Sloppy Identity and VP-Ellipsis in English and Arabic

advertisement
Sloppy Identity and VP-Ellipsis in English and Arabic
Maha Abdul Salam Kolko
Abstract: The purpose of this paper is to examine VP-ellipsis in
Modern Standard Arabic and Syrian Arabic (abbreviated
throughout as MSA and SA, respectively) within the dependency
theory which is a framework developed by Fiengo and May (1994).
Although the data in Arabic does show a similar correspondence to
English as far as the ambiguity problem is concerned, it differs in
how the eliminative puzzles are accounted for, particularly in the
cases of the Many-clauses and the Dahl puzzles.
1. Introduction
The purpose of this paper is to examine the nature of VP-ellipsis in Modern
Standard Arabic and Syrian Arabic (abbreviated throughout as MSA and SA,
respectively).1 The study of VP-ellipsis in English has had a long tradition, especially
in the realm of syntactic theory; however, the case is not similar as far as MSA and
SA are concerned. This study has twofold goal: First, I attempt to explore the nature
of VP-ellipsis in Arabic. Second, I examine this phenomenon within the dependency
analysis which is a framework developed by Fiengo and May (1994). I will show that
the ambiguity problem aroused in the different constructions of Arabic cannot be
analyzed on a par with VP-ellipsis in English.
The paper is organized as follows. In the first section, I identify VP-ellipsis in
Modern Standard Arabic (MSA) and Syrian Arabic (SA). In the second section, I
examine this phenomenon within Dependency theory or the theory of indexical types
which is a framework proposed by Fiengo and May (1994). In the third section, I will
focus on a group of central empirical issues that arise in ellipsis with respect to what
is known as the ambiguity problem; strict and sloppy identity and the eliminative
puzzles and show how Fiengo and May’s framework is not comprehensive enough to
capture the multiple sloppy readings found in many Arabic elliptical sentences. In the
fourth section, I will attempt to propose an Alternative Solution which can account for
the many readings or interpretations in the Arabic data.
1
I would like to thank Joseph Aoun, Alan S. Kaye & Jamal Ouhalla for their helpful comments in
writing this paper.
2. VP-Ellipsis in Arabic
2. 1. The Basic Idea
It is relatively well-known since the work of Sag (1976) and Williams (1977)
that VP-ellipsis in English is only permitted when an overt auxiliary precedes the
elided material. Huang (2000: 131) has defined VP-ellipsis as the phenomenon
whereby the VP of the second and subsequent clauses of a structurally parallel
construction is lexically suppressed. An elliptical construction usually involves two
clauses which are structurally parallel. Where as the first clause (we refer to it as the
source) is a complete one, the second (target) is missing the anaphoric VP which can
be recovered from the source. For example:
(1) John wrote a novel, and Bill did [ ] too.
The stranded auxiliary in the second clause of example (1) indicates the elision
of a verb phrase. A representation for the elided VP must be recovered from the
representation of another clause, in this case, the first clause. This is similar to some
cases of VP-ellipsis in MSA:
(2) (a) ʔstācmlt-u
al-Hasūwb
wa kadhaalika
use.past.1fs the-computer and
so did
baqitū T-Tullaba
(MSA)
the other students
I used the computer and so did the other students
(b) ʔkālat
ate.past.fs
al-fātatū
the-girl-nom
tūffah-n
wa kadhaalika ʔaxā-ha
apples
and
The girl ate apples and so did her brother.
so did
brother-her
(MSA)
(3) ahmed
qaraʔa
l-majallata
Ahmed-nom read.past-3ms (def) magazine-acc
wa
and
ayđā-n
kareem
Kareem-nom
(MSA)
too
Ahmed read the magazine and Kareem too
The previous two sentences exemplify the two types of VP-ellipsis in MSA where the
elliptical VP can either be introduced by the equivalent of so does in MSA which is
kadhaalika in (2) or it can follow this structure ‘wa…..[NP] ayđā-n’ (and…..[NP] too)
which is an empty VP preceded by the conjunction and in (3).
As far as kadhaalika is concerned, it is more appropriately construed as a
coordinating conjunction rather than an auxiliary verb, given that it displays no verbal
morphology. In this case, ellipsis in Arabic looks very much like ellipsis in Hebrew
(and many other languages, other than English and, apparently, Norwegian, which
allow auxiliary stranding). It involves bare NP (particularly subject NP) arguments
rather than auxiliaries that introduce an empty VP.
