kul-a
not-jut. 3-eat-subj
He will not eat.
Note that the Neg marker is temporally specified (as posterior or future),
unlike the case with neutral Neg, as we will see.
Other distributional properties of modal Negs can be accounted for along
similar lines. For example, these Negs do not occur in nominal sentences,
as shown by the ungrammaticality of (56):
(56) *Ian zayd-un mariic;l-un
not.fut. Zayd-nom sick-nom
Intended to mean: Zayd will not be sick.
If a verbal copula is inserted, carrying the appropriate TCase or Mood,
then the sentence becomes grammatical:
(57) Ian y-akuun-a zayd-un mariic;l-an
not.fut. 3-is-subj Zayd-nom sick-ace
Zayd will not be sick.
This behaviour recalls that of other modals, which must govern inflected
verbs, needed to support Mood.
Another property of this type of Negs is that they do not occur with
SVO structures, as the ungrammaticality of the following construction
indicates:
(58) *Ian zayd-un y-a ?tiy-a
not.fut. Zayd-nom 3-come-subj
Intended to mean: Zayd will not come.
It seems then that adjacency is required for TCase checking. The
adjacency effect can in fact be derived on grounds of principle. Recall
that SVO structures are nominal projections, because I there is nominalized.
I in SVO can then be seen as an NCase carrier, not a TCase (or Mood)
carrier. Consequently, Mood requirements of Neg are not satisfied, and
the construction is ungrammatical. Note that this situation recalls also that
TEMPORAL, ASPECTUAL, AND MODAL CATEGORIES 165
of other Mood markers. The complementizer ?an, for example, which marks
the verb with subjunctive is incompatible with SVO order. 20
A second type of Neg is instantiated by laysa. The latter behaves like
a copular verb in assigning accusative Case to the predicate nominal or
adjective, as exemplified in the following construction:
(59) lays-at hind-un ?uxt-a 1- ?ustaad-i
not-fs. Hind-nom sister-ace the-professor-gen
Hind is not the professor's sister.
Like a copular verb, too, Neg merges with sentential AGR. It heads either
a nominal sentence, as in (59), or an SVO sentence, as in the following
construction:
(60) lays-at hind-un t-adrii maa 1-camal-u
not-f Hind-nom f-know what the-doing-nom
Hind does not know what to do.
These properties have led traditional grammarians to treat this type of Neg
as a verb. I will adopt this essential idea here.
A third type of Neg can be seen as neutral. It is instantiated by maa.
This morpheme can occur in both verbal and nominal sentences, as
illustrated by the following examples:
(61) maa daxal-a r-rajul-u 1-qaacat-a
not entered the-man-nom the-room-ace
The man did not enter the room.
(62) maa ?anaa qaadir-un calaa haagaa
not I capable-nom of this
I am not capable of this.
It is found with VSO order, as in (61), or SVO, as in (63):
(63) maa ?anaa qul-tu haagaa
not I said-/ this
I did not say this (or: it is not me who said this).
It is also neutral with respect to whether the verb is in the anterior form,
as in (63), or the non-anterior form, as in (64):
(64) maa ?abad-un y-asukk-u fii qawl-i-ka
not one-nom 3-doubt-s.m. in saying-gen-you
No one is questioning what you said.
These distributions indicate that maa is not merged with inflection, and
that it does not have any Case marking or Mood marking properties.
166 CHAPTER 4
2.3.1. Anteriority
As indicated earlier, Neg morphemes differ with respect to whether they are
compatible with both anterior and non-anterior forms of the verb, or with
only one of these forms. This has been illustrated by comparing maa with
other Negs, which are TCase assigners. Some morphemes, however,
although not (necessarily) Mood assigners, select only non-anterior forms.
This is illustrated by the following grammaticality judgments:
(78) maa l)a~ala haagaa
not happened this
This has not happened.
(99) negates performing the prayer at the time of utterance, (98) is rather
a negation of the habit of praying.
{laarib 'beating' are possible. In contrast, a verb like salima 'to (be) safe'
yields both forms: both saalim and saliim mean 'safe' in English, but there
is a subtle difference in meaning between the two, to which I shall return.
It is sufficient to note at this point that there are distributional differences
and gaps in the occurrence of participial and adjectival forms. Under optimal
conditions, we expect these differences and gaps to follow from a princi-
pled account. In English, the dividing line between prototypical adjectives
and less prototypical ones is between basic and derived forms, but in Arabic
the dividing line is to be characterized in other terms. Aspect appears to
be the natural base for drawing the right distinctions.
(107) a. e --* s.
b. s--* s.
Moreover, I assume (following Hale and Keyser, 1991) that e is the notional
type associated with V, whereas s is associated with A. Consequently, only
participles are deverbal, whereas pure adjectives are not.
So far so good. There are participle forms, however, that can be easily
classified as stative. The latter include intransitive participles like ~aaliiJ
'good, right' ,faasid 'bad, corrupting', tjaa?iq 'narrowing, confined', (aahir
'clean, pure', baa(il 'false', ?aamin 'safe, secure', etc. There are also
transitive stative forms expressing cognition (caarif or caalim 'knowing',
jaahil 'ignoring' ,faahim 'understanding') or perception (saamic 'hearing'),
affective verbs (kaarih 'hating', mu!Jibb 'loving'), etc. These participles
derive from predicates that can be classified as stative by using some
classical tests. 33 They cannot be conjugated in the imperative (* #u/:1 'be
right!', *fsud 'be wrong!', *crif 'know!'). They do not occur with manner
adverbs such as camdan 'deliberately' (*yacrifu l-jawaaba camdan 'he knows
the answer deliberately'). They do not serve as complements of control verbs
like ?aqnaca 'to persuade' (*?aqnactuhu bi-?an yacrifa l-jawaaba 'I
persuaded him to know the answer'). They do not appear in pseudo-cleft
constructions (*ma facala huwa ?annahu ~alul:za, carafa l-jawaaba 'What
he did is be right', 'What he did is know the answer').
It is clear, however, that these operations are intended to test agency
(or volitional control), rather than only stativity: the imperative requires
volition; the subordination to some predicates requires an agent control-
ling the event; some manner adverbs require agency, etc. The do test in
pseudo-clefts assesses agency in activities. Agency implies that a
situation is dynamic, but agency is not the only test for dynamicity. The
progressive can test dynamicity of situations which are not (necessarily)
actions or activities. For example, the sentence in (108) describes a dynamic
situation, yet there is no implied agency here:
(108) d-damc-u haamir-un
the-tear-nom flowing-nom
Tears are flowing.
is that Zayd has enjoyed the position or state of being sitting before I entered
the room (a result state reading). The construction (114a) can also have
this interpretation, but crucially, it has a simultaneous process reading.
That is, Zayd's sitting was taking place during the interval of my entering.
On the other hand, present forms denote habitual or iterative meanings,
but participles do not. Compare the following pair:
(115) a. kaana r-rajul-u y-aftaJ:!-u faa-hu kull-a
was the-man-nom 3-open-indic mouth-his every-ace
marrat-in
time-gen
The man was opening (used to open) his mouth every time.
b. kaana r-rajulu faatiJ:!-an faa-hu
was the-man-nom opening-ace mouth-his
(*kulla marrat-in)
every-ace time-gen
The man was (in the state of) opening his mouth.
Similarly, present forms of passive verbs can express a generic middle
meaning, but passive participles cannot. Thus the passive participle in (116)
can only have an actual non-generic interpretation, although the present form
of the passive in ( 117) does have a middle generic reading:
(116) s-saay-u masruub-un bi-duun-i sukkar-in
the-tea-nom drunk-nom without sugar-gen
The tea is (now) drunk without sugar.
(117) y-usrab-u s-saay-u bi-duun-i sukkar-in
3-pass.drink-s.m. the-tea-nom without sugar-gen
The tea is (usually) drunk without sugar.
This contrast can be explained if we assume that present forms (by virtue
of carrying an unmarked T) can have two readings: (a) a temporal
(non-past) reading which becomes explicit if one uses a temporal adverb
like l-?aana 'now' or gadan 'tomorrow' and (b) an atemporal, habitual,
or generic reading which is not located in any particular time, and which
arises when an unspecified Tense is bound (by default) by a habitual
operator. On the other hand, the non-availability of a generic interpreta-
tion for participles follows from the absence of T in their inflection. 36
This view is further supported by the fact that passive participles cannot
be iterative or habitual, unlike present passives. Consider the following
pair of sentences:
184 CHAPTER 4
Consider once again such pure adjectives as hasan, tawiil, labyafj, etc.
The formation of these adjectives on one pattern or another is, if not
idiosyncratic, limited to some semantic/conceptual class. Suppose that these
subregularities (and idiosyncrasies) are lexical, and that the formation of
these adjectives takes place in the lexicon. Suppose moreover, that the
non-contingent requirement imposed on their formation is a property of both
the lexical root and the affix. In prototypical cases, the lexical root and
the affix match in being non-contingent states. This can be symbolized as
in (120):
(120) s+s~s
present in sentences. For example, there are adverbials that determine the
telic/atelic property of processes, i.e. whether they may come to a termi-
nating point (or end) or not. Adjectives are normally atelic, and this can
be attributed from their non-contingent nature. Consequently, they are
incompatible with telic adverbs. This is illustrated by the following contrast:
(122) katab-tu d-dars-a fii saacat-in
wrote-/ the-lesson-ace in hour-gen
I have written the lesson in an hour.
(123) * ?anaa marii9-un fii saacat-in
I sick-nom in hour-gen
*I am sick in an hour.
The construction (123) is ungrammatical because the temporally bounded
PP there is incompatible with the non-bounded nature of the adjective.
Let us turn now to participles. Suppose that the contingency require-
ment is (normally) met by both the lexical root and the participle affix. I
will represent contingency by sc. Moreover, I assume that verbs are
necessarily associated either with e or sc, and that participles are also e's
or so's (at some point in the derivation), by virtue of being verbs. Verbs
usually classified as stative like fahima 'understand' or carafa 'know' are
associated with sc, and hence eligible to participle formation in Arabic.
Participle formation, which I assume to be syntactic, can be symbolized
as follows:
(124) Sc + Sc ~ Sc
All languages (as far as I know) allow participle formation from e's. They
differ (quite fuzzily) in whether they allow total or only partial (or non
standard) participle formation from s/s.
If this is true, then some problems remain to be solved. One of them is
how to prevent participle formation from verbs like marirja, fJasuna, taala,
etc., i.e. how to preclude forms like *maarirj, *l}aasin, *taawil, etc. Recall
that these verbs have inchoative readings, i.e. they do not normally have the
pure stative reading, and can thus qualify as sc's. These verbs differ,
however, from other statives, in that their contingency value (unlike other
verbs) can be thought of as derived, rather than basic. One informal way
to put it is to assume that these verbs originate as adjectives, while others
(like fahima) originate as verbs. This can be represented as follows:
(125) mariga: A (= s) ~ V (= sc)
( 126) fahima: V ( sJ =
Suppose this is true, then a generalization about participle formation
emerges: verbs which originate as adjectives cannot form participles. This
186 CHAPTER 4
This section will be devoted to two main questions. In the first part I discuss
further the content of the notion of finiteness, which is needed to charac-
terize languages. Finiteness has been thought in recent literature as being
a property of Tense, and finite Tense (equated with[± Past]) has been treated
as a (syntactic) operator, binding a variable (left by V raising, as in Pollock,
1989a). This syntactico-semantic view of finiteness is shown to be less
TEMPORAL, ASPECTUAL, AND MODAL CATEGORIES 191
4.1. Finiteness
The appropriate notion of finiteness to characterize language differences
is not straightforward. Morphological, syntactic, and semantic criteria have
been used in the literature to characterize finiteness. These criteria are not
of the same nature, which often leads to confusion and inconsistency.
Quirk et al. (1985), for example, provide the following description for finite
verb phrases:
(137) (a) Finite verb phrases can occur in independent clauses.
(b) They have tense contrast, i.e. distinction between present
and past tenses.
(c) There is person and number concord between the subject
of the clause and the finite verb phrase.
(d) They contain a finite verb which may be an operator or a
simple present or past form.
(e) They have mood (which indicates the factual, nonfactual,
or counterfactual status of the predication).
These criteria are of various kinds, and they do not define a unified notion
of finiteness.
In languages like English and French, verbs are usually classified as finite
or non-finite depending on their inflectional declensions. The latter include
agreement, tense, and mood. Finite morphology is often thought of as
being correlated with what has been called finite T, or so-called absolute
(or referential) T (basically [± Past]). In contrast, non-finite morphology
expresses either Aspect (with participles) or non-finite T (in infinitives
(see Stowell, 1983). In other words, there is a one-to-one correspondence
between [± finite] morphology and [± absolute] T.
Even if this picture can be defended for English and French (we will
see that it cannot), it is hardly adequate for Arabic. In this language the
finite morphological form of the verb covers both contexts. As explained
earlier, a past perfect tense, for example, is expressed in English by using
a finite auxiliary followed by a non-finite participial form of the lexical
verb, as in (138):
(138) He had eaten.
But a sequence of two finite verbs is used in Arabic, as in (139):
192 CHAPTER 4
is 'a past of the past', a fact totally transparent in the Arabic case. The
difference between the two languages is then a matter of variation in
'spelling out' the embedded past under modals or auxiliary verbs, by making
use of either finite or non-finite morphology.
In Fassi (1989c), I have taken T expressions to be (basically) referen-
tial verbal expressions, parallelling D (determiner) expressions, which are
referential nominal expressions. I have adopted Higginbotham's (1985)
proposal that saturation is a condition on referentiality. A constituent is
saturated iff every role (or position) in the grid associated with its head
is saturated. If NPs and VPs have a Davidsonian E position in their grids,
then they are necessarily not saturated, and hence cannot be referential.
Higginbotham assumes that E in « ... the thematic grid of the verb is
discharged at the point where VP meets lnfl. The interpretation is existential
generalization over the £-position, as in Davidson (1966) . . . ». The
discharge of E in V is operated via the mechanism of theta-binding, T being
the binder, like D in nominals (see Chapter 5).
It is reasonable to think that T expressions bear a referential index
(R-index), in the same way that D expressions do. In a fairly standard GB
view, R-indices are assigned to nominal expressions by a predicate (see
e.g. Stowell, 1981 ). If this is so, then the R-index is assigned to the maximal
functional projection, and inherited by the functional head. It might be
that R-indices can be assigned freely to verbal expressions (IPs or CPs),
especially in matrix clauses. T, being an operator, is binding an E position
in the argument grid of V (its natural 'scope bearer', as in En~t. 1987). I
assume that only a referential T (i.e. a T bearing an R-index) can dis-
charge the E position of the verbal expression, thus saturating the expression.
Referential T expressions might be compared to pronouns, being deictic
or anaphoric, as in Partee (1973, 1984). 42 There are also cases of anaphoric
Ts, as in subjunctive Tense. The latter are assumed to have no R-index,
and have to look for an antecedent (to inherit an R-index from it) if it is
to refer. Given this framework, it is possible to define a local or 'autonomous
domain' (CP?), and to associate with it only one referential tense (as in En~t.
1987, or Bennis and Hoekstra, 1988)," ... which is interpreted as a function
of the time of utterance". Otherwise it is 'anaphoric', i.e. included in the
domain of an R-tense (see En~t. 1987 on anchoring conditions for tenses). 43
Inspired by this view, can we think of finiteness as an S-structure and/or
LF phenomenon (as in Pollock, 1989a), rather than a morphological property
(as I have proposed earlier)? Pollock argues that V raising (in languages
like French) is necessary at S-structure, to enable finite T (taken to be a
syntactic operator) to bind the variable left by V (see also Platzack and
Homlberg, 1990, who extend this idea to Germanic). If finiteness is a
semantico-syntactic phenomenon, then a sequence of two finite tenses in
the same local domain (as in the Arabic case) will have to be ruled out
by quantification theory. This is so given that there can be only one finite
194 CHAPTER 4
T in, for example, auxiliary structures because there is only one possible
E variable (which is created by the movement of the thematic verb; see
Pollock, 1989a for details). 44 But there is no evidence supporting the
uniqueness requirement on tenses. On the contrary, the model of tenses
which I have argued for earlier is binary, and a sequence of two tenses is
expected to be grammatical by the theory. Consequently, finiteness has to
be construed as morphological, rather than logical.
Since finiteness is morphological, crosslinguistic variation in its mani-
festation is expected. Thus the fact that some languages (like Arabic) allow
two finite tenses in the same local domain, while others (like English) do
not, is attributed to a parametric property of I (i.e. nominality). Such a
variation is not generally aknowledged at LF.
The construction (147), in which both objects have been omitted, is ungram-
matical:
(147) * zayd-un maanii)-un
Zayd-nom giving-nom
* Zayd is giving.
The construction (147) can be ruled out as a violation of the Projection
Principle, because internal arguments have not been projected. Thus (147)
is excluded for the same reason that the verbal construction (148) is:
(148) * manal)a zayd-un
gave Zayd-nom
*Zayd gave.
There are other [aa-i] adjectival forms, however, which do not seem to
preserve the thematic properties of the verbal base. Consider the following
pair of constructions:
(149) zayd-un naafic-u-n ?abaa-hu
Zayd-nom helping-nom father-his
Zayd is helping his father.
(150) zayd-un naafic-un
Zayd-nom helping-nom
Zayd is helping.
In (150), the adjective has no object, whereas it does have one in (149).
Yet (150) is well-formed. Other adjectival forms which can be used
intransitively (as well as transitively) are: jaahil (ignoring), caalim
(knowing), tjaalim (doing wrong), etc. They derive normally from stative
or experiencer verbs. The question then is: why is (147) excluded, whereas
(150) is not?
I have observed that the verbal source for the pair (146) and (147) is
unique, and that the syntactic alteration of the thematic properties is not
allowed by the Projection Principle. But the situation of (150) is different.
In this case, the adjective either has a different verbal source from that in
(149), or if the source is unique, the alteration of the valency of the verb
in (150) is allowed. If the latter option is correct, it suggests that the
affixation in (150) occurs in the lexicon, and the adjective (which is formed
there) is projected with no arguments in the syntax. In contrast, the adjec-
tive in (149) is verbal, and projected with a complete theta-grid in the syntax.
As it turns out, there are two different verbal sources for (149) and (150),
and no alteration of their grids to derive participial forms is necessary. These
verbs are exemplified in the following pair:
196 CHAPTER 4
b. ?a marii9-un zayd-un
Q sick-nom Zayd-nom
Is Zayd sick?
This corroborates the view that adjectives which occur in these construc-
tions must be describing an inherent property of the subject or the modifying
complement. Participles are excluded from this context because they do
not denote properties, but only actions or processes.
