Report
Reference
The hierarchical representation of subject verb agreement
SHLONSKY, Ur
SHLONSKY, Ur. The hierarchical representation of subject verb agreement . Haifa University : 1989
Available at:
http://archive-ouverte.unige.ch/unige:83251
Disclaimer: layout of this document may differ from the published version.
1 / 1
The Hierarchical Representation of Subject Verb Agreement*
Ur Shlonst.y
Univ~rsity of Haïfa 19ô9
This paper advances a model for representing agreement features
insyntax.
Iargue tnat each feature is a zero-levet category. Wllich.
in kwping with the X-bar convention, projects a phrasa! category. Verbsubject agreement is brought about
byconsecutive applications of move a, as argued by Pollock (
19ôô) andChomsky (
19ô9).The model accounts
for the patterns of agrooment in Hebrew and Arabie but I oolieve itis not
andcannot
bespecifie
to theselanguages.
The tlleory
developoo is tllen made
toaccount
ina straightiorward manner for a numoor of agreement-related phenomena such as null
subjectsand
word order variation.*
Re5earch for thi~ paper va~ carried out vhile the author vas supported by an Alon fellovship. Mditional support vas proVided by an Israeli Hinistry of .Absorption Research Develop»ent Grant (12134) and a University of Haifa faculty of Hu»anities Grant. I aa grateful to R.Tabaon for help vith the Standard Arabie data and for discussions of Arabie grauar. The Palestinian Arabie data, Wlless othervise noted, 1s solicited froa speakers froa Palestine's Galillee region. I aa
especially grateful to O. Avad and S. Hasan for assistance vith this data. Finally, I vish to thank ztudentz and colleaguez at Tel Aviv and
Haita universities. .All errors and oAissions are •Y ovn.
·Addenda 1993
This is the original manuscript distributed in 1989. It has not been revised.
Author's current address
Département de linguistique générale Faculté des lettres,
Université de Genève CH-1211 GENEVE 4 SWITZERLAND
~~l@uat~th
1
Tbe Problem ... 3
1.1 Agreement Patterns in Standard Arabie ... 3
i .2 Allreement :Patu;;,rns in Hebrew ... 4
1.3 E~dence For a Hterarcny or
4'
Features ... ) 1.4 Agrooment and its Syntactic Correlates ... 71.5 Issues to
oo
Considered and Organization of Paper.. ... 8i: ~;~~~~:r~~p~~;~-~-·.::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::·ï~
4 Agreement and T(mse in U1e Hebrew Verbal Paradigm ... 144.1 Overview of the Hebrew Verbal Paradigm ... 14
4.2 The Be>noni ... :···--·--·----15
4.2.1 'The Benoni as a TEmseless Form ... !) 4.2.2 Analysis of the Benoni and the Position of Tense ... 18
4.2
.3
The Benoni V-Raising or Affix-Lowering? ... 204.2.3.1 The Position of VP-Adverbs in Hebrew ... 20
4.2.4 The Placement of Negation in Palestinian Arabie ... 23
4.3 The Tensoo Forms ... 26
4.4 Infinitives ... 27
4.5 The Node
f
and Ule Syntactic Aspect of Sëmitic Morphotogy ... 305 The
~0~~~~:-~~;~~~bi~~·-·::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::~;
5.2 The Hebrew Pro-Drop Paradigm ... 335.3 The 'Pro-Condition' ... Le-xical Government ... 34
).4 The Third Person Past and Future Forms ... 36
S.S
Weak Nodes and Null Nodes ... 40d V . ti . A b. 42 6 Agreement and Word Or er ana on m ra tc ... .. 6.1 VS- word Order and the Position of the Oausal Subject ... 42
6.1
svo
Word Or der in Arabie ... 4ô References ... 49i The Problem.
1.1 Agreement Patterns in Standard Arabie.
Consider the basic pattern of subject verb agreement in (Standard) Arabie in ( 1) and (2) ~low. Verbs agr"' With the-ir subjects in person, number and gender in the embedded sentences (lb). (2b). They agree with the-ir subjects only in gender in (la), (2a).1 As the-se e:xamples make obvious, tlle pattern of agreement
correlat-es With the order of constituents in the Arabie sentence. When the verb is clause initial. as in the (a) e:xamples, it agrees With the subject only in gender. Full agreement, in gender, num~r and person is manifested ont y When the subject is pre-verbal, as in (lb). (2b).