This is further justified by looking into the nature of Arabic Coordinating
Clauses where in determining parallelism between clauses for the purpose of
interpreting VP-ellipsis, verbs are usually not mentioned as parallel elements, since
typically, in a clause where VP is missing, the verb is missing as well. But in
languages where verbs raise out of the VP overtly, it is to be expected that VP-ellipsis
would “strand” the verb. Languages differ according to the overt position of the verb,
whether within the VP or adjoined to the functional category T which is external to
VP.2 Some languages allow main verbs to raise to T overtly, whereas other languages
(e.g. English) allow this only for auxiliary verbs. Recent research has shown that overt
raising of V to T is attested in many languages: Semitic, Celtic, Romance, the Balkan
languages, Chinese, Japanese, Korean, and others.
This can be explained, as Koopman and Sportiche suggest, by arguing that in
the VS order, the verb, having moved to a higher functional category, just assigns
2
See Chomsky (1995)
case under government and has default number agreement. Koopman and Sportiche
use I(nflection) as the name of the functional head, rather than T and AGRs
(Koopman and Sportiche 1991; Mohammad 1989). Hence, the difference then
between Arabic-type languages and English-type languages is that the latter have
obligatory movement of the subject to IP. When partial agreement occurs, the verb
moves to I in Across-the-Board (ATB) fashion leaving the subject in the Spec of VP.
2. 2. VP-Ellipsis in Syrian Arabic
Two types of VP-elliptical sentences are found in SA which is similar to the
ones found in MSA. For instance,
(4) (a) ʔstācmlt
use.past.1fs
l-kәmputār
w
the-computer and
kāman
bāіt
so did
the other students
e- Tullab
(SA)
I used the computer and so did the other students
(b) ʔaklat
ate.past.fs
el-bint
the-girl-nom
e-tәffeeħ
w
kāman
apples
and
so did
ʔaxw-ā
(SA)
brother-her
The girl ate apples and so did her brother.
In sentence (4a, b), kāman has the same function as kadhaalika in MSA where
it can be a coordinator with a meaning similar to English 'so'. However, an additional
type of VP-ellipsis is also possible in SA exemplified in sentence (5b) which we can’t
have in MSA
(5) (a) omar
ʔakl
Omar-nom ate.past.ms
e-tәffeeħ
w
kareem
kāman
apples
and
Kareem
did so
Omar ate apples and Kareem did too
(SA)
(b) omar
ʔakl
e-tәffeeħ
w
Omar-nom ate.past.ms apples and
hetta kareem kāman
even
(SA)
Kareem did so
Omar ate apples and Kareem did too
Thus in Syrian Arabic, ellipsis does not require subject initial order in the second
clause which is contrary to other varieties of Arabic dialects, e.g. Moroccan Arabic. In
other words, in Moroccan Arabic (MA), you have to have the subject BEFORE the
equivalent of kadhaalika and PRON, not after as in SA, e.g.
(6) (a) Omar
kla
Omar-nom ate.past.ms
e-tәffeeħ
w
Kareem
apples and
Kareem
hetta huwwa.
did
(MA)
too
Omar ate apples and Kareem did too.
*(b) Omar
kla
Omar-nom ate.past.ms
e-tәffeeħ
w
apples
and
hetta huwwa Kareem.
did too
(MA)
Kareem-nom
Omar ate apples and Kareem did too.
3. Sloppy/Strict Identity
3. 1. Background
The terms strict and sloppy identity are more normally confined to anaphors.
The prototypical elliptical interpretation phenomenon is the strict/sloppy alternation,
which stems from an anaphoric dependency between a pronoun in the source clause
and its source clause antecedent, where the latter is a parallel element. As is wellknown, VP-ellipsis allows sloppy identity of pronouns. Fiengo and May argue that
sloppy identity is possible because Dependency theory contains a notion of identity-icopy-which allows β-occurrences of indices to be non-distinct if they manifest
identical patterns of indexical dependence. This has been discussed in relation to two
central empirical puzzles, which we call the many-pronouns and many-clauses
puzzles. For example:
(7) John loves his wife and so does Bill.