To sum up, I have shown that intransitive [aa-i] adjectives are ambiguous
between a participle structure, in which they are Vs at pre-Xo syntactic level,
and a pure adjective structure, in which they are just As at that level. Each
category has Case, thematic, and aspectual properties that are compatible
with the analysis adopted. The scope of affixation accounts for the depth
of internal verbality or its absence, in the syntax or the lexicon. But at
the external level, participles are APs. In the next subsection I argue that
adjectives and participles are essentially of stative (aspectual) nature.
Abney adds that " ... the semantics of adjectives [is] similar to that of
mass nouns: tall denotes a certain quantity of tallness, in the way that rice
denotes a quantity of rice" (p. 308). According to him, all APs are headed
by Deg, be it lexically filled or empty.
Abney himself notes some problems with the generalized Deg hypoth-
esis. For example, some adjectives, like those in (170), resist all degree
words:
(170) Everyone here [AP tested for drugs] has come up negative
a.* as tested for drugs as anyone else
b. *too tested for drugs
c.* more tested for drugs than me
These adjectives, he observes, rather consistently appear to be participles,
although he could not come up with an explanation for this incompati-
bility (p. 303). But given that participles are APs in his system (as they
are in ours), the problem is even more dramatic. If participles and pure
adjectives are unified, then Abney's approach to the problem raised is not
consistent.
The Deg hypothesis might be applicable to some Arabic gradable
adjective phrases, but it is not obvious how it can generalize to bare adjec-
tives (even when they are potentially gradable). Consider the following
adjectival constructions:
NOTES:
* This chapter is a revised version of Fassi (1989c), as well as Chapter V of Fassi (1989b).
Part of this material was presented in the form of various lectures at the University of Paris
VIII and the CNRS (Fall 1989 and Spring 1990), SOAS (London) and the University of Essex
(Spring 1990), the International Linguistic Institute of LSM at the University of Rabat
(Summer 1990), as well as UQAM (Montreal) and MIT (Fall 1990). I would like to thank
the audiences at these talks for stimulating remarks.
1 To my knowledge, this description of the Arabic temporal system, which has become quite
standard, is originally due to Caspari, whose grammar was translated by W. Wright in 1859,
and has subsequently undergone a number of revisions. Reckendoff (1898) adopts basically
the same approach. Among similar later developments is Cohen's (1924) monograph, based
on ideas contained in Meillet (1917).
2 See Meillet (1917) for this terminology.
3 In French, for example, the so-called imparfait expresses an imperfective past, the passe
simple a perfective past, etc.
4 For details about verbal morphology, see Wright (1974), McCarthy (1979), Fassi (1984)
and Chapters 2 and 3. For Voice morphology, see Fassi (1988b).
5 For various approaches to dependent and narrative tenses, see Partee (1973, 1984), Kamp
(1979), Kamp and Rohrer (1989), Lo Cascio (1985), and Bartsch (1989). On the depen-
dence of tenses, their scope, and various syntactic mechanisms in GB, see Higginbotham
(1985), Enc; (1987), Pollock (1989a), Picallo (1984), Jakubowicz (1985), Raposo (1985-1986),
Zagona (1988), Gueron and Hoekstra (1988), Bennis and Hoekstra (1988), and Rouveret
(1989), among others.
6 The French terminology accompli/inaccompli is used to designate an interrupted or
uninterrupted process. It is often equated with the perfect/imperfect distinction. See Marouzeau
(1951) and Wright (1974), among other references.
7 Comrie (1976) proposes a 'relative tense' approach to the Arabic temporal system which
is intuitively close to ours, although the terminology of perfective/imperfective he uses to
designate the two members of the opposition is confusing. Cohen (1989), on the other hand,
analyzes only the two forms of lexical verbs (katab/yaktub), without taking into account
TEMPORAL, ASPECTUAL, AND MODAL CATEGORIES 205
auxiliaries, and concludes that the opposition in only aspectual, in terms of accompli!
inaccompli, i.e. the equivalent to perfect/imperfect distinction. Moreover, he overlooked
the importance of the temporal adverb test used by early Arabic grammarians, as well as
that of the contrasts pointed out by Cohen (1924).
Anteriority, or time reference, is conceptually closer to absolute Tense than to the ('genuine'
aspectual) category of perfectivity/imperfectivity. The latter denotes the internal time of the
event, and can be separately encoded in verbal morphology, as in Slavic or Greek. It is not
concerned with external temporal location, nor with ordering events.
8 See Wright (1974) for details, as well as Section 2 below. Other moods discussed in the
traditional literature are: the conditional, the imperative, and the energetic. The conditional
has the same mark as the jussive, as illustated in (i), the energetic is manifested by adding
a [n] suffix (geminated or not) to the verb, as in (ii):
(i) ? in t-adxul-!ll 1-qaa'ata
if you-enter the room-ace
(ii) la-t-adxul-u-nna
will-you-enter-2.pl. -indeed
You will enter indeed.
As for the imperative, it loses its prefix mark (encoding PERS normally), at least when it
is declarative:
(iii) dxul-!ll 1-qaa'ata
enter the room-ace
Enter the room!
Note that there is no Mood marking with the past form, as exemplified by (iv):
(iv) katab-tu
wrote-/
I have written.
This distribution of Mood can be captured by redundancy rules like the following:
(v) Perfect/Past ~ Ill M
(vi) Imperfect/Nonpast ~ +M
9 Note that the present form of the copula has also 'pure' modal uses, as in the following
construction:
(i) ? -akuun-u sa'iid-an bi-liqaa? -i-ka
1-am-indic happy-ace with-meeting-gen-you
I would be happy to meet you.
The politeness use here forces the occurrence of the copula. The simple present reading obtains
without a copula, as in (ii):
(ii) ? anaa sa'iid-un bi-Iiqaa? -i-ka
I happy-nom with-meeting-gen-you
I am happy to meet you.
10 The INFL split view has been championed by Pollock (1989a). The bi-inflectional view
is tacitly assumed in a large literature. It is explicitly proposed in e.g. Fassi ( 1987 a) and
Dobrovie-Sorin (1991). A more neutral position is found in Kayne (1989). Chomsky (1991)
proposes a bi-AGR structure (designated by AGRS and AGRO), but he does not seem to
assume two temporal projections within the same clause.
206 CHAPTER 4
11 See Laka (1990) for arguments that Neg must be c-commanded by Tense. The presence
of Neg might then be seen as a diagnosis for the presence of Tense.
12 A traditional answer to the ungrammaticality of (41) or (42) has been often posed in terms
of auxiliary selection, as stated in (i):
(i) An auxiliary verb selects a VP.
Gueron and Hoekstra (1988), for example, assume the following:
(ii) An auxiliary verb governs a verbal projection [a VP; FF], a non-auxiliary verb
governs either a nominal projection or a verbal projection.
This solution is inadequate, since it does not carry over to Arabic. In very recent unpub-
lished work, these authors reach conclusions about auxiliary selection which are similar to
mine.
13 On Temporal marking, see in particular Zagona ( 1988), who proposes the following
definition (p. 94):
(i) A Temporally-marks B if A assigns a Temporal role (S, R, or E) to Bas a lexical
category.
Note, on the other hand, that if English auxiliary structures are bi-AGR structures a La
Chomsky (1991), then the question arises as to why these structures are not nominal, as in
the Arabic case. Auxiliary verbs in English or French appear to be Tense markers, rather
than Case markers. For a functional definition of categories which is compatible with my view
of auxiliaries, see Gueron and Hoekstra (1988).
14 Joan Maling pointed out to me that the fact that copular verbs assign Case to predicate
adjectives or nouns appears to be intriguing, especially from the point of view of Indo-
European (IE) languages, where the case of the NP subject spreads to the predicate. This
difference suggests that the Arabic copula is more 'verbal' than its IE counterpart. At this
point, I have no convincing analysis of why case spreading operates in some languages,
but not others. On case spreading, see Yip, Jackendoff, and Maling (1987), as well as
Maling (1991).
15 A number of generative studies, starting with Chomsky (1957), and including Emonds
(1976, 1978), and Akmajian, Steele, and Wasow (1979), among others, have argued that modal
auxiliaries in English are not verbs, but rather realizations of the category AUX (taken in
Emonds, 1985, to be an inflectional Spec of V). Their arguments center around the idea
that English modals do not undergo verbal rules (see Emonds, 1985, pp. 210-213 for a
recent illuminating discussion; see also Emonds, 1976, pp. 205-211, as well as references
in Heny and Richards, 1983). Other studies have maintained the verbal analysis of modals,
even for English. For example, Schachter (1983) questions the thesis " ... that modals lack
any [verbal] inflections whatsoever" and he argues that " ... although modals do not inflect
for number, there is evidence that they do inflect for tense" (p. 147). For the problematic
nature of the AUX analysis of modals, see below, in particular the fact that the AUX analysis
of modals is not general enough to account for crosslinguistic variation in expressing
modalities.
16 For an attempt to unify the semantics of modal qad, see Dahl and Talmoudi (1987).
17 It is possible that the finite form of the verb there is unspecified with respect to the
value of anteriority. This neutral value of the form is further corroborated by its use in
command or prohibitive statements:
(i) li-y-adxul
let-3-enter
Let he enter.
TEMPORAL, ASPECTUAL, AND MODAL CATEGORIES 207
cializes' in some category (N or V), iff it governs (and/or selects) only that category, not
another. Thus modal Neg specializes in V, copular Neg in N (or A), and neutral Neg does
not specialize, since it occurs with both. By virtue of this relational property, only the first
two types of Negs are Case assigners, and proper governors. Note that this notion is less
stipulative than selection. Baker (1988) defines selection as follows:
A selects B iff:
- (i) A assigns a theta role to B or
- (ii) A is of category C and B is its IP
- (iii) A is of category I and B is its VP.
23 For further motivation of the notion g-projection, see Kayne (1983) and Pesetsky (1982).
Note that traditional grammar distinguishes 'strong' head governors (like verbs), with can
govern to their left (in addition to their right) and 'weak' governors, with cannot (preposi-
tions, Ns, etc.).
An alternative to the g-projection analysis is to assume that the verbal head is incorpo-
rated into the other heads (including Neg) at LF, and that Case and government are checked
at LF via Baker's (1988) Government Transparency Corollary, stated in (i):
(i) A lexical category which has an item incorporated into it governs everything
which the incorporated item governed in its structural original position.
24 See Ouhalla (1990) and Benmamoun (1990) for such claims.
25 This restriction is further corroborated by the fact that maa is not generally compatible
with future interpretation. When it occurs with the non-anterior form of the verb, the inter-
pretation is limited to the present tense, as shown by its incompatibilty with future adverbs:
(i) maa ? -aakul-u (1-? aana, *gad-an)
not 1-eat-indic (now, *tomorrow)
I do not eat (now, *tomorrow).
Other differences aside, this behaviour is shared by laysa. Moreover, the latter does not
occur with the anterior form of the verb, and its interpretation is necessarily non-anterior
present, as shown by the following contrasts:
(ii) lay sa r-rajul-u y-a? kul-u
not the-man-nom 3-eat-indic
The man does not eat (now).
(iii) *lay sa r-rajul-u ?akal-a
not the-man-nom ate-3.s.m.
(iv) * laysa r-rajul-u y-a? kul-u gadan
not the-man-nom 3-eat-indic tomorrow
26 Another difference between the two Negs is illustrated by the fact that one is compat-
ible with conditionals, but the other is not:
(i) ? in lam y-a?ti
if not.past 3-come
If he does not come.
(ii) *?in lammaa y-a?ti
if not.yet 3-come
Furthermore, lam is compatible with the modal qad, which expresses uncertainty and expec-
tation, whereas lammaa is not. These differences can be accounted for only in a general theory
of illocutionary force, which must distinguish 'assertive' Negs from (potentially) hypothet-
ical Negs.
TEMPORAL, ASPECTUAL, AND MODAL CATEGORIES 209
27 The use of modal particles also affects the interpretation of reference time denoted by
verbal inflection. For example, past/perfect forms are used with conditional particles to express
anteriority in a hypothetical situation in the future, not in the past, as exemplified in (i):
(i) ? in qum-ta la-?-aquuman-na
If stood-you emph.-1-stand-emph.
If you stand up, I will certainly stand up.
It is equally used in so-called 'instigative' (ta}Jr.iiirj) sentences like (ii):
(ii) hallaa qum-ta
why.not stood-you
Why don't you stand up.
In all these modal conditioned uses, and others, the anterior form is said to emphasize the
asserted truth of the event, or its realization. But clearly the deictic T cannot be past. The
tense borne by the verbal inflection is interpreted only within the context of the modal,
which is irrealistic.
In the same vein, non-anterior forms can be used with modalities of wishes, as in (iii},
or with conditionals, as in (iv):
(iii) layta r-rajul-a y-a?kul-u
may the-man-ace 3-eat-indic
May the man eat.
(iv) ? in t-ac;Jhab ? -adhab
if you-go /-go
If you go I go.
Third, it is well-known that tenses have modal uses. For example, past tense often denotes
a hypothetical situation as in the following examples:
(v) I wish I had a car.
(vi) Deux minutes de plus et !'OM gagnait (!'OM n'a pas gagne).
These uses and interactions provide additional support for the three-valued TMA view of
temporal inflection adopted.
28 I use the traditional term 'adjective' (abbreviated as A) to designate the category
discussed. In Chomsky (1974) and (1981), A is replaced by a feature matrix, i.e. [+N, +V].
Abney (1987), on the other hand, has proposed that the traditional AP is headed by a Deg(ree)
P. See Section 4 for criticism of these views.
29 Radford (personal communication) has pointed to me that English has sentences like
the following:
(i) He sicked up his food.
(ii) Her hair yellowed in the sun.
Clearly, the readings here are processive and inchoative, respectively, and cannot be pure
stative. Moreover, it is natural to think that these verbs are derived from adjectives, rather
than basic. See below for similar cases in Arabic.
30 On properties of English adjectives and participles, see Fabb (1984), Abney (1987), Borer
(1984), Sproat (1985b), and Levin and Rapaport (1986), among others.
31 There are other forms of adjectives which are derived from triliteral roots: IJasan 'nice',
?abyad 'white', fari/J 'glad', and <a.tsaan 'thirsty' are instances of the most productive
forms. The latter are often predictible from the meaning of the root, though not always.
210 CHAPTER 4
32 This is true of triliteral roots only. With non-triliteral roots, there is no counterpart to
(a), and deverbal formation is limited only to regular participle forms.
33 Some philologists have proposed that participles are freely formed from transitives, but
are subject to a non-stative constraint when intransitives (see e.g. Wright, I, pp. 131-132).
This is incorrect, however. On the one hand, there are intransitive statives which form good
participles (~alu}Ja 'to be/become good', batula 'to be/become false', etc.). On the other hand,
there are transitive statives which do not form participles freely. For example, the transi-
tive ?asbaha 'to resemble' does not form the passive participle *musbah 'resembled'.
34 Some grammarians claim, contrary to fact, that constructions like (ll2a) are ill-formed.
For example, l:fasan, III, p. 281, rejects the following example:
(i) 1-mutasaabiq-u batii? -u 1-l}arakat-i I-? aana
the-runner-nom low-nom the-movement-gen now
The runner is slow now.
l:fasan observes that the adjective expresses an inherent property or a permanent state which is
not construed as being true only in the present, the past, or the future. This property has to be
true at all times at once, not at one particular time. Given this view, the ungrammaticality of (i)
is expected. In contrast, the participle can express a meaning only at a definite point of time
reference, hence the grammaticality of (110), for instance (p. 295). I think that f:lasan's position
arises from a confusion between the 'inherent' aspectual properties of the adjective or the
participle (the former being stative whereas the latter is not) and the temporal properties of the
construction (which is interpreted as past or non-past). f:lasan himself quotes ~abbaan, who
argues that the adjective can occur in a context in which it indicates only past, present, or
future, as in the following examples (ibid., p. 280, n. 2):
(ii) kaana zayd-un l}asan-an fa-qabul}a
was Zayd-nom good-ace then-became.bad
Zayd was good and then became bad.
(iii) sa-y~iiru l}asan-an
will-become good-ace
He will become good.
(iv) huwa 1-? aana l}asan-un
he now good-nom
He is now good.
Clearly, there is no incompatibility here between the inherent aspectual property of the
adjective and the temporal specification of the construction. If l:fasan's objection were true,
then all adjectives would have been individual level predicates in the sense of Kratzer
(1989), but this is obviously not true.
35 Note that the contrast between participles and adjectives cannot be construed in terms
of durativity. Participles can express continuous or durative aspect. Consider the following
sentence:
(i) <arnr-un <Jaarib-un zayd-an
Amr-nom beating-nom Zayd-acc
Arnr is beating Zayd.
In this construction, the process does not come to an end; it is continuous. The participle is
also understood as progressive. If the progressive is made of non-stativeness and continu-
ousness (or durativity), then the only distinction involved is between something homogeneous
throughout its duration, and something in which there is a breaking point (see Comrie,
1976).
TEMPORAL, ASPECTUAL, AND MODAL CATEGORIES 211
36 En~ (1991) proposes that there is no present tense form in English, but clearly these
differences cannot be accounted for if there is no such a form.
There are further differences between process and state verbs. For example, the former,
but not the latter, exhibit a difference in interpretation depending on whether the order of
the sentence is VSO or SVO. In VSO, the verb is neutrally interpreted as habitual:
(i) y-a? kul-u r-rajulu
3-eat-indic the-man-nom
The man eats.
In SVO, the sentence can be ambiguous, but it is neutrally interpreted as located in the present:
(ii) r-rajul-u y-a? kul-u
the-man-nom 3-eat-indic
The man is eating.
The construction (i) can be interpreted as specific present only in some marked contexts
like reportive use, or when a present adverbial shifts the orientation point appropriately. Stative
verbs, by contrast, do not seem to have a habitual reading, nor are they sensitive to order
differences in the clause.
37 For various arguments supporting the dual aspect-tense view of temporal systems, see
Smith (1991), Descl~s (1989), Guentcheva (1990), and also Dahl (1985) and Comrie (1976),
among others.
38 As is well known, the active participle has different forms, depending on whether it is
bound to a triliteral consonantal root, or to a non-triliteral root. With a triliteral, it is [aa-i],
with a non-triliteral, it is [u-a-i]. This is exemplified in (i) and (ii), respectively:
(i) zayd-un qaatil-un s-saJun-a
Zayd-nom killing-nom the-prisoner-ace
Zayd is killing the prisoner.
213
214 CHAPTER 5
the other hand, some languages (like Arabic) have obligatory N raising
(to D) at S-structure, due the affixal nature of D, whereas others (like
English) do not, resulting in different orders in nominal phrases (NSO vs.
SNO, respectively).