( 1) a. ?akal-a 1-?awtaad 1-taÇ'aam.
ete- M the- boy- M- Pl the-food The boys a~ tlle food.'
b. QuJt-u ?inna 1-?awlaad ?akal-u 1-taraam.
seid-1S thet the-boy-M-Pl ete-3-M-Pl the-food
'I said that the boys ate the food.'
In Arabie (as vell as Hebrew. See ahead.) the third person singular for• is aabiguous. It is used as a designation of third person and singular nuaber, but it also designates the absence of person and nuaber agree•ent. It is this latter, 'iBpersonal' for• which characterizes clause-initial verbs in Arabie.
(2) a. ?akal-at 1-banaat 1-ta\'aam.
ate-F the-girl-f-Pl the-food Th~ girls até th~ food'
b. Qult-u ?inna 1-banaat 7akal-na 1-ta\'aam.
~id-1S toot the-girl-F-Pl ate-3-F-Pl the-food
'1 said that the girls até th~ food.'
An adaquaté account of these Arabie facts must explain how the grammar isolaœs gender from the other components of Agr, ~rmitting th~ agreement- ooaring verb to access it directly to the exclusion of ~rson and numoor.
1. 2 Agrooment Patterns in Hebrev.
The same issue arises when one examines the contrast between the ténsed verbal forms of Hebrew and the participial, pr~ent œnse form, ~noni). In the future and past forms, (,3a,b). the verb is inflected for gender, number and ~rson,
which appear congealed into a prefix in (,3a) and into a suffix in (3b). In the
~noni. on the other hand, the verb agrees with its subjoct only in gender and numoor, (3d.
(3) a. Ata ti-smor ?al ha-xacilim.
you 2- M-S guard on the-ewplanb 'Y ou will guard the eggplants.'
b. Ata Sa.mr-ta ?al ha-xacilim.
you guard- 2- M-S oo the-ewplaot~
'Y ou guarded th~ ~plants.'
c. Ata romer ?al ha-xacilim.
you goord- M-S on the-eggph~nt$
'Y ou guard/are guarding the eggplants.'
~u~ th9 B9noni r9pr~nts only two f9atur~, num~r and g~nd~r, it has onty four forms, as in (4).
(4) masculine feminine
singular oomer-0 oomer~t
plural Somer-im /iomrim/ Som er -ot /iomrot/
As in the Arabie examples in ( 1-2), these data from Hebrew demonstraœ that the grammar can access num~r and gender featur~ to the exclusion of person features, which are otherwise available in the language since they are overtly represente<! in the tensed verbal forms.
1.3 Evidence For a Hierarchy of t Features.
Mor~ov~r, f~atures of agreement are depend~nt one upon the other: G~nder can ~ represented without number and person, as in Arabie
v
-initial clauses, gender and number without person as in Hebrew &moni forms. But crucially,there is no verb wnich is markoo for numoor and not markoo for gender and no verb whkh is markoo for person but not marke<! for numoor. O~rvationally,
then, the implkatlona1 hierarchy in (5) seems to hold.
(5) Implicational Hierarchy of Agreement Features
-If a verb is infloctoo for numoor then it is also infloctoo for gender.
-If a Vérb is infledéd for person tben itis also infloctOO for number.
On fammar learnabiHty grounds, it is highly unHkely that (5) captures an accidentai state of aff airs. Moroover, {5) genera1izes to non-SG'mitic languages.
With certain accountable differences (such as the absence or non-distinctness of gender in the non-participial verb forms). (5) adaquately describes the patterns of agreement in Romance as weil. Consider, for example, French past partldple agreement. Uke the Hebrew Benoni it bas four forms, the combinations of numoor and gender. The past participle [repaintoo] in {6) agr~ with the wb- fronted. objoct in numoor and gender. Crucially, there is no form in Frencb w:bich manjfests JXffS<>n features but no numt>e-r distinctions and no form inflected for, say, person and gender, but not numoor.2
(6)
2
Je sais combien de tables ils ont repeintes.
1 kOO'vt ho'R' mani) of tcbl~-f-Pl they hove re pei nted-f-Pl
'I know how many tables they repainted.' (Kayne (1987)) {5} is not as obvious in French since the non-participial. tensed foras are inflected for nuœber and persan but not for gender. This is due to the tact that gender is non-distinct in these foras. as in the Se•itic first person. (5) should be read as follovs: •If a form manifests nuaber then it aanifests gender only if gender is distinctive.-
S!m.Harly,
in RuSSian, verbs in the nonpast ~nses agr~ in ~rson and numoor with their subjoct; in the past tense, they agree in gender and numoor with their subject. HypothesiZing, then, let us take (5) to be descriptively valid of Universal Grammar.1.4 Agreement
andits
~tacticCorrelates.