In (7) there is an ambiguity between whether Bill loves John’s wife (strict
identity) or whether Bill loves his own wife (sloppy identity). The difference between
strict and sloppy substitutions on nonparallel NPs, as outlined above, accounts for this
kind of anaphoric ambiguity. To take just one case indicative of the role of structure
in the distribution of strict and sloppy identity, observe that (8) does not allow of a
sloppy reading, as pointed out in Indices and Identity, (p.104)
(8) Max’s mother saw him, and Oscar said Mary did, too
The reason, according to Fiengo and May, is that there is a lack of syntactic
parallelism; more precisely, that the indexical dependency associated with the βoccurrence borne by the elided pronoun will not be an i-copy of that associated with
the β-occurrence borne by the unelided pronoun.
In Indices and Identity, Fiengo and May assume that a sentence such as Max
saw his mother is structurally ambiguous, the ambiguity being attributed to a
distinction in the representation of anaphoric pronouns that indicates whether the
pronoun is formally dependent on its antecedent or not. In Arabic, VP-ellipsis does
engender a sloppy reading, e.g.
(9) raʔā
saw.past.ms
al-walad-u
the-boy
ʔmah-u
wa
mother-his
and
kadhaalika Maher
so did
maher-nom
The boy saw his mother and so did maher
This sentence can have two readings; strict and sloppy exemplified as (a) and (b)
respectively.
ʔal-walad
(a) raʔā
1
saw.past.ms
ʔma ʔal- walad
the-boy
α
ʔmah-u
1
mother-his
wa
and
maher
raʔā
saw.past.ms
2
Maher-nom
α
1
mother-of –the boy
The boy saw his mother and Maher saw the mother of the boy
ʔal-walad
(b) raʔā
1
saw.past.ms the-boy
β
ʔmah-u
1
mother-his
wa
raʔā
and saw.past.ms
maher
2
Maher-nom
β
ʔmah-u
2
mother-of –Maher
The boy saw his mother and Maher saw his mother
The ambiguity in (9) is because of the pronoun hu (his) which is understood in the
elided material may be construed as being co-indexed with the subject of the first
conjunct ʔal-walad (the boy) or with the subject of the second conjunct (Maher).
Although the data in Arabic does show a similar correspondence to English as far as
the ambiguity problem in example (9), it differs in how the eliminative puzzles are
accounted for, particularly in the cases of the many-clauses puzzle and the Dahl
puzzle.
3. 2. The Many-Pronouns Puzzle
I will begin with examining the first of the eliminative puzzles which is called
the many-pronouns puzzle. In this kind of puzzle, the number of pronouns is
increased beyond the one found in the standard examples used to illustrate strict and
sloppy identity. Here the elliptic VP does not share the same range of possible strict
and sloppy readings as its antecedents. For example
(10) Qala
said.past.3ms
maher
innah-u
yu-hibb-u
mouclmah-u
Maher-Nom
that-he
like.pres.3ms
his-teacher
wa
kadhaalika qala
and
did
omar.
said.past.3ms
Omar-Nom
Maher said he likes his teacher, and Omar said he does, too
This can have the following three well-formed representations:
(i) Qala
maher1
said.past.3ms
wa
and
Maher-Nom
qala
innah-u1α
that-he
like.pres.3ms
his-teacher
innah -u1α yu-hibb-u
omar2
said.past.3ms
mouclmah-u1α
yu-hibb-u
Omar-Nom
that-he
mouclmah-u1α
like.pres.3ms
his-teacher
Maher said that Maher likes his teacher and Omar said that Maher likes his teacher
(ii) Qala
said.past.3ms
wa
and
maher1
innah -u1β
Maher-Nom
that-he
qala
said.past.3ms
like.pres.3ms
innah -u2β
omar2
Omar-Nom
mouclmah-u1β
yu-hibb-u
his-teacher
mouclmah-u2β
yu-hibb-u
that-he
like.pres.3ms
his-teacher
Maher said that Maher likes his teacher and Omar said that Omar likes his teacher
(iii) Qala
maher1
said.past.3ms
wa
and
Maher-Nom
qala
innah -u1β
that-he
said.past.3ms
like.pres.3ms
innah -u2 β
omar2
Omar-Nom
that-he
mouclmah-u1α
yu-hibb-u
yu-hibb-u
like.pres.3ms
his-teacher
mouclmah-u1α
his-teacher
Maher said that Maher likes his teacher and Omar said that he likes Maher’s teacher
*(iv) Qala
maher1
innah -u1α
yu-hibb-u
mouclmah-u1β
said.past.3ms
wa
Maher-Nom
qala
and
that-he
innah -u1α
omar2
said.past.3ms
like.pres.3ms
Omar-Nom
that-he
his-teacher
mouclmah-u2β
yu-hibb-u
like.pres.3ms
his-teacher
For this case, there are four possible combinations of indices for the
antecedent; of these, only three give rise to well-formed elliptical structures. Given
the correspondence of variables with the sloppy reading, and constants with the strict
reading, (i) and (ii) will result in ‘across-the-board’ strict and sloppy readings,
respectively, while (iii) and (iv) will give readings mixed between sloppy and strict.