The properties of analytic genitives interact with those of synthetic
genitives in interesting ways, but the two strategies are different. For
example, analytic genitives exhibit no complementary distribution between
the possessor and the article, and no agreement in Definiteness between
the possessor and the possessee. They allow NP raising and extraction of
NPs only when D and AGR are rich, echoing the requirements on NP raising
and extraction of sentential subjects. These properties (and others) are
missing in Romance, because the latter (almost exclusively) use the analytic
(or prepositional) strategy. Analysis is available in Arabic and English as
well, but the range of its use is limited. Part of my concern is to derive
the various outcomes of the interaction of the two strategies across
languages. It is argued that synthetic genitives are regulated by Spec-Head
agreement principles (in Spec-oriented languages or constructions) applying
to AGR or D projections (and categories), whereas analytic genitives are
regulated by Head-Comp licensing principles (in Camp-oriented languages
or constructions).
The problem of theta-marking possessors is also investigated. It is argued
that the possessor is not generated in Spec of NP, but in Spec of Poss.
Poss is hypothesized to be a functional licenser for possessors. Similarly,
the properties of deverbal process nominals (called mCJ.$dars in the Arabic
tradition) are examined, as are those of subject nominals (the latter are closer
to, but not identical with, agent nominals in English, as we will see). In
order to derive the 'mixed' categorial nature of ma~dars (the fact that they
are internally verbal, and externally nominal), and their categorial differ-
ences across constructions, it is proposed that they are formed in the syntax.
A nominalizing affixation process converts the verbal root to N, at
different stages of the derivation. On the other hand, the thematic proper-
ties of process and subject nominals are shown to be problematic for the
current version of the Projection Principle (which is too strong). A weaker
version of this principle is proposed.
This chapter is organized as follows. In the first section I investigate N
and NP raisings in ordinary noun phrases, and especially in synthetic
genitives. I establish and discuss a number of 'complementary distributions'
in the DP system. I also examine the contexts of genitive Case assign-
ment, and the role played by AGR and D in Case and order licensing. I
close the section with a typology of AGR and D properties. In Section 2
I turn to the treatment of ma~dars, how they are derived, and how their
thematic structure is satisfied by making use of Higginbotham's (1985)
mechanism of theta identification. In Section 3, I investigate the preposi-
tional strategy used thoroughly in Romance, but also in Arabic and English.
INFLECTIONAL PROJECTIONS IN NOUN PHRASES 215
(3) DP0
~
D NP
11 D~~ N DP2
L___ _ _ ____JI
Arabic nominal phrases exhibit a number of properties that appear (at least
superficially) to be in total complementary distribution. As we have just
seen, one of the combinatorial properties of nouns is their ability to occur
with the definite article. But note that the article is missing in the indefi-
nite NP in (4):
(4) daxal-tu daar-a-n
entered-/ house-acc-n
I entered a house.
The noun here carries no (indefinite) article, although it carries another
affixal form which is suffixed to the noun (but is missing in (1)). It is the
[n] form which is termed tanwiin in the Arabic tradition, and nunation by
Western philologists. The nature of nunation has been (and is still) a real
puzzle for Arabic grammarians. Most of them treat this form in (4) as an
indefinite article, but this view is hardly tenable. 4 It is true that common
singular nouns carrying the article lack this form, as ( 1) indicates, but proper
nouns may carry it, as illustrated by (5):
(5) a. hind-u-n
Hind-nom-n
Hind.
INFLECTIONAL PROJECTIONS IN NOUN PHRASES 217
b. mul).ammad-u-n
Muhammad-nom-n
Muhammad.
On the other hand, definite dual or plural common nouns carry both the
article and the [n] form, as shown by (6): 5
(6) a. r-rajul-aa-n
the-man-dual-n
The two men.
b. 1-muslim-uu-n
the-moslem-pl. -n
The moslems.
These contrasts indicate that both nunation and the article can occur on
definite nouns, but only nunation occurs on indefinite nouns (and some
proper nouns). Moreover, indefinite nouns have no (overt) article.
On the other hand, genitive NP complements can carry neither nunation
nor the article, as illustrated by (7):
(7) daxal-tu daar-a r-rajul-i
entered-/ house-ace the-man-gen
I entered the man's house.
These restrictions on genitive possessive constructions are further illustrated
by the ungrammaticality of the following constructions:
(8) a.* daxal-tu d-daar-a r-rajul-i
entered-/ the-house-ace the-man-gen
I entered the house of the man.
b.* daxal-tu daar-a-n r-rajul-i
entered-/ house-acc-n the-man-gen
I entered a house of the man.
In these constructions the head noun can bear neither the article nor
nunation, be it definite (as in (8a), or indefinite, as in (8b)). That the head
noun loses both combinatorial possibilities can be seen more explicitly when
a noun like that in (6b) is constructed with a genitive, as in (9):
(9) muslim-uu 1-madiinat-i
moslem-nom.pl. the-city-gen
The Moslems of the city.
Note that the lack of articles is a property of nominal genitive con-
structions only, and does not extend to adjectives heading genitive
218 CHAPTER 5
These observations point to the conclusion that the head noun inherits the
same value of definiteness that the genitive noun has. When the genitive
NP is indefinite, the head N is also indefinite, and the modifying adjec-
tive cannot bear the article, as the contrast in (15) shows:
(15) a. daxal-tu daar-a rajul-i-n waasicat-a-n
entered-/ house-ace man-gen-n large-acc-n
I entered a large house of a man.
b.* daxal-tu daar-a rajul-i-n 1-waasicat-a
entered-/ house-ace man-gen-n the-large-ace
I entered the large house of a man.
The facts of relative clause modification corroborate this view. When the
head noun of a genitive construction is modified by a relative clause, the
relative complementizer (which is inherently definite) may or may not
head the relative clause, depending on whether the head noun is definite
or not, as in the following examples:
(16) a. daxal-tu daar-a r-rajul-i 1-latii l}.taraq-at
entered-/ house-ace the-man-gen the-that-f. bumed-3.fs.
I entered the man's house which burned.
b. daxal-tu daar-a rajul-i-n l}.taraq-at
entered-/ house-ace man-gen-n bumed-3.fs.
I entered a man's house (which) burned.
c.* daxal-tu daar-a r-rajul-i l}.taraq-at
entered-/ house-ace the-man-gen bumed-3.fs.
d.* daxal-tu daar-a rajul-i-n 1-latii l}.taraq-at
entered-/ house-ace man-gen-n the-that-! bumed-3.fs.
The relative complementizer in these constructions behaves like the article
with adjectives: it must occur with definite head nouns, and cannot occur
with indefinites. Note that the modification is a modification of the head,
not of the possessor, although there is inheritance of definiteness from the
latter. This inheritance of definiteness is one of the important properties
of Arabic genitive constructions that needs to be accounted for. 7
Summarizing, I have shown that nominal phrases may or may not bear
definite articles and/or nunation, and that the head noun in genitive con-
structions lacks these affixes. Moreover, there is a matching requirement
in definiteness between the head noun and the genitive NP.
220 CHAPTER 5
b. talaat-u nisaa?-in
three-nom women-gen
Three women.
With these numerals, the agreement is 'privative'. That is, when the genitive
noun is masculine, as in (20a), the numeral bears a feminine marker, but
when the noun is feminine, as in (20b ), the numeral bears no mark of the
feminine. 12
222 CHAPTER 5
~D~sP
\ ~ss'
~o~P
~~
If AGR is projected higher than the functional projection where the
subject is located at S-structure - i.e. DP -then we can derive NSO order,
as well as its agreement properties. Assuming that AGR in Arabic nominal
phrases is limited to the poor option, we can account for both the non-
availability of SNO order and the non-extractability of possessor NPs. The
fact that (almost) no scrambling options are available in genitive co1
structions can also be derived. 13 Ordering possibilities are examined in the
next subsection.
Suppose that these constructions are ruled out as ECP violations. The reason
can be that since AGR is poor in this case, it does not license an NP in
its Spec, and antecedent government cannot go through. The impossibilty
of extracting in this case recalls the impossibility of extracting a subject
in VSO, without moving through Spec of AGR, and without triggering
rich AGR. That amounts to extracting from an SVO structure. But there
is no rich AGR in DPs. We know this because the language has N raising,
but no surface SNO structure, as the ungrammaticality of (24a) indicates.
Only (24b) is possible:
(24) a.* zur-tu r-rajul-i ?umm-a
visited-/ the-man-gen mother-ace
I visited the man's mother.
b. zur-tu ?umm-a r-rajul-i
visited-/ mother-ace the-man-gen
I visited the man's mother.
I have argued in Chapter 2 that (rich) AGR must be checked at S-
structure, and that only rich AGR licenses a NP in its Spec. Poor AGR,
on the other hand, is not a Spec-Head agreement. Furthermore, rich AGR
not only licenses a NP in its Spec, but requires it, otherwise AGR itself
is not licensed.
I have suggested that this bidirectional requirement is regulated by the
AGR Criterion, which states essentially the following:
(25) AGR Criterion. (Rich) AGR is licensed by a NP in its Spec,
and a NP in Spec of AGR is licensed by (rich) AGR.
The AGR Criterion rules out NP raising to Spec of AGR in Arabic
noun phrases. Consequently, extraction of possessors to Spec of CP (or to
adjoin to CP), as in (22) and (23), will have to be a 'long movement',
hence violating the antecedent government requirement on traces.
The absence of (significant) scrambling possibilities within NPs can be
derived by appealing to quite standard assumptions. Consider the following
contrast:
(26) a. ?aqlaqa-nii ntiqaad-u zayd-in camr-an
annoyed-me cntzczzmg-nom Zayd-gen Amr-acc
Zayd's criticizing Amr annoyed me.
b.* ( aqlaqa-nii camr-an ntiqaad-u zayd-in
annoyed-me Amr-acc criticizing-nom Zayd-gen
Different reasons can be advocated to rule out this construction. If NP is
an inherent barrier, and adjunction is limited to[- N] categories (as proposed
224 CHAPTER 5
in Chapter 2), then the movement there is precluded. On the other hand,
if NP is not a barrier (on the assumption that all lexical projections are
not), then the structure can still be excluded if we assume that PossP is
of nominal nature, and that adjunction to nominal projections is not allowed.
One way or another, we derive the fact that ONS order, the counterpart
of OVS order which is permitted in clauses, is ruled out here. 15
Likewise, scrambling of the possessor to a postnominal position (to yield
NOS) is also excluded, as indicated by the ungrammaticality of (27):
(27) a.*? aqlaqa-nii ntiqaad-u camr-an zayd-in
annoyed-me criticizing-nom Amr-acc Zayd-gen
b. *?aqlaqa-nii ntiqaad-u camr-in zayd-un
annoyed-me criticizing-nom Amr-gen Zayd-nom
In these constructions the possessor has been shifted to the right of the
object. The grammatical judgments are the same, whether the shifted subject
surfaces with Nominative or with Genitive, and whether the object surfaces
in Accusative or Genitive. It is possible to exclude (27) by assuming, as I
have done earlier, that the adjunction to PossP (to the right), which is needed
to shift the subject, is not licit. 16
The structure in (21) seems then to account for the distributional and
movement possibilities in nominal phrases. It suggests that the DP system
is parallel to that of IP. The introduction of an AGR node in the structure
of possessives is not without motivation. It is realized as 'rich' AGR in
possessive constructions in Hungarian, Chamorro, and Turkish, for example
(encoding number and person features). In contrast, AGR in Arabic is
limited to the poor option. 17
There are problems, however, which cast doubt on the adequacy of
(21). First, in a number of languages (including Hungarian and Chamorro),
DP can be shown to be higher than AGRP, rather than the other way around,
suggesting that DP parallels CP, not IP. Second, in languages (like
Hungarian) in which an overt rich agreement is present, both NSO and SNO
structures are possible, but the possessor in NSO is in Spec of AGR at
S-structure, not Spec of DP. Third, the possessor and the article are not in
complementary distribution in these languages, nor does the possessor agree
in Definiteness with the possessee. Fourth, the extraction of the possessor
NP is allowed. In the following subsections I compare the Hungarian and
Chamorro type of languages (which have the properties mentioned) with
Arabic and English (which do not). I argue that the DP inflectional system
mirrors that of CP, and that the two types of languages have basically the
same nominal inflectional structure, except that AGRP in the Arabic and
English type appears to play no (significant) syntactic role.
INFLECTIONAL PROJECTIONS IN NOUN PHRASES 225
b. A Peter hang-ja
the Peter-nom voice-Poss.3.s.
Peter's voice.
c. Peter-nek a hang-ja
Peter-dat the voice-Poss.3.s.
Peter's voice.
These data show that there is no cross-linguistic complementary distribu-
tion between possessors and articles. Recall that in order to capture this
distribution in English, articles used to be treated as specifiers of NPs in
the standard account, and this was seen as one of the important advan-
226 CHAPTER 5
tages of the latter. 18 But this treatment cannot extend to bnguages like
Chamorro and Hungarian, nor in fact to Romance languages (where the
article and the possessor cooccur, as we will see).
Note incidentally that interpretively, the possessor is distinct from the
article. In Higginbotham's (1985) thematic theory, for example, the article
saturates (via binding) an open position in N (an E position a la Davidson,
or an R position a la Williams, 1981). But the possessor bears a relation
R to N, which is different from E (or R of Williams). Thus although
Higginbotham treats syntactically the article as a specifier of NP, he has
to state the relation of the article to the noun and the relation of the
possessor to the noun as two different interpretive relations, as shown in
(31) (see Higginbotham, 1983):
(31) [the x: N (x) & R (x, NP)]
In (31 ), the article is binding the position x in N (the equivalent of the
Davidsonian E), and R is establishing a relation between x and another
NP. Suppose R is a theta role relation (borne by Pass), i.e. Pass' theta-marks
NP, as I have suggested, then R cannot be equated with the function of
operator that the article is performing. 19
the article. When the possessor is after D, it must bear Nominative case,
which is assigned under Spec-Head agreement with AGR (at S-structure).
When the possessor precedes D, then it must bear dative case, which we
assume to be assigned by D, under Spec-Head agreement with D (compare
(30b) and (30c) above). In contrast, the extracted possessor can bear only
Dative case, not Nominative, as the following examples show:
(35) Akarom bogy halljad a · Peter hang-ja-t
want-! that hear-you the Peter-nom voice-poss.3s.acc
I want you to hear Peter's voice.
(36) *Peter akarom bogy halljad a hang-ja-t
Peter-nom want-! that hear-you the voice-poss.3s.acc
Peter, I want you to hear (the) his voice.
These facts provide a motivation for hypothesizing that AGR and D are
playing separate roles in Case assignment in Hungarian. Moreover, they
show that possessor extraction has taken place from Spec of DP, and not
Spec of AGRP. There are no facts of this kind in Arabic or English.
How can we account for this variation between the two types of languages?
I will pursue the idea that the structure of noun phrases is identical in
both types. The variation can be captured either (a) in terms of the absence
vs. presence of some functional projections, or (b) the content of these
projections (correlated with morphological differences between languages).
The latter solution is more compatible with our general strategy which
minimizes phrase structure differences between languages.
In order to execute this solution I rely on differences between AGR
and D systems in languages. I claim that the properties of these systems
regulate NP raising and possessor extraction. Only languages with mor-
phological Spec-Head AGR agreement allow internal NP raising. They
also allow (potential) extraction of possessor NPs, the properties of the D
system permitting. A prototypical instance of languages with NP raising
at S-structure and extraction is Hungarian. An instance of non-NP raising
at S-structure and non-extraction is Arabic.
Let us suppose that the inflectional structure of DPs has three layers,
DP, AGRP, and PossP. The possessor is generated in Spec of Poss, assuming
that it is licensed there by thematic principles. It can then raise to Spec
of AGR, at S-structure, and stop there. Because DP is higher than AGRP,
the Hungarian (30b) is generated. The possessor there is Nominative, and
Case checking will require it to be a sister of AGR at S-structure. The
possessor may also raise higher, to Spec of D. The case it receives there
is different, as (30c) illustrates. I assume that Dative in this context is
228 CHAPTER 5
The doubly filled Camp Filter can be derived in the same fashion (see Rizzi,
1990 for other details).
The analysis proposed derives the effects of Abney's condition as well
as Rizzi's rule, while accounting for a wide variation among languages,
and correlating a number of observed properties. It also corroborates the
conclusion reached by other researchers (although with different motiva-
tion): that D in the nominal system is parallel to C, at least in some
languages (see Horrocks and Stavrou, 1987; Horrocks, 1987; and Szabolcsi,
1989; who argue that Spec of D is an 'escape hatch'). The parallelism
between D and C has been generalized here to all languages, and the
escape hatch property derived, rather than stated.
To sum up the results of this section, I have analyzed how distribu-
tional, Case, thematic, agreement, definiteness, as well as extraction (and
scrambling) characteristics of Arabic nominal phrases can be accounted for.
I have shown that Arabic instantiates N raising to D at S-structure, although
it does not instantiate Spec to Spec DP raising nor extraction of DP
possessors. These limitations have been attributed to poorness of AGR
and D. In contrast, languages with rich AGR and D do not exhibit these
limitations. On the other hand, it has been argued that Genitive is
licensed/checked in Spec of D at LF. The agreement in Definiteness between
the possessor and the possessee has also been analyzed. Finally, three
functional layers have been motivated in the structure of DPs, including
PossP. DP Possessors are base generated and theta marked in Spec Pass.
In the next section I examine the internal structure and properties of deverbal
process nominals (ma$dars), and how they behave with respect to the
mechanisms involved in the derivation of ordinary noun phrases analyzed
in this section.
(56) a. ntaqad-tu.
criticized-/
I criticized.
b. nabbah-tu
call.attention.past-1
I called (someone's) attention to.
(57) a. laa ?-uriidu 1-intiqaad-a
not /-want the-criticizing-ace.
I do not want to criticize.
Then, in order to derive the right order, not only has N to move to D, as
I have explained earlier, but the 'agent' (if generated inside NP or VP)
must also move higher than the modifier, to yield the correct order. The
potential position where it surfaces is presumably Spec of Poss. Note,
however, that if the subject were to move from Spec of NP or VP to Spec
of Poss, then the chain created will receive two theta roles (Agent and Poss),
resulting in a theta Criterion violation. If that is true, then the only way
to avoid this problem is to assume that subjects of deverbal nouns receive
only Poss roles (compositionally). That suggests that the subject has not
been generated lower than the adverb, and then raised to Spec of Poss,
but that it has originated there. There are different ways to derive this
fact. One of them is to assume that external roles (among which is the
Poss role) are not discharged inside VP or NP, but only under government
by a functional category. 29
Summarizing, I have proposed that nominal conversion takes place either
at the V level, or at the VP (or V") level, giving rise to different catego-
rial properties. Result nominals are derived syntactically from a verbal
source at the xo level, but they differ from process nominals formed at
the same level in that the nominalizing affix has no argument structure. This
takes care of differences between result and process nominals, on the one
hand, and various process nominals, on the other hand. However, there is
242 CHAPTER 5
stay in situ, or the Case Criterion will be violated. That the reflexive
cannot stay in situ (and receive Accusative) is established by the ungram-
maticality of the following examples:
(67) a.* y-uriidu ntiqaad-a-n r-rajul-a
he-wants criticizing-ace the-man-ace
He wants to criticize the man.
b.* y-uriidu ntiqaad-a-n nafs-a-hu
he-wants criticizing-ace self-ace-his
He wants to criticize himself.