The degree to which a verb is richty inflected bas indisputable syntactic correlations in many languages. Thus, in Arabie, tully inflected verbs cannot precede their subjocts, but must follow them whereas the semi-impoverished forms (manifesting only gender) appear dause-initially, giving rise to VSO word or der.
Word or der variation is dearly a syntactic phenomenon, driven, it is commonly assumed, by the exigencies of the case Module and constrainoo by the
Government and Bounding modules.3
In Hebrew, the degree of infloctional richness correlates with the possibility of (argument) nun subjects.4 Those are permitœd in the ~nsed forms but ruled out 3 See. e.g .. Koo~ (1961). Sproat {1965). Travis {1981) and .any others.
4 for discussion of non-arguacnt null subjects in Hebrev. vhich are licenced under different conditions. see Shlonsky (1969a).
in the &noni.5 Although not strictly relate<! to the question of how agreement is represente<!, a satisfactory account of agr~ment should ba expected to shed light on wh y null subje<:ts are ~rmitted with full agreement and how the degree of infloction affects word order.
1. 5 Issues to
beConsidere<!
andOrganization of Paper.
Consider, then, the following questions:
(7) I. By what grammatical process do verbs acquin~ agreement?
1 I. a. How is agreement information represented in the grammar so that verbs can have access to specifie features within it?
b. How is the implicational hierarchy of agreement represented?
II 1. What accounts for the correlation bet~n particular grammatical processes such as word order or ~-drop and the degr~ of inflectional richness?
As far as question I is concerne-ct, 1 adopt the analysis of inflectional morphology defended by Pollock { 19ôô) and Chomsky { 1989). to which I have HWe to add. Section 2 below sketches their theories. Sections
3
and 4 develop and defend an answer to question II. Finally, the interaction of inflectional richness and NQ.-drop is discussed in Section5
and word order variation in Arabie in Section 6.5 See ahead, 1.3, tor discussion ot the third person past and tuture forBs, vhich permit null subjects under aore restricted circu.stances.
2. Inflectional Morphology.
According to Pollock ( 19ôô). clauses have the following structure:
(8) /)~ns~ (=IP) NP ~nse·
Te~ "'-A~
/·',A~
Agr ~
f
a.Gender V
flNumber
1
PersonPollock argues that TEmse and Agr constitute distinct heads in the X-bar sense, (i.e., Xûs). Moreover, he takes the view that where Agris transparent (or strong), it attracts the verb to it. Move a applies and V raises to Agr and amalgama tes with it, forming an infle<:ted verb. This, he claims. is the situation in French finite clauses. In English, Agris opaque or weak and V cannot raise toit. Consequently, the agreement affixes move down to V. Th us, inflectional morphology in French- type languages takes the form of V-Raising (9a). and in English-type languages it takes the torm of Agr-lo~ring. (9b).
(9} a. French-typy languages: b. English-typy languages:
Both instances of Move a are transformations afte<;ting
xo
elements Which Chomsky ( 1986b, 1989) argues, involve adjunction to a category on the same levet, Le. anxo.
Pollock further sho'IIS that the same prinicples extend to Tense so that a strong (finite) Tense attracts movement wbile a weak (non-finite tense) moves down.
The Head Movement constraint. ( 10). blocks direct movement of V to T, so that the 3r·(l.•Y'IrNJ.1'(il(
options are diagrammed in ( 11).6
(1 0) Head Movement Constraint {Travis 0984)).
An
xo
may only move into ayo
wbich pro~rly-governs it.6 Baker (1988), Choasky (1989) arque that the HnC is fully reducible to the Empty Category Principle (ECP). For clarity of exposition, I continue to eaploy the descriptive ter• HUC.
10
However, ( 10) appears to rule out all cases of
xo
lowering, since the e.c.left by the lowered head is not c-commanded (and bence not governed) by its antecedent, the lowered head. Chomsky argues that in LF the verbal complex tormed by lowering the affixes onto it raises and covers up the e.c.'s formed at $-structure.At LF, French-ty~ and English-type languages are identical, with V and Agr adjoined to Tense. The HMC can be satistied in LF since the raised heads c- command their traces. Crucially, this relies on the grammar's ability to sus~nd application of the HMC until LF as well as on a numt>er of other assumptions discussed at length in Chomsky ( 1989).