However, a fifth interpretation (the availability of a fifth reading) is also possible
which is:
(v) Qala
said.past.3ms
wa
Maher-nom
qala
and
innah -u3β
maher1
said.past.3ms
that-he
omar2
Omar-nom
yu-hibb-u
like.pres.3ms
innah -u3 β
yu-hibb-u
that-he
like.pres.3ms
mouclmah-u3 β
his-teacher
mouclmah-u3 β
his-teacher
Maher said that Kareem likes his teacher and Omar said that Kareem likes his
teacher
In the previous sentence, hu can refer to a third person/ referent ‘K’ (refers to
someone else than Omar) whose name is not mentioned in the sentence. If it is
replaced by a name then the relevant intuition or interpretation vanishes.
3. 3. The Many- Clauses Puzzle
It is important to note that these puzzles arise when we increase the
complexity of the data, either by increasing the number of pronouns contained in the
ellipsis or by increasing the number of clauses that contain ellipses. Fiengo and May
assume that the solution to the many-clauses puzzle for a sentence such as:
(11) maher
mouclmah-u wa omar ayđā-n wa lysa zayd
yaħtarm
Maher-nom respect.pres.3ms his-teacher
and Omar
too
but not Zayd
Maher respects his teacher and Omar too but not Zayd
Sentence (11) can have the following readings: One might expect there to be two
readings or two well-formed representations (a) and (b) are derived, that is, either all
the three pronouns are α-occurrences or all of them are β-occurrences (according to
Fiengo and May). However, in Arabic, five readings are derived.
(a) maher1
mouclmah-u1α
yaħtarm
Maher-Nom respect.pres.3ms
mouclm maher1α
wa
teacher-Maher
his-teacher
lakna zayd3
but Zayd-Nom
la
wa
omar2
yaħtarm
and Omar-Nom respect.pres.3ms
yaħtarm
not respect.pres.3ms
mouclm maher1α
teacher-Maher
Maher respects his teacher and Omar respects Maher’s teacher but Zayd doesn’t
respect Maher’s teacher
(b) maher1
Maher
yaħtarm
respect.pres.3ms
mouclmah-u2β
wa
his-teacher
mouclmah-u1β
wa
omar2
his-teacher
and
Omar
lakna zayd3
la
but
not
Zayd
yaħtarm
respect.pres.3ms
yaħtarm
respect.pres.3ms
mouclmah-u3β
his-teacher
Maher respects Maher’s teacher and Omar respects Omar’s teacher but Zayd doesn’t
respect Zayd’s teacher
The other three extra readings are: (c) another sloppy reading and (d) a fourth
reading which is derived by the use of the dual pronoun huma and (e) where the
pronoun hu (his) which might not be co-indexed with the NPs in the sentence. It can
refer to another person whose name is not being mentioned in the sentence.
(c) maher
Maher
mouclmah-u β wa
yaħtarm
1
respect.pres.3ms
mouclmah-u2β
wa
his-teacher
omar
1
his-teacher
lakna zayd3
but
Zayd
and
la
yaħtarm
2
Omar
respect.pres.3ms
mouclm Omar
yaħtarm
respect.pres.3ms teacher-omar2β
not
Maher respects Maher’s teacher and Omar respects Omar’s teacher but Zayd
doesn’t respect Omar’s teacher.
(d) maher1
mouclmah-u1α
yaħtarm
Maher
respect.pres.3ms
his-teacher
mouclmah-u2β wa lakna zayd3 la
his-teacher
but
Zayd
wa omar2
and
Omar
yaħtarm
respect.pres.3ms
mouclimy-hma3α β
yaħtarm
not respect.pres.3ms teachers-maher and omar
Maher respects Maher’s teacher and Omar respects Omar’s teacher but Zayd
doesn’t respect both (Maher and Omar’s) teachers.