In the same vein, a NP with an article cannot occur in an accusative
configuration, as (68b) shows:
(68) a. tamma hadm-u 1-madiinat-i
achieved destroying-nom the-city-gen
The city's destruction has been achieved.
b.* tamma 1-hadm-u 1-madiinat-a
achieved the-destroying-nom the-city-ace
The city's destruction has been achieved.
In order to account for the ill-formedness of (68b) and (67), I propose the
following condition on Case discharge:
(69) Object Case is discharged only if subject Case is discharged.
In essence, this condition prohibits predicates which discharge cases from
discharging the case of the object, without discharging that of the subject.
Given this condition, and assuming that PRO (being caseless) does not
absorb Genitive case, we can ensure that the thematic object in (65) will
surface in Genitive, not Accusative. 31
Another possibilty is that V is nominalized early in the derivation, at
the X 0 level, and projects only a NP structure. Then N (at the low level
of the tree) can be thought of as assigning Genitive to the object in situ.
That will solve the Genitive problem. There are other problems which
arise with this solution, however.
Consider the class of Genitive case assigners. Conceptually, Ds and
prepositions are relational, and seem to assign Genitive by virtue of this
property, but N is not. If N were a Genitive case assigner then it is diffi-
cult to see what would unify the class of Genitive assigners. Descriptively,
this option also raises problems. If Genitive were (directly) assigned by
N to its complement, and by D to its Spec, then we expect that, in the context
of a transitive N, two (directly) assigned Genitives would occur. The expec-
tation is not borne out. In the context of a derived nominal, for instance,
244 CHAPTER 5
the agent behind, but the result is ruled out. Whether the subject is marked
as Nominative or Genitive does not affect the judgment. The question, then,
is why affixation over a thematic subject is possible with inflected verbs,
as (75) shows, but is not possible with process nominal constructions
(above):
(75) ntaqada-hu zayd-u-n
criticized-him Zayd-nom
Zayd criticized him.
Note that, since the affixation process in (75) 'violates' the Specified Subject
Condition of Chomsky ( 1973), it is unlikely that the ungrammaticality of
(74) can be attributed to this condition.
There are two ways to derive the ungrammaticality of (74). The latter
can be construed either as a Case theory violation, or as an ECP viola-
tion. First, it is possible to think of (74) as a Case theory violation. If the
affix on the process nominal is Case marked as Genitive, then the only Case
left to assign is Accusative. Second, if the pronominal affix were to move
in one step to the head noun in D, then its trace will not be properly
governed, assuming that either NPs or PossPs are barriers. On the other
hand, the movement cannot proceed via adjunction to nominal projections,
by hypothesis.
Summarizing, I have discussed how various ma~dar constructions are
formed, as well as Case, thematic and argumental properties of ma~dar
affixes. In the next section I turn to investigating the characteristics of
analytic genitives.
3.1. Arabic
Arabic makes use of prepositions as Case marks of NPs. The preposition
li, which has the same form as the dative one, may be used (as an alter-
native to accusative Case; see (73)) to Case mark the direct object of a
monotransitive deverbal predicate, as in the following construction:
(76) ntiqaad-u r-rajul-i li-1-masruuc-i
criticizing-nom the-man-gen to-the-project-gen
The man's criticism of the project.
The same form of the preposition is also used with a dative object of a
ditransitive predicate, as in (77):
(77) bay"-u r-rajul-i 1-kitaab-a li-zayd-in
selling-nom the-man- gen the-book-ace to-Zayd- gen
The man's selling of the book to Zayd.
I think, however, that the dative preposition is different from the 'objective'
one, although they have the same form. While the dative preposition is a
theta marker, the objective one is only a Case marker. Evidence for this
comes from the fact that the dative preposition cannot Case mark the first
object of non-dative ditransitive nominals. For example, with salb the
preposition used must be min, and cannot be li, as the following contrast
shows:
(78) a. salb-u zayd-in r-rajul-a maal-a-hu
depriving-nom Zayd-gen the-man-ace money-ace-his
Zayd's depriving the man of his money.
b.* salb-u zayd-in maal-a-hu li-r-rajul-i
depriving-nom Zayd-gen money-ace-his of-the-man-gen
Literally: Zayd's depriving his money of the man.
c. salb-u zayd-in maal-a-hu mina r-rajul-i
depriving-nom Zayd-gen money-ace-his from the-man-gen
Literally: Zayd's depriving from the man his money.
The preposition min theta marks the 'source', and the preposition li
the beneficiary or goal, hence the ungrammaticality of (78b). Note, by
way of comparison that the same case patterns obtain with inflected
verbs.
Another important difference between the two prepositions is that the
dative preposition occurs with verbal predicates, whereas the objective
preposition occurs only with nominal ones. Consider the following
examples:
248 CHAPTER 5
3.2. Romance
Consider the following examples from French:
(94) Le portrait de Marie de Rembrandt.
the portrait of Mary of Rembrandt
Rembrandt's portrait of Mary.
(95) Le portrait de Rembrandt de ce collectionneur.
the portrait of Rembrandt of this collector
This collector's portrait by Rembrandt.
As argued by Milner (1977-1982) for French, Cinque (1978-1980) and
Giorgi and Longobardi (1987-1991) for Italian, and Torrego (1988) for
Spanish, there is a hierarchical structure of de-di NPs in Romance, where
the possessor is generated higher than the agent, which is in tum higher than
the theme, according the following hierarchy:
(96) Poss < Agent < Theme.
On the other hand, a NP like (97) is three way ambiguous, unlike the
situation I have described for Arabic and English:
(97) Le portrait de ce collectionneur.
The portrait of this collector
This collector's portrait.
Observe that there are two problems here: theta marking and Case marking
(see Emonds, 1985). The differences with Arabic (and English) can be
described by stating that the preposition is inserted only between N and
its sister complement in Arabic, but extend to complement of N' and of
Poss' in Romance. That is, I assume the following structure for Romance
NPs:
(98) DP
~
D PossP
I~
le Poss' de ce collectionneur
~
Poss NP
I~
portrait N' de Rembrandt
~
N de Marie
I
e
252 CHAPTER 5
In this structure, the head noun has moved to Poss, and NPs are gener-
ated as complements of N, N', and Pass', respectively, in conformity with
the hierarchy in (96). Given this structure, the conditions under which the
preposition is inserted appear to be uniform: it is inserted under head gov-
ernment by N or any of its extended projections in the sense of Grimshaw
(1991). 33 The alternative will be that these positions are Spec positions to
the right (as in Torrego, 1988, and Giorgi and Longobardi, 1991). I return
to this problem later on, but it is tempting, in order to account for the
distribution of of prepositions in a uniform manner to take these NPs to
be simply Comp(lement)s, not Spec(ifier)s.
Given the differences outlined, the question is: why can't Arabic (and
English) prepositions have the same distribution? Why can't the preposi-
tion be inserted higher in the tree, just like the Romance one? There is
another set of differences between the Arabic and Romance D systems
which, if correlated with the prepostion marking system, can help us
characterize the variation described. As I have explained earlier, Arabic (and
English) Genitive is assigned via agreement with D. In Romance, by
contrast, there is no (straightforward) sense in which D can be taken as
assigning Case to the possessor (or the agent). First, there is no agree-
ment in Definiteness between the possessor and the possessee, as the
following example shows:
(99) Le chapeau d'un gar9on.
Second, there is no complementary distribution between the possessor and
the article. Third, there is no evidence for NP raising in NP, be it at
S-structure, or at LF.
These properties differ from those of the Arabic system. In the latter,
D is involved in Case marking and agreement, as I have explained. Suppose
that D in Arabic possessive constructions is a Case assigner (or has a Case
to discharge), whereas Romance D is not. Suppose, too, that of insertion
is a last resort Case mechanism. Then a language can make use of this
mechanism only when no other Case mechanism is available. Since Genitive
case is available for Arabic (and English) through D, then that mecha-
nism has priority over the of one, thus accounting for the ungrammaticality
of (89), (90), and (93). Note that the preposition is a 'Case sign', not a
Case assigner (as in Pollock, 1989b and Borer, 1983). Consequently, it
has to be licensed by a functional category. The licensing category may
be AGR or D in Romance, but we may speculate that these categories do
not prepositionally license case-marked NPs in Arabic (and English) when
they are empty. The latter solution seems to be more attractive than that
which limits insertion of prepositions, without correlating it with other
properties of D and AGR in the nominal system.
As it turns out, this solution makes the right predictions for further data
INFLECTIONAL PROJECTIONS IN NOUN PHRASES 253
3.3. Extraction
As is well known, extraction of de-NP from within NPs is possible in
French, although it appears to be subject to the following limitation (on
this matter, see Ruwet, 1972; Milner, 1977-1982; Pollock, 1989b; and
Godard, 1990):
(102) Only de-NP which can surface as a possessive pronoun can be
extracted.
254 CHAPTER 5
the article on the head noun. When the article is present, extraction is
blocked. In the case of (115b), where the NP is the subject, we can derive
this result from Case theory by appealing to the 'elsewhere' requirement
advocated earlier, which forces the subject to pick the Genitive from D,
rather than from li. In the case of (114b ), if the extracted element is the
object of 'picture', the solution does not seem to carry over in a straight-
forward way. There are two options that are already available, and which
suggest themselves. The first one is to derive the ungrammaticality of (114b)
from Case theory, in the same way that we did for (85b). The second one
is to derive it from agreement theory. In both cases the extracted object
has to be treated on a par with a (derived) subject. The issue is whether it
escapes from a post-N position (call it Comp), or from a pre-N position (call
it Spec).
Note that, given the assumptions we have made so far, either alterna-
tive will be descriptively adequate. The question is only to favour one option
over the other. Suppose li-NP were to move through Spec of D, then we can
account for the ungrammaticality of (114b) as a result of the failure of
agreement between the article and the subject, as we did for other cases
discussed earlier. But a number of problems arise with this solution,
however.
First, if raising through Spec were allowed, then we are left with no
explanation of why neither li-NP nor genitive NP can surface in Spec.
That is why a construction like (24a), for example, repeated here as (116),
is excluded:
(116) * zur-tu r-rajul-i ?umm-a
visited-! the-man-gen mother-ace
*I visited the man's mother.
A surface filter to rule out this outcome can hardly be motivated. 34
A second problem arises with regard to extraction from indefinite noun
phrases. As observed earlier, li-NPs are extractable in this context, but
genitive NPs are not, as the ungrammaticality of the following example
indicates:
(117) *rajul-in btaraq-at daar-u
man-gen burned-! house-nom
*A man, a house burned.
A third problem arises with respect to scrambling. We have seen above
that scrambling of the different arguments in synthetic gentive constructions
is impossible, but it is possible with analytic li-NP:
(118) a. qara?-tu maqaal-an la-ka can 1-?abdaaH
read-! article-ace of-you on the-events-gen
I read an article of yours on the events.
INFLECTIONAL PROJECTIONS IN NOUN PHRASES 257
There are Arabic nominal forms which correspond roughly to the English
agent nominals formed by suffixing er to the verbal stem (as in driver).
These nominals have been misleadingly termed agent nominals by some
Western scholars (see e.g. Wright, 1951 ). As will become clear, however,
the role they name is not necessarily the agent (nor that of the actor, as
proposed by Marantz, 1984 and Sproat, 1985b for English). They simply
refer to the subject of an active verb. It is more appropriate to call them
subject (or active) nominals, after traditional grammarians. Their form is
identical to that of active participles (see Chapter 4). 36 Syntactically,
however, they are nouns, and their behaviour is close to that of English agent
nominals, although they differ from the latter in thematic characteristics,
as we will see. In this section I examine their essential properties, and
how they are derived.
established, despite the fact that the two categories are morphologically
indistinguishable. The participle has none of the combinatorial and
categorial properties illustrated by (119) and (120). In (119) the subject
nominal heads a genitive construction. It also inherits Definiteness from
the genitive NP, as in possessive constructions, although the genitive NP
is thematically a complement. Moreover, the subject nominal is assigned
Nominative by virtue of being subject, and not predicate (unlike the
participle). In ( 120), the subject nominal bears a definite article, and
functions exactly like any regular definite NP. Participles can also bear a
definite article (as adjectives do), but they function only as predicates.
On the other hand, the subject nominal in (120) is modified by an
adjective. It does not allow modification by (manner) adverbs, as shown
by the ungrammaticality of (122):
(122) *jaa? a
qaatil-u s-saJnn-1 bi-watlsiyyat-in
came killing-nom the-prisoner-gen with-brutality-gen
* [The killer of the prisoner brutally] came.
The construction ( 122) is deviant in the reading whereby the killing was
rapid, thus indicating that adverbial modification of subject deverbal nouns
is ruled out.
A further important nominal property of subject nominals is that they can
form broken plurals. The latter form of plurals is exclusively nominal, unlike
so-called sound plurals which can be adjectival as well as nominal. Thus
nominals like caamil 'working/worker', {iaabit 'officiating/officer', taajir
'trading/trader'' ~aanr 'making/maker' pluralize as cummal 'workers''
{iubbaat 'officers', tujjaar 'traders', ~unnaac 'makers, artisans', etc., when
taken as nouns. Broken plurals are non-concatenative. When taken as
participles, however, these forms concatenate with a plural suffix, forming
a 'sound form' (see e.g. Wright, 1974). Plurals of these forms are then:
caamil-uun 'working-pl.', {iaabit-uun 'controlling-pl.', taajir-uun 'trading-
pl.', ~aanic-uun 'making-pl.', etc. In fact, the plural number is part of the
AGR marker heading the participle construction (as has been argued in
Chapter 4). These properties (and others) can be used to test whether the
form is nominal or participial/adjectival.
Now consider some properties that seem to push in the other direction,
i.e. to distinguish these forms from ordinary nouns. First, subject nominals,
unlike ordinary nominals, may assign accusative Case to their second object
(the first being marked as Genitive), as in the following example:
(123) jaa?a waahib-u s-saJnn-1 1-maal-a
came donating-nom the-prisoner-gen the-money-ace
The donor of the money to the prisoner came.
Second, subject nominals exhibit dative alternations. The dative object
260 CHAPTER 5
can be marked with Genitive, and the direct object with accusative, as in
(123), or the direct object is marked with Genitive, and the dative object
marked with the dative li preposition, as in (124):
(124) jaa?a waahib-u 1-maal-i li-s-sajiin-i
came donating-nom the-money-gen to-the-prisoner-gen
The donor of the money to the prisoner came.
These Case and thematic properties recall similar properties of deverbal
event nominals (and participles as well; see Chapter 4). In order to account
for the latter I have postulated there that the lexical X0 head of the
construction is V, although it is later converted to N. A similar analysis
can be adopted for the nominalizing process here. But before providing such
an analysis I have to examine in more detail the thematic and selectional
properties of the nominal.
c. zayd-un mursil-un
Zayd-nom sending-nom
Zayd is a sender.
These examples clearly show that the distribution of internal arguments
in active nominals is not arbitrary. In fact, apart from the constructions
(138c) and (140c), in which no internal argument occurs, the occurrence
of internal arguments here seems to be exactly parallel to that of verbs.
For example, the subject nominal in the dative construction cannot occur
with the dative only, as shown by the ungrammaticality of (138b) and
(140b). This is exactly what happens with tensed verbs or active participles,
differences in Case assignment aside. 39 How can these facts be accommo-
dated?
266 CHAPTER 5
Thus the APP accounts for the facts of thematic preservation in subject
and event nominals. The question is then whether it generalizes to verbs,
and how. We will see that it does, although at first glance the facts seem
to suggest the contrary. I return to this problem after I provide an account
of how the subject nominal is formed, and how its thematic structure
composed.
To begin with let us characterize the [aa-i] affix, which forms active
nominals by combining with a verbal root. An adequate lexical entry for
this affix will necessarily include the following information:
(142) [aa-i]: a. (af. (1))
b. (V, N)
Part (a) of the entry provides the argument structure of the affix, and (b)
its categorial properties (namely the fact that it converts V to N). I have
labelled the subject role with 1 in (a). Where the semantics is compositional,
the subject nominal is formed (in syntax), by affixing [aa-i] to V. Thematic
marking proceeds via theta identification. As with event nominals, the
different Case properties are accounted for via the 'scope' of affixation.
There are different ways to execute these ideas. For example, when the
internal structure of the nominal is verbal, we can take the nominalizing
affix to head a separate projection, and the role it identifies not to be
assigned within the clausal VP. To illustrate, consider again the construc-
tion (123), repeated here as (143):
(143) jaa?a waahib-u s-saJnn-1 1-maal-a
came donating-nom the-prisoner-gen the-money-ace
The donor of the money to the prisoner came.
The nominal phrase there presumably has the following structure:
(144) DP
~
D'
~
D NP
~
N'
~
N VP
~ /~
V1 N sajiin V'
whb
I I~
[aa-i] Vo DP
I
e
I
maal
268 CHAPTER 5
In this section I place the descriptive results reached earlier in a more the-
oretical context, in particular within a principled approach to D and AGR
agreements. I also reexamine the parallelism between systems D and C, and
provide a way to derive the specificity requirement on indefinite geni-
tives. Finally, I speculate on Poss realization.
270 CHAPTER 5
Observe that the latter view depends crucially on the idea that Genitive is
'inherent'; that is, it is assigned by virtue of a thematic relationship. I
have provided reasons to think that Genitive case is structural, and that
the two Case marks belong to different strategies. The latter are checked
under different relations, and are subject to different licensing principles.
One advantage of this view is that it enables us to parametrize the distri-
bution of the two strategies across languages by making use of the Elsewhere
licensing requirement stated in (152).
Another important advantage of the Checking view is that it accounts
more accurately for the fact that a possessor may have different cases
depending on whether AGR is involved, or D is involved. As we have
seen, Hungarian possessors, for example, can be marked either as Dative,
or as Nominative, depending on where they are located. Likewise, Arabic
subjects can be genitive (see above), Nominative, or Accusative. By analogy,
I have taken Genitive and OF to be different Case marks, checked under
different conditions. This view is more in line with recent proposals made
by Chomsky.
Observe that the typology I have established takes account of AGR and
D properties, and their effects on order and Case. What looks like SNO
order, then, can be three-way ambiguous: (a) base generated as it is (with
no movement); (b) N has moved to AGR, and the subject to Spec of AGR
(and D is empty); (c) N has moved to D, and the subject to Spec of D.