Thus, in (11), V must move to Agr in order to raise to Tense. Tense, however, may lower diroctly to V. This is so precisely because the HMC can be suspended until LF.
(11)
For Chomsky, V -Raising is obligatory at S-structure whenever possible (i.e., whenever Agr and Tense are strong). Otherwise, Agr and/or Tense must Iower onto V. In LF. the strong/weak distinction no longer constrains V -raising which then must apply, even if it could not apply at $-structure.
3.
The Structure of Agr.
ln Section 1 above, we argue<! that the
t
features constituting Agr must be so represenW<i as tooo
accessible i_Ddividually and in a particular or der. As it stands, (ô) faits to articulate any internat structure for Agr, because it treats Agr as a . single node containing the features in a bundle.Suppose, therefore, that there is no Agr node perse but rather, that each feature of agreement (
t
feature) heads its own maximal projection. In other words, consider the hypothesis tllat Pollock's AgrP is merety an abbvreviation of the more articula te structure in (12 ).(12) Per~·
~erS{)n·
Per~n Numoor"
~
yumœr·
Number
~ender"
'""'
Gender' 1Ge<nder
Given ( 12) and the tlleory of inflectional morphology outlined above,
we
canans~r (7-II). Verbs have access to the contents of Agr because tlley associate witll its features one by one. subjoctto the conditions on head movement and the st.ronglweak distinctions in each and every head.
12
Moreover, the generalization in (5) is accounted for by the HMC.
v
can only adjoin to NumoorO if it previously adjoins to GenderO and a verb "W'iU manifest person agreement only if it manifests number and gender agreement. since movement to PersonO must first go tllrough NumoorO and GenderO. When V cannot adjoin to sorne xo because xo is weak and doos not licence V -Raising. thenv
can not raise to yo, where yOc-commandsxo .
In Arabie, the fact that the verb agrees with the subjoct ont y in gender in verb- initia1 constructions, implies that the verb bas adjoined to GenderO, but has failed to move up further. Conversety, in subjoct-initial clauses, where a verb manifests full agreement, V has raiood and adjoinoo to PersonO. In Hebrew, a verb in tbe
~noni form has only movoo up to NumoorO but not to PersonO.
The Phrase-marker in (12) thus provides a framework in which verbal agreement patterns can be perspicuously described. Given ( 12) we can state the following two questions:
( 13) I. What blocks V from moving further up to NumoorO and PersonO in Arabie?
II. What prevents a Benoni verb from moving to PersonO and acquiring
P~rson features?
In Section 4 oolow, I atœmpt to answer ( 13-II). Question ( 13-I) is taken up in Section 6.
4 Agreement and Tense in the Hebrew Verbal Paradigm.
4.1 Overviev of the Hebrev Verbal Paradigm.
Hebrew verbs manifest eitller no agreement witll tlleir subjects (infinitives), agreement in number and gender only (Benoni forms. See (4) above.) and full, i.e., gender, numoor and ~rson agreement in tlle past and future tenses, as shown ootow in (14)_7.8
? Since verbs in the future or past tense display all three agreeaent features. they should aanifest twelve distinct foras for each tense. In reality, only eight foras are attested in Uodern colloquial Hebrev, as the table in (14) shows.
The reasons for which the paradiga in (14) is defective are varied:
first, Hebrew (and Arabie) are non-distinct tor gender in the tirst person, (cf. 5.5). Second, the past and future second person plural feainine fora and the future third person plural feainine fora are considered archaic in colloquial Uodern Hebrev (but not in Arabie or in more toraal registers of nodern Hebrev). Tbird, the absence of a third person plural feainine fora in the past tense is probably an accidentai gap, since the equivalent fora exists in Arabie.
The overall aorphological opacity of the agreeaent affixes, i.e., the difficulty in breaking thea down into the coaponent 'aorpheaes' (gender, nuaber and person) should not deter us froa assigning thea discrete syntactic representations. The absence of isoaorphy between inflectional affixes ~ syntactic objects and ~ aorphological ones is quite
faailiar to students of aorphosyntax. A specifie argument to this effect
14
(14) PAST FUTURE
singular plural 2ingular plural
Samar-ti Samar-n-u ?e-Smor ni-Smor 2m Samar-ta Samar -te-rn ti-Smor ti-Smor-u
2f Saœar-t ti-Sœor-i
3m
Sa.mar-0 $amar-0-u yi-Smor yi-Smor-u3f
Samar-0-a ti-SœorThe Benoni is discussed in 4.2 oolow. The past and future forms in 4.3 and the infinitive form in 4.4.