(e) maher1
Maher
mouclmah-u4 β wa
yaħtarm
respect.pres.3ms
mouclmah-u4β
his-teacher
wa
but
his-teacher
lakna zayd3
Zayd
and
omar2
Omar
yaħtarm
respect.pres.3ms
mouclmah-u4β
la
yaħtarm
not
respect.pres.3ms teacher-his
Maher respects his teacher (k) and Omar respects his teacher (k) but Zayd doesn’t
respect his teacher (k) (the missing VP is resolved outside the sentence).
As shown by Fiengo and May, the Parallel Dependency approach would only permit
the first two readings but not the other three which are ruled out. However, in Arabic,
all the previous five representations are allowed. This is also similar to what is known
as the Dahl puzzle.
3. 4. The Dahl Puzzle
This type of puzzle is exemplified in (12)
Yactaqid
(12)
innah-u
maher
think.pres.3ms Maher-nom that-he
ʔma ʔbuh-u
but
qāwy
wa
strong
and
kadhaalika
so does
omar
Omar-nom
fāla
father-his not
Maher thinks that he is strong and so does Omar, but his father doesn’t
Sentence (12) can have the typical strict and sloppy readings of simple many-clauses
sentences. In addition, we get the following extra readings:
(c) Yactaqid
maher
α
innah-u
1
1
think.pres.3ms Maher-nom that-he
qāwy
ʔma
strong
but
ʔbuh-u
2
2
2
strong and think.pres.3ms Omar that-he
β
innah-u
Yactaqid
fāla
qāwy
2
father-his does-not think.pres.3ms that- he
strong
Maher thinks that Maher is strong and Omar thinks that Omar is strong, but his
father does not think Omar is strong
(d) Yactaqid
maher
4
think.pres.3ms Maher-nom
β
innah-u
4
qāwy
that-he strong
qāwy
strong
ʔma
but
α
innah-u
1
β
omar innah-u
qāwy wa Yactaqid
qāwy
wa
Yactaqid
omar
2
that-he strong and think.pres.3ms Omar
ʔbuh-u
3
father-he
fāla
does-not
Yactaqid
think.pres.3ms
β
innah-u
4
that-he
Maher thinks that he (k) is strong and Omar thinks that he (k) is strong, but his father
(k) does not think that he is strong.
In (d), ‘he’ (k) is a reference to another person who is neither ‘Maher/ Omar’ or even
the ‘Father’.
(e) Yactaqid
maher
α
innah-u
1
qāwy
1
think.pres.3ms Maher-nom that-he
qāwy
ʔma
strong
but
β
ʔbuh-u
2
2
strong and think.pres.3ms Omar that-he
α
innah-u
Yactaqid
fāla
4
α
omar innah-u
Yactaqid
wa
1
father-his does-not think.pres.3ms
that- he
qāwy
strong
Maher thinks that he is strong, Omar thinks that Maher is strong, but his father (k)
does not think Maher is strong
(f) Yactaqid
maher
β
1
innah-u
qāwy
1
β
omar innah-u
wa Yactaqid
2
2
think.pres.3ms Maher-nom that-he strong and think.pres.3ms Omar that-he
qāwy
ʔma
strong
but
ʔbuh-u
4
β
innah-u
Yactaqid
fāla
2
father-his does-not think.pres.3ms
that- he
qāwy
strong
Maher thinks that Maher is strong and Omar thinks that Omar is strong, but his
father (k) does not think Omar is strong
(g) Yactaqid
maher
1
1
think.pres.3ms Maher-nom that-he
qāwy
ʔma
strong
but
β
innah-u
ʔbuh-u
4
fāla
strong
Yactaqid
β
omar innah-u
qāwy wa Yactaqid
2
2
and think.pres.3ms Omar that-he
β
innah-u
4
father-his does-not think.pres.3ms that- he
qāwy
strong
Maher thinks that Maher is strong and Omar thinks that Omar is strong, but his
father (k) does not think he-father is strong
To sum up, it appears that the complex Parallel Dependencies approach is
necessary to account for certain complex ellipsis examples, involving multiple
occurrences of pronouns and clauses in English as has been argued by Fiengo and
May(1994). However, this is not the case in Arabic. I have exemplified that the
interpretations of Arabic data is similar to the English ones as far as the ambiguity of
strict and sloppy readings and the Many- Pronoun constructions are concerned,
however, it differs in the cases of the Many-Clauses and the Dahl Puzzles where we
can have at least four interpretations for each case.