Similarly, NSO can be the result of moving N to AGR and to D, and either
leaving the subject in situ, or moving it to Spec of AGR. Most of these
possibilities (if not all) have been shown to be instantiated in languages.
Moreover, they seem to indicate that the Case checking theory is on the
right track.
Milner (1982) observes that the head noun of a genitive in French must
carry a definite article. He conjectures that examples (155) are marginal
or excluded as genitives (although they may have a partitive reading), and
those in (156) are canonical genitives:
(155)a.*Un fils d'un voisin
b. ?Un fils du voisin
(156) a. Le fils du voisin
b. Le fils d'un voisin
If Milner's judgments are correct, then specificity in French will be
expressed by the definite article, and it will be a (possible) property of
analytic genitives as well.
In Chapter 2 I have treated specificity in clauses as a property of a weakly
referential chain, one member of which being a rich agreement marker. In
the case discussed here, the rich member (in the absence of rich AGR) is
D, which agrees with its Spec, as I have suggested earlier. The latter result
confirms the thesis that D is an agreeing category. Note that checking D
agreement occurs at LF. It does not need to take place at S-structure.
Throughout this chapter I have defended the idea that Poss projects as a
separate category. The motivation of this projection has been essentially
thematic. We have also seen the role it plays in word order. Let us turn
now to a puzzling problem raised at the beginning of the chapter with respect
to the nature of nunation. At this stage, we can hypothesize that nunation
is a realization of Pass. Nunation realizes Pass just in case the Possessor
is not present. If this is correct, then it will have (at least) two advan-
tages: (a) it will account for the complementary distribution of Poss marking
and nunation; and (b) it will keep Definiteness and Pass marking separate.
Recall that both marks are lacking in nominal genitives, whereas only
nunation is lacking in adjectival genitives.
This view of Pass marking is further supported by the existence of
languages which distinguish between 'absolute' (or 'free') forms of nouns
and 'construct' forms which occur in possessives. For example, in Uta-
Aztecan (Northern Paiute) as described by Langacker (1982), nouns are
marked with absolutive, which" ... effectively marks unspecified possessor"
when they are not possessed. When they are possessed, they have a different
marking, indicating that Poss marking may be (overtly) grammaticalized
in languages. 41
274 CHAPTER 5
6. CONCLUSION
NOTES
' Part of this materi1 is contained in Chapter 4 of Fassi (1989b and c), which is in turn
based on Fassi (1%7a). Another part has been presented as a lecture in Paris VIII (Winter
1990) entitled: 'Quels chemins pour NSO?'.
1 The idea that D is the head of nominal phrases is originally due to Brame (1982). Abney
has elaborated on this idea within GB.
INFLECTIONAL PROJECTIONS IN NOUN PHRASES 275
2 On this terminology, see Marouzeau (1951). On construct states in Hebrew, see Borer
(1983), among others.
3 The preposition li introduced in (2) is a Case mark, as will be explained later on. In
Fassi (1985b) and (1987a), I have proposed deriving the order in Arabic NPs via N fronting,
generalizing Emonds' (1980) original V fronting rule. Ritter (1987/1988, 1991) has worked
out a DP analysis for deriving Hebrew NP structure which differs in essential ways from mine.
See notes 21 and 28, in particular, for comparison.
4 Traditional grammarians have also pointed out that nunation is a mark of tamakkun, that
is of 'complete nominality'. Some nouns do not have such a property (see Sibawahyi, 800,
and Wright, 1974, on the matter).
5 This [n] form occurs only on so-called sound plurals, and duals. Broken plurals do not
carry [n] when they are definite. Traditional grammarians make a distinction between this [n],
which they call nuun (i.e. a simple n), and tanwiin (nunation). The distinction is simply
graphic, in my view. The two [n] are unified here, and provided a single treatment.
6 Vocatives share the lack of these two marks with genitive constructions. The following
examples illustrate this fact:
(i) yaa xaalid-u
0 Khaled-nom
0 Khaled!
(ii) yaa rajul-u
0 man-nom
0 man!
(iii) * yaa r-rajul-u
0 the-man-nom
*0 the man!
(iv) yaa bn-a ? ax-ii
0 son-ace brother-me
0 son of my brother'
7 On the properties of Arabic modification, see Fassi (1976 and 1981). See also Fassi (1987a)
for tests of Definiteness.
8 On this view, see e.g. Grimshaw (1987-1990) and Zubizaretta (1987). According to
Emonds (1985), nouns theta mark their arguments, but only indirectly. In contrast, Chomsky
(1986a) assumes that nouns and verbs do not differ in theta marking capacities, but only in
Case marking properties. I will adopt Emonds' proposal in this work.
9 See Anderson (1983) and Chomsky (1986a) on a similar idea, though linked to Genitive
Case marking. On indirect theta marking in the same spirit, see Marantz (1984) and more
recently Pollock (1989b). Traditional grammarians have observed that there is a sort of
predication (they call it nisba 'attribution') within the genitive construction. If the subject role
is to be assigned by virtue of predication, then the possessor role would be, too. I think,
however, that theta role assignment and predication are different mechanisms.
10 An analysis of the complementary distribution of Poss marking and nunation will be
proposed in Subsection 5.4 below.
11 Other alternatives are either that Genitive is assigned to the Possessor by D in an ECM
configuration (as proposed by Fassi, 1987a), or by N to its complement. Both options will
be shown to be inadequate.
12 On the intricacies of the Arabic numeral system, see Fassi (1981 ), and for numeral
agreement, Fassi (1984).
13 One might wonder why AGR in Arabic nominal phrases should be limited to the poor
276 CHAPTER 5
option, while this is not so in clauses, as pointed out to me by a SNLL T reviewer. The
way I think of this problem at this point is that this limitation is simply accidental (and/or
lexical). AGR nominal morphemes are not specified for NUM in Arabic, while they are so
in other languages (Chamorro, Turkish, Hungarian, etc.).
14 Note that these structures can be 'saved' if a resumptive affix is attached to the NP, as
shown in (i):
(i) man qara? -ta kitaab-a-hu
who read-you book-ace-his
Literally: Who did you read his book?
Whose book did you read?
There is no reason, however, to equate pronominal affixation with rich agreement. On
significant differences, see Chapter 3.
15 Similar facts are ruled out in Romance. For an analysis, see Giorgi and Longobardi (1991).
16 The only scrambling possibility that we know of with synthetic genitives is to place an
adverb between the possessor and the object, as in (i):
(i) ntiqaad-u-ka 1-yawm-a 1-? ustaag-a ? a'jaba-nii
criticism-nom-you today-ace the-teacher-ace pleased-me
Your criticism today of the teacher pleased me.
A modifying adjective of the head cannot be scrambled between the possessor and the
object. Only the analytic strategy is used in this case, as in (ii):
(ii) ntiqaad-u-ka 1-waarid-u li-1-? ustaac;!-i (*1-?ustaag-a)
criticism-nom-you the-relevant-nom of-the-teacher-gen (*the-teacher-ace)
Your relevant criticism of the teacher.
See below for an anlysis of these judgments.
17 See Abney (1987) on agreement in DPs. On Hungarian, see Szabolsci (1989) and Maracz
(1989), on Greek, Horrocks and Stavrou (1987), and on Turkish, Underhill (1976). In Fassi
(I 989c), D is taken to be the counterpart ofT, and AGR in the noun phrase the counterpart
of AGR in the sentence. There are other properties of D and T which make them parallel.
For example, both T and D may be deictic. They can be treated as 'binders', in the sense
that both of them discharge a position in a theta grid of a predicate via theta binding (see
Higginbotham, 1985, 1986 for details). As for AGR, it seems peculiar, at first glance, to
hypothesize that it heads the genitive construction, given the fact that nouns do not usually
bear agreement markers, the latter occurring only with predicates. The hypothesis is supported,
however, by empirical as well as theoretical evidence, as we will see.
18 In Jackendoff (1977), for example, the head noun is generated under N, and the article
as Spec of N'.
19 It remains to be seen whether R is a predication or modification relation. The latter
view has been defended by Grimshaw (1990), who treats possessors as argument-adjuncts.
20 I take (38) to be an instance of NP raising in 'active' nominals, following an idea of
Pollock (1989b). In contrast, I think that the genitive NP in constructions like The city's
destruction may be base generated as a possessor, or treated as a 'passive' possessor. I assume
that passive raising is regulated by different principles than AGR or D raising (see Fassi,
1988b for details).
An SNLLT reviewer wonders why (37) is not grammatical in English, but is perfectly well-
formed in Hungarian. The explanation I have provided is that AGR and D in Hungarian
are rich, and hence license Spec to Spec movement. In English, by contrast, these struc-
tures violate the AGR Criterion.
21 In order to handle the facts of complementary distribution in Hebrew construct states,
Ritter (198711988) proposes that the D position is filled by an article which is lowered onto
INFLECTIONAL PROJECTIONS IN NOUN PHRASES 277
the constituent following it. Thus, in ordinary DPs, the definite Hebrew marker ha- is prefixed
to the head noun, but in construct states it is attached to the genitive. According to her,
"This accounts for the fact that ha- is realized on the genitive phrase in a CS but on the
head noun otherwise" (p. 17). But this lowering operation has no independent motivation,
as far as I can tell. Since the head noun and the possessor can bear different articles, it is
unclear why there should be only one article on the head noun in the CS, lowering onto
the genitive.
22 As will be explained later on, any extraction is in fact blocked in the context of an
article. It is not clear from the data whether the possessor can raise to Spec of AGR, nor is
it clear how the internal structure of nominals is derived in Chamorro. Chung ( l990b) proposes
that at D-structure, Spec of D (which is to the right of the head) hosts the possessor. The latter
is then lowered to adjoin to the head N at S-structure, in the same way that the subject in
IP lowers to adjoin to V. This proposal raises a number of theoretical and descriptive problems,
as observed in Chapter 2. See also Section 6, below.
23 I assume that (in)definiteness is a necessary semantic property of nominal expressions,
and that D has to be there for interpretive reasons. For an overview of different approaches
to (in)definiteness, see Reuland and ter Meulen ( 1987). If D is filled by an article, it will count
as a head and will block antecedent government when it does not agree.
24 Ibn as-Sarraaj, an Arabic grammarian of the 11th century, defines the noun as 'what
expresses a single meaning, and this meaning might be an individual or a non-individual.
Examples of individuals are: rajul 'man', faras 'horse', f;ajar 'rock', balad 'country', 'Amr,
and Bakr. Non-individuals are: (i.arb 'beating', ?akl 'eating', qann 'thinking', 'ilm 'knowing',
l-yawm-a 'today', l-laylat-a 'tonight', and s-saa'at-a 'now;. See Al-?u,wul, Vol. 1, p. 36.
He then adds: "I said that it expresses a sole meaning because I want to distinguish it from
the verb, since [the latter] expresses some meaning and time, and time is either past, present,
or future. You might object that among nouns are 'today', 'tonight', and 'now', and these
are times. What is then the difference from verbs? Our answer is that the verb is not solely
time, in the same way that 'today' is solely time. 'Today' has only a time meaning, and it
does not convey any another meaning. The verb is [also] divided according to three times
[... ]. Hence if the word expresses time only, it is a noun. If it expresses a meaning and a
situated tense (mu/l~al), it is a verb" (Ibid., p. 37).
On the other hand, Siiraafii defines the noun as "a word which expresses a meaning by
itself, without been associated with a situated tense". See Ibn Ya'iiS, Vol. 1, p. 22. In the
traditional trichotomy of parts of speech (verb, noun, and particle), only verbs and nouns
express meanings by themselves, while the particle expresses a meaning 'with' the noun or
the verb. Moreover, only (inflected) verbs are associated with tense. Thus nouns and masdars
are grouped together as not being associated with so-called situated tense. Nouns stand not
only for individuals, but also for events (?af;daat). As observed by Ibn as-Sarraaj, " ...
events denote vague (mubham) tenses, because an event can only be in a certain tense.
Verbs denote a determined tense (ma'luum), be it past or non past" (Ibid., p. 22).
Beside these definitions, grammarians have gathered 'the things by which the noun is
identified', that is, its properties and marks. Among these is the ability to take a definite article,
to receive nunation (when indefinite), to be constructed with a genitive complement, to
pluralize, to be able to have a broken plural, to undergo diminutive formation, to be modified
by an adjective, etc. Suyuutii provides more than thirty nominal marks and features in
Al-?asbaah wa n-nadaa?ir (Vol. 2, p. 9). Obviously, these properties, distributions, and marks,
are not realized in every case. Take, for example, masdars, which are classified as nouns,
according to the tradition. As is well known, this class of nouns does not allow pluraliza-
tion, nor diminutive formation. Or take their definition of verbs as necessarily expressing a
'situated' tense. This requirement excludes infinitive verbs in English or French from the
category of verbs, and we have to classify them as nouns.
25 For some concrete proposals, see Seghrouchni (1989).
26 For a different theoretical approach to Hebrew nominalizations, see Hazout (1990).
278 CHAPTER 5
27 Traditional grammarians claim that having the ability to be constructed in the vocative
is the nominal property par excellence, because the noun in the vocative is assimilated to
an object both grammatically and semantically (see Suyu\ii, Ham', Vol. 1, p. 5). But ma~dars
are odd in the vocative, as illustrated by (i):
(i) * yaa duxuul-a d-daar-i'
0 entering-ace the-house-gen
28 Following Sproat ( 1985b ), I assume that " ... thematic grids are required to percolate
from stems to dominating nodes of a different category only when the grids are associated
with (the grids of) the affixes by some regular method of discharge such as theta marking,
theta binding, or theta identification" (p. 188). Sproat states the following convention (which
I assume):
- Cross-Categorial Theta Grid Percolation Convention. In a structure [A (A, B)h, where
A is a lexical item of category A, (A, B) an affix forming words of category B from
lexical items of category A, and the whole structure is thus of category B, the theta grid
of A is only required to percolate to B where A and (A, B) are associated by some regular
process of thematic discharge.
This convention enables us to generate both result and process nomina1s in the syntax, although
the percolation takes place only in process nominal formation.
29 If discharge of the external role is licensed only under a predication relation, and if the
latter obtains only under government by a functional category, then the NP subject will receive
a theta role only in the context mentioned.
On the basis of similar facts in Hebrew, Ritter (1991) argues that the subject of the deverbal
noun has to raise to the specifier of a functional category she identifies with Number. I
think, however, that the latter analysis makes incorrect predictions with respect to raising, and
extraction, as well as agreement. As I have already explained, languages like Arabic (as
well as Hebrew) have no raising and no extraction because a NP is not licensed in the Spec
of their functional projections at S-structure.
30 The non-Arabist reader might wonder why the possessive pronoun in naf>i-hi is hi rather
than hu. The form hi is a surface realization of hu when it is immediately preceded by an
[i] vowel.
31 Yip, Maling, and Jackendoff (1987) as well as Maling (1991) have proposed that
(grammatical) cases form an autonomous Case tier, and they are mapped onto arguments
(or GFs) by principles analogous to those of autosegmental phonology. Following Zaenen,
Maling and Thn:iinsson (1985), Maling (1991) assumes the following mapping principle:
(i) The highest GF is assigned NOM case, the next highest ACC.
Consequently, NOM is assigned before ACC. See also Kiparsky (1985) and Fassi (1986a)
for arguments that Case assignment observes a hierarchy in which Nominative is the highest.
32 As pointed out to me by Joan Maling, the thematic restrictions on the use of li might
be taken to indicate that it is a theta-marker, in addition to being a Case marker. She also
observes that the use of English of is thematically restricted, as illustrated by (i):
(i) Our help to/*of the poor
33 Grimshaw defines the notions of extended head and projections as follows:
x is the (extended) head of y and y is an (extended) projection of x iff: y dominates x;
x and y share all categorial features and all nodes intervening between x and y share all
categorial features; and the value of [F] for y is greater than for x.
F is a multivalued feature assigned to lexical and functional projections. It can have the
following values: (a) [Fl] is assigned to lexical categories (V and N), [F2] to functional
categories (0 and 1), and [F3] to what she calls super-functionals (C and P).
34 In order to rule out similar outcomes in Italian, Giorgi and Longobardi (1991) appeal
INFLECTIONAL PROJECTIONS IN NOUN PHRASES 279
homophonous with determiners of nominal phrases, as pointed out to me by Ken Hale. Joan
Maling (personal communication) has also observed that this fact is quite common. In English,
French, and Welsh (among a number of other languages), complementizers of relative and
factive clauses are homophonous.
43 Variants of the NP hypothesis have been proposed for Hebrew by Borer (1983, 1988),
Hazout (1990), and Shlonsky (1990). Possessor lowering has been advocated by Chung
(1990b) for Chamorro. I think, however, that the analysis adopted here is descriptively more
adequate, in addition to being theoretically preferable.
CONCLUSION
281
282 CONCLUSION
It has been argued that auxiliary, modal, and negation structures are
bi-temporal. The properties of the rich Arabic Neg system have been
investigated. A typology of Neg morphemes has been proposed, depending
on whether they are mood assigners, Case assigners, or neutral. It has
been argued that Neg morphemes are heads, and their extraction proper-
ties have lead me to adopt a rigid version of Minimality, together with a
version of ECP a La Kayne ( 1984), based on the notion of g-projection.
Participle inflection has been shown to carry no tense. Moreover, its
aspectual value has been characterized in terms of contingency, a notion
which has proved to be more adequate than the dynamic/static dichotomy.
Tense inflections have been shown to be sensitive to categories, and to
interact with situation aspect and mood characteristics of predicates. The
(non)visibility of the copula has been investigated in this context. Particular
attention has been paid to the derivation of adjectives and participles, what
properties they share, and how they differ. I have proposed that adjectives
are associated with (pure) states in the lexicon, and that they are not
deverbal. In contrast, participles (like verbs) are associated with processes,
events, or contingent states. Moreover, they are deverbal, and their cate-
gorial properties vary according to the level at which their stativization
and/or adjectivalization takes place.
I have also investigated the inflectional structure of nominal phrases, and
argued that they project three separate functional layers: PossP, AGRP,
and DP. Both N and NP raisings have been motivated in the grammar of
Arabic DPs. The properties of AGR and D in Arabic nominals account
for the fact that their order is essentially limited to NSO (excluding any
significant scrambling or extraction of the subject or the object). The two
strategies used in genitive constructions have been shown to correlate with
different AGR and D properties. Synthetic genitives (like their analytic coun-
terparts) have been analyzed as DPs, not NPs. It is proposed that the
Principle of Elsewhere Licensing regulates the distribution of synthetic
and analytic options. DP raising and possessor extraction depend on whether
AGR is rich or not, and legitimate options observe the AGR Criterion. N
(to D) raising at S-structure has been motivated for Arabic, while DP raising
is precluded. A number of complementary distributions in the DP system
have been parametrized. Furthermore, the derivational properties of deverbal
process and subject nominals have been dealt with in the syntax. A nominalizing
affixation process converts a verbal root toN, at different levels in the tree.