4 . 2 The Be no ni.
4.2.1 The Benoni as a Tenseless fora.
Benoni forms in Hebrew differ from the future and past t.ense forms in two ways. First, as we have seen, tlley are inflected for gender and number only and is given in footnote 21. for aore discussion, see Anderson (1982), (1986), Broaberger and Halle (1989) aaong others.
8 The status of the third person past and future foras is not so clear.
On the one band, the third person tora lacks any overt aanitestation of person. On the other hand. it is discretely represented aaong the other persons in the sense that the absence of a person affix is itself distinctive and unaabiguo~ly identifies the fora as third person. These problematic tor•~ are di~c~~ed in 1.3.
not for ~rson. They differ from the other verb forms also in that they are not spedfied tor tense _
The Benoni. as &rman ( 197ô) sho-ws, ap~ars in a wider range of environments than, say, the Englisb present tEmse. In addition toits distribution as the present tense form, the Benoni can be a gerundive complement to a perception verb, in ( 15a) and a partidple, ( 15b). both of which are forms witbout
an
inde~ndenttEmse specification.
(15) a. Ra?i-ti ?etDani holex.
~"111'-M-S-1 ccc D6ni 'W'OU::-BEOONI-M-S 'l saw Dani walking·
b. Dani baya margiz ?et k. ulam k~-bu baya
~ni oo-PAST-M-S- 3 armoy-BEPVNI-M-S ace everyooo vhen-he be-PAST-M-S-3
ox:el xacilim.
est- BENON!-M-S eggplsnb
l>ani used to annoy e:veryone wben he used to eat eggplants'.
To account for the distribution of the Benoni as botha present tense torm and as a participle, Berman ( 197ô) and Doron ( 19ô3) propose tbat Hebrew t&nsed forms can be either past or future. The present tense, in their view, is formally tenseless; its employment as the present tense is so by default. 9
9 Standard Arabie and the modern colloquial dialects have a for• vhich is morphologically identical to the Hebrev Benoni. (called ?ism Faa.m1
'verbal noun' in traditional Arabie gra..ars.) Its syntaetie
16
distribution, hovever, is mueh more restrieted than that of the Benoni:
In general. it cannot be used as a present tense verb but only as a partieiple.
Hebrev differs from both Standard Arabie and the modern dialects in having a tense rather than an aspect system (which also charecterized Biblical Hebrev.) In Arabie, the present tense is expressed by the imperfect verbal for• am not by the ?ïsm. Feffiil. Contrast the Hebrev Benoni (i) vith its Palestinian equivalent, (ii), where the verb bears the i:llperfective prefix, lb-/. The ?ism faaÇU cannot be used here, as shown by the ungra.uaticality of (iii). It can only be used as a participle, as in (iv).
(i) Dan pote?ex ?et ha-delet.
(ii)
(Hi}
(iv)
Dan open- Bff«JNI-M-S ace the-door 'Dan opens the door.' Dan 1>-i-ftah H>aa.b.
O. IMP-M-S-3-openthe-door
Sa~~~e as: (i).
*Dan faatih 1-bsa.b.
Dan open-BfNONI-M-S the-door
(Het>rev)
(Palestinian)
(Patestinian)
1-bsa.l> (lamma lrunna
O. be-PI.Sl- M-S-3 open-Bcooni-M-S U.e-door (vhHc be-PAST-Pl-t
naymiin) (Palestinian)
sleep-BEN.JNI- M- Pl)
'D. bad opened the door vhile ve vere sleeping.'
It is thus reasonable to asstœe that the Benoni for• is, indeed, a pa.rtieiple vhich bas been converted to default present tense use in nodern Hebrew as part of the shift fro• an aspectual to a tense syste:ll.
The difference ootwwn the ten~less infinitive and the tenseless Benoni is characterized by Berman and Doron by assigning the value [0} to the feature
(!Ten~] in the &moni and [-Ten~] to the infinitive.
(16) -Infinitives are [-Tense].