4. Proposal
A question arises as to how strict and sloppy identity may be described in the
system outlined here. In fact, a theory of sloppiness should be general enough to
encompass the various configurations in which sloppy/ strict ambiguity arise while
correctly accounting for the parallel constraint. In order to resolve VP-ellipsis in
Arabic, it is important to find a theory which will accounting for the many readings in
the cases of Eliminative Puzzles. Where as the Dependency theory is not capable of
capturing all of the previous readings in the previous examples, DSP has been found
to accommodate all of them except the one with the dual pronoun.
4. 1.The Equational Method ( DSP): An Alternative Solution
Dalrymple et al. (1991) resolve ellipsis through the use of unification to
resolve equations relating the meaning of the ellipsis antecedent to that of the ellipsis.
For example
(13) Dan likes golf. So does George
We can represent the meaning of the antecedent sentence as
(14) likes (dan, golf)
The meaning of the ellipsis is
(15) P (george)
Where P is a variable over some property, which when applied to George gives the
resolved meaning of the ellipsis.
Noting that dan and george are parallel expressions in antecedent and ellipsis, the
property P applied to dan should give the meaning of the antecedent. That is
(16) P (dan) = likes (dan, golf)
Using high-order unification to resolve this equation, we derive
(17) P = λх. likes (х, golf)
Where typed λ-terms represent phrase meanings. Applying this solution for the value
of P in the ellipsis, we obtain as the resolved meaning
(18) likes (george, golf)
Applying high-order unification to example (19) previously (11) to resolve VPellipsis in Arabic, we derive the following:
(19) maher
yaHtarm
mouclmah-u wa omar ayđā-n wa lysa zayd
Maher-Nom respect.pres.3ms his-teacher
and Omar
too
but not Zayd
Maher respects his teacher and Omar too but not Zayd
(20) Source = respect (Maher, teacher (Maher))
Target = P (Omar)
Where P (Maher) = (Source meaning)
(21) Source P (Omar) = respect (Omar, teacher (Maher))
P (Omar) = respect (Omar, teacher (Omar))
Target = Q (Zayd)
Where Q (Omar) = respect (Omar, teacher (Maher))
Q (Zayd) = respect (Zayd, teacher (Maher))
From the strict reading of clause (2)
Where Q (Omar) = respect (Omar, teacher (Omar))
Q (Zayd) = respect (Zayd, teacher (Omar))
= respect (Zayd, teacher (Zayd))
from the sloppy reading of clause (2).
(22) Source = respect (Maher, teacher (Kareem))
Target = P (Omar)
Where P (Maher) = respect (Maher, teacher (Kareem))
P (Omar) = respect (Omar, teacher (Kareem))
(Strict reading)
(Sloppy reading)
Here the pronoun is resolved outside the clause. This approach to resolve VP-ellipsis
seems to be more adequate in accounting for the readings we get in Arabic which
have been discussed in relation to VP-ellipsis in English in Kehler (1997)
5. Conclusion
To conclude, in this paper, I have identified VP-ellipsis in MSA and SA. I
have shown how Arabic VP-elliptical sentences exhibit the strict/ sloppy identity and
how different they are from the English data. Also, I have shown how Fiengo and
May’s theory is not adequate enough to account for the many readings which we
derive in Arabic and have proposed Dalrymple et al. (1991) the equational method as
an alternative frame work which can account for the sloppy/strict readings of VPelliptic constructions in Arabic.
References
Chomsky, N. (1995), The Minimalist Program. Cambridge, MA: MIT Press.
Dalrymple, M., S. Shieber, & F. Pereira (1991), ‘Ellipsis and higher-order
unification’, Linguistics and Philosophy14: 399-452.
Fiengo, R. & R. May (1994) Indices and Identity. Cambridge: The MIT Press.
Huang, Y. (2000). Anaphora. Oxford: Oxford University Press.
Koopman, H. & D. Sportiche (1991) ‘The Position of Subjects’, Lingua 85: 211-258.
Mohammad, M. A. (1989). The Sentential Structure of Arabic. Ph.D.thesis. University
of Florida.
Sag, I. (1976). Deletion and Logical Form. Cambridge, MA: MIT Press.
Williams, E. (1977). ‘Discourse and Logical Form’, Linguistic Inquiry 8:101-139.
Download