Throughout the book I have defended a multi-valued functional view
of parametrization. According to this view, languages vary depending on
the properties of their various inflectional morphemes. The properties of
one morpheme may trigger a particular value of the parameter involved,
while those of another morpheme may trigger another value. One conse-
quence of this view is that languages may instantiate either the positive
value of a parameter, its negative value, or both. The latter option is
CONCLUSION 283
available because nothing precludes the co-occurrence in the same language
of inflectional morphemes having opposite triggering properties. Arabic, for
example, has been shown to have both pronominal and non-pronominal
AGR, as well as nominal and non-nominal AGR, depending on construc-
tions.
The rise of functional ambiguity has been analyzed as another con-
sequence of this view. In this case the same morpheme receives different
specifications (depending on contexts), and it triggers different gram-
matical options.
I have also shown that different strategies are used within the same
functional system, and that the choice of one strategy over another is neither
free nor arbitrary. On the contrary, it is regulated by general principles
like the Elsewhere Licensing Principle.
Clearly, this view of variation within the same language and across
languages may have significant consequences for language learning. The
picture which emerges is that languages exhibit a substantial variety of
options which arise from the acquisition of properties of inflectional
categories. But these options appear to be very limited indeed, especially
when faced with the potential 'vastness' of languages.
BIBLIOGRAPHY
284
BIBLIOGRAPHY 285
Barwise, John and Robin Cooper: 1981, 'Generalized Quantifiers and Natural Language',
Linguistics and Philosophy 4:2, 159-219.
Beedharn, C.: 1982, The Passive Aspect in English, German, and Russian, Gunter Narr Verlag,
Tiibingen.
Belletti, Adriana: 1982, "'Morphological" Passive and Pro Drop: The Impersonal Construction
in Italian', Journal of Linguistic Research 2:4, 1-24.
Belletti, Adriana: 1986/1988, 'Unaccusative as Case Assigners', Lexicon Project Working
Papers 8, Center for Cognitive Science, MIT.
Belletti, Adriana and Luigi Rizzi: 1988, 'Psych Verbs and Theta Theory', Natural Language
and Linguistic Theory 6:3, 291-352.
Benmamoun, Abbas: 1990, 'Inflectional Morphology', ms., USC, CA.
Bennis, Hans and Teun Hoekstra: 1988, 'The Tense Connection', GLOW Newsletter, Faris,
Dordrecht.
Benveniste, Emile: 1950, 'Actif et moyen dans le verbe', in Problemes de linguistique
generale, Gallimard, Paris, 1966, pp. 168-175.
Benveniste, Emile: 1952, 'La construction passive du parfait transitif', in Problemes de
linguistique generate, Gallimard, Paris, 1966, pp. 176--186.
Besten, Hans den: 1977-1983, 'On the Interaction of Root Transformations and Lexical
Deletive Rules', in Werner Abraham (ed.), On the Formal Synt(IJC of Westgermania, John
Benjamins, Amsterdam, pp. 47-131.
Blachere, Regis and Godefroy Demombynes: 1952, Grammaire de l'arabe classique, Adrien-
Maisonneuve, Paris.
Borer, Hagit: 1983, Parametric Synt(IJC, Faris, Dordrecht.
Borer, Hagit: 1984, 'The Projection Principle and Rules of Morphology', in Charles Jones
and Peter Sells (eds.), Proceedings of the Fourth Annual Meeting of NELS, Amherst,
pp. 16--33.
Borer, Hagit: 1986, '!-subjects', Linguistic Inquiry 17:3, 375-416.
Borer, Hagit: 1988, 'On the Morphological Parallelism Between Coumpounds and Constructs',
ms., Univ. of California, Irvine.
Bouchard, Denis: 1984, On the Content of Empty Categories, Faris, Dordrecht.
Brame, Michael: 1970, Arabic Phonology, Ph.D., MIT, Cambridge, Mass.
Brame, Michael: 1982, 'The Head-Selector Theory of Lexical Specifications and the
Nonexistence of Coarse Categories', Linguistic Analysis 10:3, 321-327.
Brame, Michael: 1983, 'Bound Anaphora Is Not a Relation Between NPs: Evidence for
Local Word Grammar (Without Trees)', Linguistic Analysis 11:2, 139-165.
Brandi, Luciana and Patricia Cardin: 1989, 'Two Italian Dialects and the Null Subject
Parameter', in Oswaldo Jaeggli and Kenneth Safir (eds.), pp. 111-142.
Bravman, Meyer: 1971, 'The origin of the Arabic object pronouns formed with ?iyyaa',
Journal of Semitic Studies, XVI: I pp. 50-52.
Bresnan, Joan W.: 1970, 'On complementizers: toward a syntactic theory of complement
types', Foundations of Language 6, pp. 297-321.
Bresnan, Joan W.: 1972, Theory of Complementation in English Syntax, Ph.D., MIT.
Bresnan, Joan W.: 1980, 'Polyadicity', in Teun Hoekstra et al. (eds): Lexical Grammar,
Faris Publications, Dordrecht, pp. 97-121.
Bresnan, Joan W.: 1982a, 'Control and Complementation', Linguistic Inquiry 13:3,343-434.
Bresnan, Joan W. (ed.): 1982b, The Mental Representation of Grammatical Relations, MIT
Press, Cambridge, Mass.
Bresnan, Joan W. and Sam A. Mchombo: 1987, 'Topic, Pronoun, and Agreement in
Chichewa', Language 63:4, 741-782.
Brody, Michael: 1984, 'On Contextual Definitions and the Role of Chains', Linguistic Inquiry
15:3, pp. 355-380.
Burzio, Luigi: 1986, Italian Synt(IJC, A GB Approach, D. Reidel, Dordrecht.
Bybee, Joan L.: 1985, Morphology, John Benjamins, Amsterdam.
286 BIBLIOGRAPHY
Bybee, Joan L. and Osten Dahl: 1990, 'The Creation of Tense and Aspect Systems in the
Languages of the World' ms., Univ. of Goteborg and Suny at Buffalo.
Carlson, Greg N.: 1987, 'The Semantic Composition of English Generic Sentences', ms.,
Univ. of Iowa.
Carter, Richard: 1984, 'Lecture Notes', ms., MIT.
Chambelland, Galves Charlotte: 1990, 'V movement, levels of representation, and the
structure of S', Paper delivered at the GLOW Conference, Cambridge-London.
Choe, Hyon-Sook: 1985, 'An SVO analysis of VSO languages and parametrization: A Study
of Berber', ms., MIT.
Chomsky, Noam: 1966, Cartesian Linguistics, Harper and Row, New York.
Chomsky, Noam: 1970, 'Remarks on Nominalizations', in Roderick A. Jacobs and Peter A.
Rosenbaum (eds.), Readings in English Transformational Grammar, Ginn, Waltham, Mass,
pp. 184-221.
Chomsky, Noam: 1973, 'On Wh Movement', in Peter Culicover et al. (eds.), Formal Syntax,
Academic Press, New York, pp. 71-132.
Chomsky, Noam: 1974, The Amherst Lectures, Documents Linguistiques, Universite de Paris
VII.
Chomsky, Noam: 1980, 'On Binding', Linguistic Inquiry 11:1, 71-132.
Chomsky, Noam: 1981, Lectures on Government and Binding, Faris Publications, Dordrecht.
Chomsky, Noam: 1982, Some Concepts and Consequences of the Theory of Binding, MIT
Press, Cambridge, Mass.
Chomsky, Noam: 1986a, Knowledge of Language, Praeger Publications, New York.
Chomsky, Noam: 1986b, Barriers, MIT Press, Cambridge, Mass.
Chomsky, No am: 1989a/1991, 'Some Notes on Economy of Derivation and Representation',
MIT WPL 10. Also in Robert Freidin (ed.): 1991, Principles and Parameters in
Comparative Grammar, MIT Press, Cambridge, Mass, pp. 417-454.
Chomsky, Noam: 1989b, 'Linguistics and Adjacent Fields: the State of the Art', ms.,
MIT.
Chomsky, Noam: 1989c, 'Prospects for the Study of Language and Mind', ms., MIT.
Chung, Sandra: 1982, 'Unbounded Dependencies in Chamorro Grammar', Linguistic Inquiry
13:1, 39-77.
Chung, Sandra: 1983, 'The ECP and Government in Chamorro', Natural Language and
Linguistic Theory 1:2, 207-244.
Chung, Sandra: 1990a, 'VPs and Verb Movement in Chamorro', Natural Language and
Linguistic Theory 8:4, 559-619.
Chung, Sandra: 1990b, 'Functional Heads and Proper Government in Chamorro', ms., Cowell
College, UCSC.
Chung, Sandra and Alan Timberlake: 1985, 'Tense, aspect, and mood', in Timothey Shopen
(ed.), Language Typology and Syntactic Description, Vol. III, Cambridge University Press,
Cambridge, pp. 202-258.
Chung, Sandra and James McCloskey: 1987, 'Government, Barriers, and Small Clauses in
Modern Irish', Linguistic Inquiry 18:2, 173-237.
Cinque, Guglielmo: 1978-80, 'On Extraction from NP in Italian', Journal of Italian Linguistics
5:1, 47-99.
Cinque, Guglielmo: 1984, 'A-bar Bound pro vs. Variable', ms., Univ. of Venice.
Cinque, Guglielmo: 1991, Types of A' Dependencies, MIT Press, Cambridge, Mass.
Cohen, David: 1989, L'aspect verbal, PUF, Paris.
Cohen, Marcel: 1924, Le systeme verbal semitique et /'expression du temps, Paris.
Comrie, Bernard: 1976, Aspect, Cambridge University Press, Cambridge.
Dahl, Osten: 1979, 'A Typology of Sentence Negation', Linguistics 17, 79-106.
Dahl, Osten: 1985, Tense and Aspect Systems, Basil Blackwell, Oxford.
Dahl, Osten and Fathi Talmoudi: 1987, 'Qad and laqad, Tense, Aspect, and Pragmatics in
Arabic', ms., Univ. of Goteborg.
BIBLIOGRAPHY 287
Davidson, David: 1966, 'The Logical Form of Action Sentences', in David Davidson (1980),
pp. 105-122.
Davidson, David: 1980, Essays on Actions and Events, Clarendon Press, Oxford.
Dechaine, Rose Marie: 1991, 'Bare Sentences', ms., Univ. of Mass., Amherst.
Demirdache, Hamida: 1989, 'Nominative NPs in Modern Standard Arabic', ms., MIT.
Descles, Jean Pierre: 1989, 'State, Event, Process, and Topology', General Linguistics 29:3,
159-198.
Di Sciullo, Anna Maria and Edwin Williams: 1987, On the Definition of Word, MIT Press,
Cambridge, Mass.
Dixon, R.M.W.: 1970, 'Where Have All the Adjectives Gone?' And Other Essays in Semantics
and Syntax, Mouton Publishers, The Hague.
Dobrovie-Sorin, Carmen: 1991, The Syntax of Romanian, to be published by Foris, Dordrecht.
Dowty, David: 1979, Word Meaning and Montague Grammar, D. Reidel, Dordrecht.
Emonds, Joseph: 1976, A Transformational Approach to English Syntax: Root, Structure
Preserving, and Local Transformations, Academic Press, New York.
Emonds, Joseph: 1978, 'The Verbal Complex V'-V in French', Linguistic Inquiry 9:2,
151-175.
Emonds, Joseph: 1980, 'Word Order in Generative Grammar', Journal of Linguistic Research
1:1, 33-54.
Emonds, Joseph: 1985, A Unified Theory of Syntactic Categories, Foris, Dordrecht.
En~. Miirvet: 1985, 'Temporal Interpretation', ms., USC.
En~, Miirvet: 1987, 'Anchoring Conditions for Tense', Linguistic Inquiry 18:4, 633-657.
En~. Miirvet: 1990, 'The Semantics of Specificity', Linguistic Inquiry 22:1, 1-25.
En~, Miirvet: 1991, 'On the Absence of the Present Tense Morpheme in English', Univ. of
Wisconsin, Madison.
Ettougani, NaYma: 1989, Aspect and Arabic passive participles, Third Cycle Thesis, Univ.
of Casablanca II.
Evans, Gareth: 1980, 'Pronouns', Linguistic Inquiry 11:2, 337-362.
Fabb, Nigel: 1984, Syntactic AffiXation, Ph.D., MIT, Cambridge, Mass.
Fassi Fehri, Abdelkader: 1976, 'Relatives et adjectifs en arabe: le probleme de Ia determi-
nation', Lingua 38, pp. 125-152.
Fassi Fehri, Abdelkader: 1980, 'Some Complement Phenomena in Arabic, the Complementizer
Phrase Hypothesis and the Non-Accessibility Condition', Analyse!Thiorie, Universite
de Paris-VIII Vincennes, 54-114.
Fassi Fehri, Abdelkader: 1981, Linguistique arabe: forme et interpretation, Publications of
the Faculty of Letters, Rabat.
Fassi Fehri, Abdelkader: 1984-1988, 'Agreement, Binding and Coherence', in Michael Barlow
and Charles Ferguson (eds.), Agreement in Natural Language, Approaches, Theories,
Descriptions, CSLI, Stanford, pp. 107-158.
Fassi Fehri, Abdelkader: 1985a, 'Word order variation in Standard Arabic', Lecture
delivered at MIT, McGill Univ., and the Univ. of Utah.
Fassi Fehri, Abdelkader: 1985b, l-lisaaniyaat wa lluga 1-<arabiyya, Tubqal Publishers,
Casablanca.
Fassi Fehri, Abdelkader: 1986a, l-mu<jam f.Carabii, Tubqal Publishers, Casablanca.
Fassi Fehri, Abdelkader: 1986b, 'Anti-Causatives in Arabic, Causativity and Affectedness',
Lexicon Project Working Papers 15, Center for Cognitive Science, MIT, Cambridge, Mass.
Fassi Fehri, Abdelkader: 1986c-1988, 'A propos du conceptuel et du grammatical', Hommage
a Bernard Pottier, Klinksieck, Paris.
Fassi Fehri, Abdelkader: 1987a, 'Case, Inflection, VS Word Order and X' Theory',
Proceedings of the First International Conference of the Linguistic Society of Morocco,
Vol. 1, Oukad Publishers, Rabat, pp. 189-221.
Fassi Fehri, Abdelkader: 1987b, 'Category changing affixes and Homonymy', ms., Faculty
of Letters, Rabat.
288 BIBLIOGRAPHY
Fassi Fehri, Abdelkader: 1988a, 'Cognate objects as arguments', ms., Faculty of Letters, Rabat.
Fassi Fehri, Abdelkader: 1988b, 'Arabic Passive Affixes as Aspectual Predicates', ms.,
MIT, Cambridge, & Faculty of Letters, Rabat.
Fassi Fehri, Abdelkader: 1988c, 'On Pleonastics in Arabic', in Jochen Pleines (ed.), La
linguistique au Maghreb, Oukad Publishers, Rabat.
Fassi Fehri, Abdelkader: 1989a, 'Agreement, Incorporation, Pleonastics, and VSO-SVO order',
ms., MIT.
Fassi Fehri, Abdelkader: 1989b, al-binaaP l-muwaazii, Oukad Publishers, Rabat.
Fassi Fehri, Abdelkader: 1989c, Inflectional Categories in Arabic, ms., Faculty of Letters,
Rabat.
Fassi Fehri, Abdelkader: 1989d, 'Temporal reference, finiteness, and the internal structure
of S in Arabic', ms., Faculty of Letters, Rabat.
Fassi Fehri, Abdelkader: 1990, 'Arabic AGR, Incorporation, and Null Anaphora', Paper
delivered at the GLOW Workshop, Cambridge/London.
Ferguson, Charles A.: 1984, 'Grammatical Agreement in Classical Arabic and the
colloquial dialects', Paper presented at the meeting of the Middle Eastern Studies
Association of North America.
Ferguson, Charles A. and Michael Barlow: 1988, 'Introduction', in Michael Barlow and
Charles A. Ferguson (eds.), pp. 1-22.
Fleisch, Henri: 1958, 'Etudes sur le verbe arabe', Melanges de l'Universite Saint-Joseph,
Beyruth, pp. 153-181.
Foley, Wiliam A. and Robert van Valin: 1984, Functional Syntax and Universal Grammar,
Cambridge University Press, Cambridge.
Frampton, John: 1990, 'Parasitic Gaps and the Theory of Wh-Chains', Linguistic Inquiry 21:1,
49-77.
Fukui, Naoki: 1986, A Theory of Category Projection and its Applications, Ph.D., MIT,
Cambridge, Mass.
Fukui, Naoki and Margaret Speas: 1986, 'Specifiers and Projections', MITWPL 8, pp. 128-176.
Giorgi, Alessandra and Giuseppe Longobardi: 1987/1991, The Syntax of Noun Phrases,
Cambridge University Press, Cambridge.
Givan, Talmy: 1976, 'Topic, Pronoun, and Grammatical Agreement', in Charles N. Li (ed.),
Subject and Topic, Academic Press, New York, pp. 149-188.
Givan, Talmy: 1984, Syntax: A Functional-Typological Introduction, John Benjamins,
Amsterdam.
Godard, Danielle, 1990, 'Extraction out of NP in French', to be published by Natural
Language and Linguistic Theory.
Greenberg, Joseph: 1966, 'Some universals of grammar with particular reference to the
order of the meaningful elements', in Joseph Greenberg (ed.), Universals of Language,
MIT Press, Cambridge, Mass, pp.73-113.
Grimshaw, Jane: 1979, 'Complement Selection and the Lexicon', Linguistic Inquiry 10:2,
279-326.
Grimshaw, Jane: 1981, 'Form, Function, and the Language Acquisition Device', in Lee Baker
and John McCarthy (eds.), The Logical Problem of Language Acquisition, MIT Press,
Cambridge.
Grimshaw, Jane: 1986-1988, 'Adjuncts and Argument Structure', Lexicon Project Working
Papers 21, Center for Cognitive Science, MIT, Cambridge.
Grimshaw, Jane: 1987, 'Psych Verbs and the Structure of Argument Structure', ms., Brandeis
University.
Grimshaw, Jane: 1990, Argument Structure, MIT Press, Cambridge, Mass.
Grimshaw, Jane: 1991, 'Extended Projection', Glow Newsletter 26, Faris Dordrecht, pp.
20-21.
Gruber, JeffreyS.: 1965-1976, Lexical Structures in Syntax and Semantics, North Holland,
Amsterdam.
BIBLIOGRAPHY 289
Holmberg, Andreas: 1991, 'The Distribution of Scandinavian Weak Pronouns', ms. Univ.
of Uppsala.