-Benoni forms are [0 Tense). (Berman (1978), Doron (1983))
4.2.2 Analysis of the Benoni and the Position of Tense.
The problem, as stat.ed in ( 13-11 ), is to block V -raising in the Benoni from moving
v
ooyond NumoorO. Since the Benoni form is [0 Ten~]. it surely qualifies as an opaque node in Polloc1C:s oonoo, (Gquivalent to his [-finite]). As such it d09S not tolera tev
-Raising. suppose, now, that we incorpora te Tenoo into the dausal structure given in ( 12) by ma.k.ing it a complement of PersonO, as in ( 17). (Tense"should probably 00 Tense"/Aspect". I retain the shorter label for convenience.)
In .Arabie. The
"ism raam
aay be used as a present tense for:a. only vith a s.all class of perception and stative verbs (e.g., see, hear, sleep, etc ... )lô
(17)
The derivation of the Benoni now proceeds as followrs. V raises to GenderO and NumoorO. TenseO, being weak, lowers to the verbal complex in NumoorO. The HMC blocks V -raising to PersonO since it would cross TenseO, viola ting the HMC.
Tbus, the tenselessness and the absence of Person inflection in the Benoni is not coincidental. The Benoni forms lack Person marking precisely because they are [0 Tense].
To further substantiate this analysis, 1 turn to sorne independent evidence Which showrs that V doos indeed raise to GenderO and to Numt>erO -as opposed to their lowering onto V-and that V -raising doos indeed fait to reach higher that NumoorO, crucially, that it faits to reach TenseO. This evidence is presented in 4.2.3 and 4.2.4.
4.2.3 The Benoni: V-Raising or Atfix-Lowering?
Th~ fact lliat th~ &tnoni is inftoc~ for gender and for number can, in principle,
oo
accounted for either by V-Raising or by Affix-lo~ring. A Ulird possibility is that the verb moves to GenderO and that NumoorO lowers. Note that all three options are compatible with the explanation for why the &noni lacks person marking. This is so since ~aktense
node would, in any event. block V -raising to PersonO.Prima fa cie, bolli GenderO and NumoorO are strong in Hebrew since they are full y distinctiv~ in aH th~ &noni forms. Th us, th~y ought to attract V -Raising.
4.2.4.1 attempts to provide independent evidence for V-Raising as opposed to Affix-1o~ring. Once it is empirically estabHshed that V raises (to GenderO).
pre ven ting it from raising furtb~r (to NumoorO) would either have to be
inde~ndenUy motivatoo - which it is not- or stipula~. On grounds of simplictty,
lli~m. if V is raised to GenderO , then it is also raised further to NumoorO.
4.2.3.1 The Position of VP-Adverbs in Hebrew.
One of the standard diagnostics for V -Raising is the placement of VP adverbs.
Englisb disallo'v'r'S a VP-adv~rb to appear oo~n a verb and its obj~ white French allo'v'r'S it, as is illustrated by the contrastin (lô).
( l ô) a. *John kisses often Mary.
b. John often kisses Mary.
c. *jean souvent embrasse Marie.
20
d. Jean embrasse souvent Marie.
Following Emonds 0978) the pa~rn in Oô) is taken to constitute evidence that V moves
to
Agr over the adverb in the syntax of French, giving the or der V Adv 0, as in ( 19a) but remains in place, triggering affix lo~ring in English, ( 19b).(19) a. b.
With ( 19) in mind, consid~r possible adverb positions in Hebr~w. (2ûa), and the example, (20b). ("''"'possible adv~rb position.)
(20) a. Dan ·sam • 7~t ha-k~f • ?al ha-sutxan •.
D. put-BEOONI- M-S ace the-money on the-toble
Dan • puts A the money • on the table A_
b. Dan sam 1~-?itim krovot ?et ha-kesef ?al ha-sulxan.
D. put-BEOONI-M-S at-ti Ille$ frequent ocdhe-money on the-table
'Dan often puts the money on the table.'