Hornstein, Norbert: 1981, 'The Study of Meaning in Natural Language: Three Approaches
to Tense', in Norbert Hornstein and David Lightfoot (eds.), Explanation in Linguistics,
Longman, London, pp. 116-151.
Hornstein, Norbert: 1990, As Time Goes By, Tense and Universal Grammar, MIT Press,
Cambridge, Mass.
Horrocks, Geoffrey: 1987, 'Movement rules within "Noun Phrases'", Proceedings of the First
International Conference of the Linguistic Society of Morocco, Vol. I, Oukad Publishers,
Rabat, pp. 97-105.
Horrocks, Geoffrey and Melita Stavrou: 1987, 'Bounding theory and Greek syntax: evidence
for wh-movement in NP', Journal of Linguistics 23, 77-108.
Huang, James: 1982, Logical Relations in Chinese and the Theory of Grammar, Ph.D.,
MIT.
Ibn 'Aqiil, Bahaa?-u-d-diin A.: 14th century, Sar/J ?alfiyyat Ibn Maalik, daar al-fikr, Cairo,
1979.
Ibn Ma9aa?, Abuu-1-'Abbaas A., 12th century, Kitaab r-radd 'alaa n-nu/Jaat, Shawqi I;>ayf
ed., Cairo, 1947.
Ibn as-Sarraj, Abuu Bakr M.: 9th century, ?usuul n-nafJw, Fatii, M. (ed.), mu?assasat
ar-risaala, Beyruth, 1985.
Ibn Ya'm, Abuu-1-baqaa? M.: 12th century, SarfJ al-mufaHal, daar Hibba'a 1-muniiriyya,
Cairo.
Jackendoff, Ray S.: 1972, Semantic Interpretation in Generative Grammar, MIT Press,
Cambridge, Mass.
Jackendoff, Ray S.: 1975, 'Morphological and Semantic Regularities in the Lexicon',
Language 51:4, 639-671.
Jackendoff, Ray S.: 1977, X' Syntax, MIT Press, Cambridge, Mass.
Jackendoff, RayS.: 1983, Semantics and Cognition, MIT Press, Cambridge, Mass.
Jaeggli, Oswaldo: 1982, Topics in Romance Syntax, Foris Publications, Dordrecht.
Jaeggli, Oswaldo: 1986, 'Passive', Linguistic Inquiry 17:4, 587-622.
Jaeggli, Oswaldo and Kenneth Safir (eds.): 1989, The Null Subject Parameter, Kluwer
Academic Publishers, Dordrecht.
Jakubowicz, Celia: 1985, 'Do Binding Principles Apply to !NFL?', Proceedings of NELS
15, 188-206.
Kamp, Hans: 1979, 'Events, Instants, and Temporal Reference', in Rainer Bauerle et al. (eds.),
Semantics from Different Points of View, Springer-Verlag, Berlin, pp. 376-471.
Kamp, Hans: 1981, 'Ev6nements, repr6sentation discursive et r6f6rence temporelle', Langages
64, 39-64.
Kamp, Hans and Christian Rohrer: 1983, 'Tense in Texts', in Rainer Bauerle et al. (eds.),
Meaning, Use, and Interpretation of Language, W. de Gruyter, Berlin, 250-269.
Kamp, Hans and Christian Rohrer: 1989, 'A Discourse Representation Theory of Tense in
French', ms., Univ. of Stuttgart.
Kayne Richard: 1975, French Syntax, MIT Press, Cambridge, Mass.
Kayne, Richard: 1984, Connectedness and Binary Branching, Foris, Dordrecht.
Kayne, Richard: 1987-1989, 'Facets of Romance past participle agreement', in Paola
Beninca (ed.), Dialect Variation and the Theory of Grammar, Foris, Dordrecht, pp. 85-
103.
Kayne, Richard: 1990, 'Romance Clitics and PRO', NELS 20:2, Carnegie Mellon Univ. and
Univ. of Pittsburg, pp. 85-103.
Keenan, Edward L. and Jonathan Stavi: 1986, 'A Semantic Characterization of Natural
Language Determiners', Linguistics and Philosophy 9, 253-326.
Kiefer, Ferenc: 1982, 'The Aspectual System of Hungarian', in Ferenc Kiefer (ed.), Hungarian
Linguistics, John Benjamins, Amsterdam.
BIBLIOGRAPHY 291
Kiparsky, Paul: 1985, Morphology and Grammatical Relations, ms., CSLI, Stanford
University.
Kitagawa, Yoshihisa: 1986, Subjects in Japanese and English, Ph.D., Univ. of Mass.,
Amherst.
Koopman, Hilda: 1984, The Syntax of Verbs, Foris, Dordrecht.
Koopman, Hilda and Dominique Sportiche: 1982, 'Variables and the Bijection Principle', The
Linguistic Review 2, 139-160.
Koopman, Hilda and Dominique Sportiche: 1988, 'Subjects', ms., UCLA.
Kuno, Susumo: 1987, Functional Syntax, Anaphora, Discourse, and Empathy, The Univ. of
Chicago Press, Chicago.
Kratzer, Angelika: 1989, 'Stage-Level and Individual-Level Predicates', ms., Univ. of Mass.,
Amherst.
Kroeber, A.!. and George W. Grace: 1960, The Sparkman Grammar of Luiseno, Univ. of Calif.
Press, Berkley and LA.
Kuroda, Sige-Yuki: 1986, 'Whether We Agree or Not', Linguisticae Investigaciones 12:1,
1-47.
Kurylowicz, Jerzy: 1973, Studies in Semitic Grammar and Metrics, Curzon Press,
London.
Laka, Itziar: 1990, Negation in Syntax, Ph.D., MIT.
Langacker, Ronald: 1969, 'On Pronominalization and the Chain of Command', in David A.
Reibel and Sanford A. Schane (eds.), Modern Studies in English, Prentice Hall, Englewood
Cliffs, New Jersey, pp. 160-186.
Langacker, Ronald: 1982, Studies in Uteco-Aztecan Grammar, Vol. 3, Publications of the
Summer Institute and the University of Texas at Arlington.
Lapointe, Steven: 1981, 'General and restricted agreement phenomena', in Michael Moortgat
et al. (eds.), The scope of lexical rules, Foris, Dordrecht, pp. 125-159.
Larson, Richard K.: 1988, 'On the Double Object Construction', Linguistic Inquiry 19:3,
335-391.
Lasnik, Howard: 1976, 'Remarks on Coreference', Linguistic Analysis 2:1, 1-22.
Lasnik, Howard: 1981, 'Restricting the Theory of Transformations: A Case Study', in Norbert
Hornstein and David Lightfoot (eds.), Explanation in Linguistics, Longman, London,
pp. 152-173.
Lasnik, Howard: 1987, 'Subjects and the Theta Criterion', Natural Language and Linguistic
Theory 6:1, 1-17.
Lasnik, Howard: 1989a, Essays on Anaphora, Kluwer Academic Publishers, Dordrecht.
Lasnik, Howard: 1989b, 'Case and Expletives', ms., Univ. of Connecticut, Storrs.
Lasnik, Howard and Mamuro Saito: 1984, 'On the Nature of Proper Government', Linguistic
Inquiry 15:2, 235-289.
Levin, Beth (ed.): 1985, 'Lexical Semantics in Review', Lexicon Project Working Papers
1, Center for Cognitive Science, MIT.
Levin, Beth and Maika Rapaport: 1986, 'The Formation of Adjectival Passives', Linguistic
Inquiry 17:4, 623-661.
Lieber, Rochelle: 1983, 'Argument Linking and Compounds in English', Linguistic Inquiry
14:2, 251-285.
LoCascio, Vincenzo: 1985, 'Temporal Deixis and Anaphor in Sentence and Text: Finding
a Reference Time', in Vincenzo Lo Cascio and Co Vet (eds.), Temporal Structure in
Sentence and Discourse, Foris, Dordrecht, pp. 229-249.
Lyons, John: 1977, Semantics, 2 Vol., Cambridge University Press, Cambridge.
Mahajan, Anoop K.: 1989, 'Agreement and Agreement Phrases', MITWPL 10, Cambridge,
Mass.
Mahajan, Anoop K.: 1990, The A/A-Bar Distinction And Movement Theory, Ph.D., MIT,
Cambridge, Mass.
Maling, Joan: 1988, 'Variations on a Theme: Existential Sentences in Swedish and Icelandic',
292 BIBLIOGRAPHY
Ouhalla, Jamal: 1990, 'Sentential Negation, Relativized Minimality, and the Aspectual
Status of Auxiliaries, The Linguistic Review 7:2, pp. 183-231.
Quirk Randolph et al. (eds.): 1985, A Comprehensive Grammar of the English Language,
Longman, London.
Palmer, Frank R.: 1985, Mood and Modality, Cambridge University Press, Cambridge.
Partee, Barbara Hall: 1973, 'Some Structural Analogies between Tenses and Pronouns in
English', The Journal of Philosophy 10, 601-609.
Partee, Barbara Hall: 1984, 'Nominal and temporal anaphora', Linguistics and Philosophy
7:2, 243-286.
Parsons, Terence: 1990, Events in the Semantics of English, MIT Press, Cambridge.
Pelletier, F. J. and L. K. Schubert: 1987, Three Papers on the Logical Form of Mass Terms,
Generics, Bare Plurals, and Habituals, ms., Univ. of Alberta, Canada.
Penner, Zvi: 1988, The Grammar of the Nominal Sentence, Institut fiir Sprachwissenschaft,
University of Bern.
Perlmutter, David: 1978, 'Impersonal Passive and the Unaccusative Hypothesis', BLS 4, Univ.
of CA, 157-189.
Perlmutter, David and Paul Postal: 1983, 'Some Proposed Laws of Basic Clause Structure',
in David Perlmutter (ed.), Studies in Relational Grammar 1, University of Chicago Press,
Chicago.
Perlmutter, David and Paul Postal: 1984, 'The !-Advancement Exclusiveness Law' in David
Perlmutter and Carol Rosen (eds.), Studies in Relational Grammar 2, The University of
Chicago Press, Chicago.
Pesetsky, David: 1982, Paths and Categories, Ph.D., MIT.
Pesetsky, David: 1985, 'Morphology and Logical Form', Linguistic Inquiry 16:2, 193-
246.
Pesetsky, David: 1987, 'Binding Problems with Experiencer Verbs', Linguistic Inquiry 18:1,
126-140.
Picallo, Carme: 1984, 'The Infl Node and the Null Subject Parameter', Linguistic Inquiry
15:1, 75-102.
Platzack, Christer: 1983, 'Existential Sentences in English, Swedish, German, and Icelandic',
in F. Karlsson (ed.), Papers from the Seventh Scandinavian Conference of Linguistics,
Helsinki, pp. 80-100.
Platzack, Christer and Andreas Holmberg: 1990, 'The Role of AGR and Finiteness in some
VO Languages', ms. Univ. of Lund and Uppsala.
Pollock, Jean Yves: 1983a, 'Sur quelques propri~t~s des phrases copulatives en fran~ais',
Langue Franr;aise 58, pp. 89-105.
Pollock, Jean Yves: 1983b, 'Accord, chaines impersonnelles et variables', Linguisticae
Investigationes VII:l, pp. 131-181.
Pollock, Jean Yves: 1988, 'Extraction from NP in French and English: A Case Study in
Comparative Syntax', ms., Univ. de Rennes 2.
Pollock, Jean Yves: 1989a, 'Verb Movement, Universal Grammar, and the Structure of IP',
Linguistic Inquiry 20:3, 365-424.
Pollock, Jean Yves: 1989b, 'Opacity, Genitive Subjects, and Extraction from NP in English
and French', Probus 1:2, 151-162.
Postal, Paul: 1966-1969, 'On So-Called Pronouns in English', in David Reibel and Sanford
Schane (eds.), pp. 201-224.
Pottier, Bernard: 1974, Linguistique generate: theorie et description, Klinksieck, Paris.
Pranka, Paula: 1982, Syntax and Word Formation, Ph.D., MIT.
Radford, Andrew: 1988a, Transformational Syntax, Cambridge Univ. Press, Cambridge.
Radford, Andrew: 1988b, 'Small Children's Small Clauses', Transactions of the Philological
Society 86, 1-46.
Radford, Andrew: 1990, Syntactic Theory and the Acquisition of English Syntax: the Nature
of Early Child Grammars of English, Basil Blackwell, Oxford.
294 BIBLIOGRAPHY
Rapaport, Tova: 1987, Copular, Nominal, and Small Clauses: A Study of Israeli Hebrew,
PH.D., MIT.
Raposo, Eduardo: 1985-6, 'Some asymmetries in the binding theory in Romance', The
Linguistic Review 5:1, 75-110.
Raposo, Eduardo: 1987, 'Case Theory and Infl-to-Comp: the Inflected Infinitive in European
Portuguese', Linguistic Inquiry 18:1, 85-109.
Reckendorf Hans: 1895, Die syntaktischen Verhiiltnisse des arabischen, Leiden.
Reichenbach, Hans: 1947, Elements of Symbolic Logic, Macmillan, New York.
Reinhart, Tanya: 1976, The Syntactic Domain of Anaphora, Ph.D., MIT.
Reinhart, Tanya: 1983, Anaphora and Semantic Interpretation, The University of Chicago
Press, Chicago.
Reinhart, Tanya and Eric Reuland: 1990, 'Anaphors and Logophors: An Argument Structure
Perspective', in Jan Koster and Eric Reuland (eds.), Long Distance Anaphora, to be
published by Faris, Dordrecht.
Reuland, Eric: 1983, 'Governing -ing', Linguistic Inquiry 14:1, 101-136.
Reuland, Eric and Jan Koster: 1990, 'Long Distance Anaphora: An Overview', in Jan Koster
and Eric Reuland (eds.).
Ritter, Betsy: 1987-1988, 'A Head Movement Approach to Construct State Noun Phrases',
Linguistics 26:6, 909-929.
Ritter, Betsy: 1988, A Case Study in the Syntax of Agreement: Hebrew Noun Phrases and
Benoni Verb Phrases, Ph.D., MIT.
Ritter, Betsy: 1991, 'Evidence for Number as a Nominal Head', Paper delivered at the GLOW
conference, Leiden.
Rizzi, Luigi: 1982, Issues in Italian Syntax, Faris, Dordrecht.
Rizzi, Luigi: 1986, 'Null Objects in Italian and the Theory of pro', Linguistic Inquiry, 17:3,
501-557.
Rizzi, Luigi: 1987, 'On Chain Formation', in Hagit Borer (ed.), pp.65-95.
Rizzi, Luigi: 1990, Relativized Minimality, MIT Press, Cambridge, Mass.
Rizzi, Luigi: 1991a, 'Proper Government and the Definition of A Positions', Paper
delivered at GLOW, Leiden.
Rizzi, Luigi: 199lb, 'Residual Verb Second and the Wh-Criterion', Technical Reports in
Formal and Computational Linguistics, Universite de Geneve.
Rizzi, Luigi and Ian Roberts: 1989, 'Complex Inversion in French', Probus 1:1, 1-30.
Roberts, Ian: 1985 The Representation of Implicit and Dethematized Subjects, Doctoral
dissertation, USC, LA.
Rochemont, Michael: 1978, A Theory of Stylistic Rules, Ph.D., Univ. of Mass., Amherst.
Roeper, Tom: 1986, 'Implicit Arguments and Implicit Roles', ms., Univ. of Mass., Amherst.
Roeper, Tom: 1987, 'Implicit Arguments and the Head-Complement Selection', Linguistic
Inquiry 18:2, 267-310.
Roman, Andre: 1990a, Grammaire de l'arabe, Presses Universitaires de France, Paris.
Roman, Andre: 199Gb, 'De !'accord et du pseudo-accord du feminin en arabe', Annates
Islamologiques 25, pp. 27-56.
Ross, John R.: 1967, Constraints on Variables in Syntax, Ph.D., MIT, Cambridge, Mass.
Rothstein, Susan: 1983, The Syntactic Forms of Predication, Ph.D., MIT.
Rouveret, Alain: 1989, 'Cliticisation et Temps en portugais europeen', Revue des langues
romanes XCIII:2, 337-371.
Rouveret, Alain: 1990, 'X-Bar Theory, Minimality, and Barrierhood in Welsh', in Randall
Hendrick (ed.), pp. 27-79.
Rouveret, Alain: 1991, 'Functional Categories and Agreement', ms., Paris VIII Saint Denis
(to appear).
Rouveret, Alain and Jean Roger Vergnaud: 1980, 'Specifynig Reference to Subject', Linguistic
Inquiry 11:1, 97-202.
Russell, Robert A.: 1984, 'Historical aspects of subject-verb agreement in Arabic',
BIBLIOGRAPHY 295
Proceedings of the 1st Eastern States Conference on Linguistics, Ohio State Univ.,
Columbus, pp. 116-127.
Ruwet, Nicolas: 1972, Thiorie syntaxique et syntaxe du franrais, Le Seuil, Paris.
Ruwet, Nicolas: 1982, Grammaire des insultes et autres etudes, Le Seuil, Paris.
Saad, George N.: 1982, Transitivity, Causation and Passivisation, Kegan Paul International,
London.
Safir, Kenneth: 1986, 'On Implicit Arguments and Thematic Structure', NELS 16.
Safrr, Kenneth: 1987a, 'Subject Clitics and the Nom Drop Parameter', in Hagit Borer (ed.),
The Syntax of Pronominal Clitics, Academic Press, New York, pp. 333-356.
Safir, Kenneth: 1987b, 'The Syntactic Projection of Lexical Thematic Structure', Natural
Language and Linguistic Theory 5:4, 561-601.
Saito, Mamuro: 1985, Some Asymmetries in Japanese and their Theoretical Implications,
PH.D., MIT.
Samarra?ii, Ibrahim: 1968, n-na}Jw ar-<arabii, naqd-un wa-binaa?, daar as-saadiq, Beyruth.
Schachter, Paul: 1983, 'Explaining Auxiliary Order', in Frank Heny and Barry Richards (eds.),
pp. 145-204.
Selkirk, Lisa: 1982, The Syntax of Words, MIT Press, Cambridge, Mass.
Seghrouchni, Driss: 1989, Les schemes verbaux en arabe, ms., Faculty of Letters, Rabat.
Shlonsky, Ur: 1987, Null and Displaced Subjects, Ph.D., MIT.
Shlonsky, Ur: 1990, 'Construct State Nominals and Universal Grammar', Paper delivered
at the GLOW Workshop, London.
Siibawayhi, Abuu Bi§r Q.: (8th century), al-kitaab, Bulaaq, Cairo, 1938.
Smith, Carlota: 1986, 'A Speaker-Based Approach to Aspect', Linguistics and Philosophy
9, 97-115.
Smith, Carlota: 1991, The Parameter of Aspect, to be published by Kluwer Academic
Publishers, Dordrecht.
Sobin, Nicholas J.: 1985, 'On Case Assignment in Ukrainian Morphological Passive
Constructions', Linguistic Inquiry 16:4, 649-662.
Speas, Margaret: 1986, Adjunctions and Projections in Syntax, Ph.D., MIT, Cambridge, Mass.
Sportiche, Dominique: 1988, 'A Theory of Floating Quantifiers and its Corollaries for
Constituent Structure', Linguistic Inquiry 19:3, 425-450.
Sproat, Richard: 1985a, 'Welsh Syntax and VSO Structure', Natural Language and Linguistic
Theory 3:2, 173-216.
Sproat, Richard: 1985b, Deriving the Lexicon, Ph.D., MIT, Cambridge, Mass.
Stowell, Tim: 1981, Origins of Phrase Structure, Ph.D., MIT, Cambridge, Mass.
Stowell, Tim: 1982, 'The tense of infinitives', Linguistic Inquiry 13:3, 561-570.
Suyuu\ii, Jallal-uddiin: 15th century, ham< al-hawaami", daar al-buiJuu\ al-<ilmiyya, 1977.
Suyuu\ii, Jallal-uddiin: 15th century,l-?Sbaah wa-n-nat}aa?ir, daar 1-kitaab al-<arabii, Beyruth,
1984. .
Szabolcsi, Anna: 1987, 'Functional Categories in the Noun Phrase', in lstvhn Kenesei (ed.),
Approaches to Hungarian, Vol. 2, Jate Szeged, Institute for Science Management and
Informatics, Budapest, pp. 167-189.
Szabolcsi, Anna: 1989, 'Noun Phrases and Clauses: is DP analogous to IP or CP?', to
appear in John Payne (ed.), Proceedings of the Colloquium on Noun Phrase Structure,
Mouton, The Hague.
Taraldsen, Tarald: 1978, 'On the NIC, vacuous application, and the that-trace filiter',
distributed by Indiana Univ. Linguistics Club.
Tellier, Christine: 1990, 'Underived Nominals and the Projection Principle', NELS 20, pp.
472-486.
Tenny, Carol: 1987, Grammatical Aspect and Affectedness, Ph.D., MIT.
Torrego, Esther: 1984, 'On Inversion in Spanish and Some of its Effects', Linguistic Inquiry,
15:1, 103-130.
Torrego, Esther: 1988, 'Evidence for DPs', ms., Boston Univ., Mass.
296 BIBLIOGRAPHY
Travis, Lisa: 1984, Parameters and Effects of Word Order Variation, Ph.D., MIT, Cambridge,
Mass.
Underhill, Robert: 1976, Turkish Grammar, MIT Press, Cambridge, Mass.
Vendler, Zeno: 1967, Linguistics and Philosophy, Cornell Univ. Press, Ithaca, New York.
Vergnaud, Jean Roger: 1985, Dependances et Niveaux de Representation en Syntaxe, John
Benjamins, Amsterdam.
Verkuyl, Henk J.: 1972, On the Compositional Nature of the Aspects, D. Reidel Publishing
Co., Dordrecht.
Vinet, Marie-Therese: 1987, 'Empty pleonastics and a parametrized INFL', ms. University
of Sherbrooke.
Vinet, Marie-Therese: 1991, 'Observations theoriques sur les expletifs: le cas du creole
ha'itien', Revue quebecoise de linguistique 20:1, pp. 195-213.
Vlach, Frank: 1981, 'The Semantics of the Progressive', in Philip J. Tedeschi and Annie
Zaenen (eds.): Tense and Aspect, Academic Press, N.Y., pp. 271-292.
Wasow, Thomas: 1979, Anaphora in Generative Grammar, Ghent: E. Story-Scientia.
Weerman, Fred: 1989, The V2 Conspiracy, Foris, Dordrecht.
Wehrli, Eric: 1987, 'On Some Properties of French Clitic Se', in Hagit Borer (ed.), pp.
263-283.
Williams, Edwin: 1980, 'Predication', Linguistic Inquiry 11: 1, 203-238.
Williams, Edwin: 1981, 'Argument Structure and Morphology', The Linguistic Review I: 1,
81-114.
Williams, Edwin: 1985, 'PRO and Subject of NP', Natural Language and Linguistic Theory
3:3, 297-315.
Woisetschlaeger, Eric F.: 1976, A Semantic Theory of the English Auxiliary System, Indiana
Univ. Linguistics Club, Bloomington, Indiana.
Woisetschlaeger, Eric F.: 1983, 'On the Question of Definiteness in "An Old Man's Book"',
Linguistic Inquiry 14:1, 137-154.
Wright, William: 1858/1974, A Grammar of the Arabic Language, third edition, 2 Vols.,
Cambridge University Press, Cambridge.
Yafei, Li: 1990, 'On V-V Compouds in Chinese', Natural Language and Linguistic Theory
8:2, 177-207.
Zaenen, Annie: 1980, Extraction Rules in Icelandic, Ph.D. Thesis, Harvard University, Garland
Publishing Co., New York.
Zaenen, Annie, Joan Maling, and Hoskuldur Thrainsson: 1985, 'Case and Grammatical
Functions: The Icelandic Passive', Natural Language and Linguistic Theory 3:3, 441-483.
Zagona, Karen: 1982/1988, Verb Phrase Syntax, A Parametric Study of English and Spanish,
Kluwer Academic Publishers, Dordrecht.
Zajjajii, Abuu-1-qaasim A.: lOth century, al·?iidaah fii cilal-i n-nal]w, daar n-nafaa?is,
Beyruth, 1979.
Zanuttini, Raffaella: 1989, 'The structure of negative clauses in Romance', NELS 20, pp.
517-530.
Zribi-Hertz, Anne: 1982, 'La construction se-moyen du fran~ais et son statut dans le triangle:
moyen, pass if, reflechi', Linguisticae Investigationes VI:2, pp. 345-401.
Zubizaretta, Maria Luisa: 1987, Levels of Representation in the Lexicon and in the Syntax,
Faris, Dordrecht.
INDEX OF NAMES
297
298 INDEX OF NAMES
301
302 ANALYTIC INDEX
110 Kase 53
Subject 116 Criterion 207n 19
Incorporating discharge 94n45, 207n19
language 116, 125, 127, 139n29 phrase, KP 190, 207n19
Indefinite 216, 219, 225,245,253, 255, Kase Resistance Principle 53, 158
256,269,277n24 Kinds
Indicative of sentences 180, 181
mood 145, 151, 153, 164, 170 of situations 145
Individual level predicate 210n34
Inflection, I, INFL 96, 97, 98, 108, 110, Language
116, 117, 121, 122, 124, 127, 130, acquisition xii, 2, 15n2
137n15&17&21, 142, 161 faculty 1
Pure 97, 116, 134 learning 15n1&2
pronoun alternation 97 Lasnik's Filter 19, 100
Rich 109, 122, 135, 139n31 Left
Split analysis 142, 205n10 dislocation 27, 28, 40, 64, 66
Inflectional Legitimate object 94n45
AGR 133 Lexical 48, 72
agreement 96 aspect 143, 186
agreement marker 96, 109, 112 category xii, 4, 5, 7, 10, 77, 78, 79,
analysis, INFLA 96, 97, 109-116, 95n58,279n34
127, 135, 139n29, 140n35 expletive 41, 57, 58
category, i-category 5, 26, 27, 47, 62, head 239, 260
75,79,82-84,213 projection 220, 241, 274, 278n29
projection 5, 124, 213, 274 redundancy rule 76
structure 142, 230, 274 root 79, 184
subject 12 support 47, 48, 156
support 12 Lexicalism xiii, 76-77, 85n6
Inherent Lexicalization 47, 58, 89n2, 152
case 8, 61, 200 Licensing 17, 18, 19, 34, 37, 38, 39, 41,
Initial stage 143 42,43,44,48,51,52,53,62, 74,
Interval 75,86n8,89n5,91n23,
oftime 142, 146 93n32&34,94n48,214,215,220,
Invariant system xii 223,228,229,235,250,252,257,
Irish 62, 63, 89n6, 90n14, 92n27, 94n53, 270,271,278n29,279n34
96, 125, 136n13, 137n17 Expletive 19, 40, 53-61
!-subject 94n48&50 strategy vii
ltalian96,97, 109,110,112,125,134, Light 54, 70, 71
135, 136n2, 137n15, 139n29, Linear 149, 190
140n35,251,279n34 morphology 76
Iterative 176, 183 order45
process 184 precedence 22
Ixtisaas 207n22 Listedness 86n7
L-marking 10, 65, 66, 67, 95n58
Japanese 1, 5, 139n33 Local
Jussive 84, 151, 161, 164, 169, 172, domain 156
205n8 tense domain 147
Location 141
308 ANALYTIC INDEX
checking 49, 54, 55, 62, 75, 137n14 Parameter xi, xii, 2
discharge 55 see also Argumental, Nominal, Null
Non-anterior 170, 184 subject, Pronominal
Non-concatenative 82 Participant 143
Non-contingent state 176 Participle xiii, 141, 142, 175-190,
Non-cumulative 82 194-203
Non-finite 158 Active 177-181, 191, 192, 194-203,
morphology 191 2lln40
object 83 Passive 176-178, 183, 184, 194-203,
T 191 210n33
Non-linear 82 Partitive case 198-201
Non-nominative 98, 99, 100, 104, 105, Passive 183, 184, 276n20; see also
110,120-122,12- 130, 135n11, Participle
137nl4 Past 141, 144-147, 152-155, 159-163,
Non-past 141, 144, 145, 147, 181 169-171, 174, 175, 181, 182, 191,
Non-preterite 146 193,205n8,209n2 7
Non-stativity 176, 179 imperfect 148, 162, 205n8
Northern Paiute 273 imperfective 204n3
Notional type 86n8, 179 perfect 83, 84, 141, 148, 191, 192,
N-to-D raising 213, 216, 222 209n27
Noun, N 299, 329, 359nl8 perfective 143, 204n3
Noun phrase, NP progressive 192
movement 12, 62, 63 tense 149-150, 192
raising xiii, 213,215,223,227 ,228, Patient nominal 266, 268, 269
252,270,276n20 Perfect 141, 144, 148, 151, 162, 169,
Novelty Condition 212n43 170, 173,205n7
NSO 213, 216, 222, 224, 276 tense 149
Null Perfective 76, 78, 83, 91n23, 143, 144,
anaphor 137n22, 139n28 176, 181, 190, 204n7
argument 96, 97, 109, 112, 114, viewpoint 143
136n1&9, 138n26 Perfectivity 143, 205n7
argument language xi, 136n2 Permanent
subject 96, 112 property 184
subject language 96, 97, 109, 136n2, situation 178, 180
139n29 state 210n34
Number, NUM 31, 32, 35, 36, 39, 40, 41, Person, PERS 31, 52, 91n23, 104, 109,
42,43,93n23, 109,117,118, 117-127, 131, 132
l38n26 activation 132
activation 43 checking 132
licensing 41, 43 Constraint 104-105
specification 16, 37, 43 matching 132
Nunation 216, 219, 273, 275n4&5&10, specification 126, 131
277n24 Personal pronoun 57, 116-119, 122, 123,
132, 204n6
ONS 223 Phi-feature 96, 97, 113, 116, 120, 122,
Operator 95n61, 190, 193 135
ovs 213,223 checking 132
hierarchy 131
310 ANALYTIC INDEX
Quantifier 28, 73,74 Saturation 11, 95n61, 117, 123, 124, 193
Floating 71, 73 Scandinavian 90n11
Quasi argument 118 Scope 12
Question 64, 65, 67,68 of affixation 176, 267, 274
bearer 193
Raising 17, 18, 56, 59, 89n5, 213,216, Scrambling 20, 21, 90nl5, 119,213,222,
220,227,228,250,270,278n29, 223,256,276nl6
279n34 Selection 158, 159, 208n22
Subject 17, 18,27-34,62,75 Semitic 96, 141, 142
V 16,19-27,68, 71,85,90n8,95n61 Sequence of tense 192, 193, 194
Reciprocal 81 , 82 Simple
Recursion 77, 85n7 future 149
Redundancy past 143, 149
rule 76, 86n7 present 149, 179
view 85n2 tense 142, 149, 174
Redundantism 76, 77 Simultaneity 144, 146, 147, 153, 154,
Reference 175, 183, 184, 190
time 144, 174 Siouan 280n42
Referentia140, 41, 105, 123, 182, 193 Situated tense 141
agreement 341, 364n30 Situation 142, 143
chain, R-chain 21, 37, 44, 54, 55, 75, aspect 143, 184
93n37 determiner 286n30
dependence 93n37 Slavic 83, 205n7
expression, R-expression 10, 44, 53, Small
117 clause analysis 87, 88, 89nl, 93n30
index, R-index 193 SNO 213, 222, 224, 272
NP, R-NP 44, 53, 93n37 Sound plural 137n24, 259, 275n5
tense, T, R-tense 193 Spanish 251
Referentiality Specific 28, 40
strong 29, 114 Specification 16, 34, 37, 38, 42, 44, 75,
weak 29, 30, 9ln22 86n8, 96, 109, 112, 116-127, 141,
Reflexive 81, 90nll 144, 146, 161-162, 163, 169, 173,
Relational Grammar 94n49 174,181,182,190,229
Relative Specificity 29, 9ln22, 215, 269, 272, 273
312 ANALYTIC INDEX
Specified configuration 62
Mood 156 Structure Preserving
tense 156 Constraint 12, 19
Specified Subject Condition, SSC 246, Subjacency 64
254 Subject 19,27-34,213,214,216,220,
Specifier, Spec 5, 16, 17, 20, 28, 31, 32, 224,232,234,239,243,246,249,
34,36,44,62,63,89n6 250,253,254,255,272,277n22,
licensing 272 278n29
position 18 adjunction 70, 274
Spec-Head case 45-53, 243
agreement 47, 58,223,227,229,231, Genitive 274, 276n16
257,270 nominal214, 238, 269
relation 19, 32, 38, 39, 40, 41, 54, 62, raising 16, 18, 27-32, 62, 75
220 role 226, 275n9
sharing 43, 44, 91n23 Subjunctive 84, 151, 153, 164, 169, 172,
Speech time, S 144, 147 173,207n20
Spell out 79, 82 Substitution 12, 14
Spelling rule 79, 81, 86n9 Support 116, 121, 164
Split svo
head 156 language 16, 112, 133, 134, 135
INFL 142, 160-161 order xi, xiii, 16, 18, 19, 27-32,
S-structure 4, 16, 17, 19, 20, 21, 23, 24, 45-47, 133, 140n35, 213, 274
26,27,31,33,40,49,62,63,68, Synthetic
193,213,214,216,220,224,227, genitive 213, 214, 246, 255, 256, 257,
228,229,242,252, 269,277n22, 270,272,274,276nl6
278n29 strategy 246, 274
Stage 143 system 213, 246
Standard Arabic, SA xi, xii, 109 Syrian Arabic 92n28
State 142, 143, 176, 178, 180
State-of-affairs 142, 155, 180 Tamakkun 275n4
Static 276 T -Case see temporal
Stative 144,152-156,177,179, 180, Telic 185
181, 185, 186 Temporal
Absolute 176, 183 adverb 97, 144, 145, 146, 182, 183,
Pure 175, 178, 179, 181, 182,184 205n7
Stati vity 179, 180 Case, TCase 163, 164, 169, 172,
Stati vization 179, 181 207nl9
Stem xii, 13, 75, 77, 78, 79, 83, 84, Case assignment 163, 166, 170, 172
86n8&10 Case checking 164, 166
Strategy 246, 255, 257, 272, 274 dependence 147
Strong discharge 163
Crossover 90-91 n 15 event 148
INFL 48, 50, 51, 53, 75, 77, 85n6 extension 181
1exicalism 77, 85n6 interval 149
referentiality 29 location 205n7
Strongly referential41 Tense, T 141, 142-146, 147-150,
Structural 152-156, 158, 159, 162, 174-175,
case 8, 45 181-184
ANALYTIC INDEX 313
Managing Editors
Joan Maling, Brandeis University
James McCloskey, University of California, Santa Cruz
Ian Roberts, University of Wales, Bangor
Publications
1. L. Burzio: Italian Syntax. A Government-binding Approach. 1986.
ISBN Hb 90-277-2014-2; Pb 90-277-2015-0
2. W.D. Davies: Choctaw Verb Agreement and Universal Grammar. 1986.
ISBN Hb 90-277-2065-7; Pb 90-277-2142-4
3. K. E. Kiss: Configurationality in Hungarian. 1987.
ISBN Hb 90-277-1907-1; Pb 90-277-2456-3
4. D. Pulleyblank: Tone in Lexical Phonology. 1986.
ISBN Hb 90-277-2123-8; Pb 90-277-2124-6
5. L. Hellan and K. K. Christensen: Topics in Scandinavian Syntax. 1986.
ISBN Hb 90-277-2166-1; Pb 90-277-2167-X
6. K. P. Mohanan: The Theory of Lexical Phonology. 1986.
ISBN Hb 90-277-2226-9; Pb 90-277-2227-7
7. J. L. Aissen: Tzotzil Clause Structure. 1987.
ISBN Hb 90-277-2365-6; Pb 90-277-2441-5
8. T. Gunji: Japanese Phrase Structure Grammar. A Unification-based Ap-
proach. 1987. ISBN 1-55608-020-4
9. W. U. Wurzel: Inflectional Morphology and Naturalness. 1989
ISBN Hb 1-55608-025-5; Pb 1-55608-026-3
10. C. Neidle: The Role of Case in Russian Syntax. 1988 ISBN 1-55608-042-5
11. C. Lefebvre and P. Muysken: Mixed Categories. Nominalizations in Quechua.
1988. ISBN Hb 1-55608-050-6; Pb 1-55608-051-4
12. K. Michelson: A Comparative Study of Lake-Iroquoian Accent. 1988
ISBN 1-55608-054-9
13. K. Zagona: Verb Phrase Syntax. A Parametric Study of English and Spanish.
1988 ISBN Hb 1-55608-064-6; Pb 1-55608-065-4
14. R. Hendrick: Anaphora in Celtic and Universal Grammar. 1988
ISBN 1-55608-066-2
15. 0. Jaeggli and K.J. Safir (eds.): The Null Subject Parameter. 1989
ISBN Hb 1-55608-086-7; Pb 1-55608-087-5
16. H. Lasnik: Essays on Anaphora. 1989
ISBN Hb 1-55608-090-5; Pb 1-55608-091-3
17. S. Steele: Agreement and Anti-Agreement. A Syntax of Luisefio. 1990
ISBN 0-7923-0260-5
Studies in Natural Language and Linguistic Theory