Note that th~ Fr~nch ord~r is possible in H~br~w. with th~ adverb appearing
bet~n verb and objoct.to Itis highly unlikely that the objoct is moved
10 Note that the English order. Adv V O. is also possible. This doe3 not necessarily iaply that the verb optionally fails to raise vhich.
incidentally, is impossible given Chomsky's (1989) Least Effort
21
rightmrd over the adverb beeause it is followed by a subcategorized PP. Thus, we can
oo
fairly c~rtain that th~ objoct NP is ind*<i insid~ VP and not (Heavy NP-) shiftoo out of VP. The order V Adv NP is therefore obtained by V-Raisin& as in French.Similarly, the objoct of a passive verb such as nignav <-was stolen·) must be indefinite, as the contrast between (21a) and (21b) illustrates. This is typical of in -situ objects in passives of null subject languages. tt The presence of the locative NP y ester day ootween V and it's objoct is once again indicative that V -Raising has taken place, as in French.
(2 1) a. Nignav etmol sefer me-ha-sifriya.
stoleo 1Je3terd6y book from-the-1ibrory
'A book was stolen yesterday from the library.'
Guideline. Rather, I take i t to i:ndicate that adverbs may be :aoved clause-·internally in Hebrew. This is apparently not the case in French am !:ng lish.
Clause internai adverb :aove:aent .ay perhaps be due to the fact that aside fro:a a very saall, lexically-aarked class, the Hebrew lexicon contains no distinct category tor adverbs. Adverbs are rendered as PP's, as in (20b) in the text. Qua PP's they :aay be :aoved (scra~led) clause- internally.
11 See Belletti (1968), Borer (1963), Shlonsky (1969c) tor discussion.
22
b. *?Nignav etmol ha-seter me-ha-sifriya.
~tolen ~pterooy the- book from- the-11 br&r y The book was stol en y ester day from the library .'
c. Ha-sefer nignav etmol me-ha-sifriya.
tt.e-book stoleo yesterd&y from-the-llbr&ry Same as (2 le).
I now turn to evidence in support of the claim that V -Raising in the ~noni doos not reach TenseO, but ratller stops at Numb&rO. The evidence is taken from the placement of negation in modern Arabie dialects Which have a Benoni morphology identical to that of Hebrew, as discussed in Footnote 9.
4.2.4 The Place:aent of Negation in Palestinian Arabie.
The negative element in Palestinian consists of two formatives, (ma) and (s].
very much like French [ne] and [pas]. When a non-Benoni verb, (i.e., a perfect or imperfect verb), is negated, the two negative elements appear as a prefix and a suffix on the verb, as in (22a). There is sorne, sub-dialectal variation asto Whether both or only one of the elements are actually realized on Ule surface. However, in all sub-dialocts of Palestinian, When a ~noni form is used, the negative elements are congealed into a single word, mis. (or mus) to the left of the verb, (22b).t2
12 To illustrate this point, I have cbosen the Benoni of a perception verb which per:aits a present-tense (stative) use. See Footnote 9.
(22) a.
b.
Ma-fhimt-s 1-qi~.
009-uoderstood-PAST -1- M-S- 009 tlie-story 1 didn 't und~rstand tb~ story. ·
Mis faahim 1-qi~.
009 uodentooo-BEOONI-M-S the-story
1/you/h~ do not und~rstand the story. ·
This parall~ls th~ contrast oo~n infinitiv~ and ten500 verbs in Fr~nch, as shown in (23), wbich Pollock discu~s.
(23) a. Pierre n~ mang~ pas.
b. *Pi~rr~ n~ pas mang~.
C. N~ pas mang~r ...
d. *N~ manger pas .. _
Following Pollock, let us assume t.hat th~ Neg ~l~ment is g~n~rated bèlow T~nseO, as in (24), taking ma-to be a seo~ marker with ditic pro~rti~.13
This distribution of ais vs. a-V-s is tound in Cairene Arabie as vell (Abdel-tlassih (1981)). Daaascene Arabie distinguishes two formatives,
~ before tensed verbs and auu before Benoni verbs (Covell (1964)}.
Finally, Hebrev bas a fora. (?)ein. vhich is restricted to occur vith
~ verbs. These and other patterns of negation are discussed in forthcoaing chapters.
13 .na behaves as a scope aarker in constructions involving polarity negation. e. g.. M ... l)ada -' nobody'
24
(24) P~rso(
P~r~n·
/ /
'
Persôn Tè~e •
Te se·
Teb~À
-s
Nu~r .."'
Numoor·Num~r ~~nder"
~
/~~·
G~rtd~r
;.r v
To derive a finite form sucb as (22a}, V must move to th~ left of Neg. i.e. it must adjoin TenseO. If movement to TenseO is blockw, as I have claimoo it is in the
~noni form (22b), th~ verb r~ach~s numoor and can go no furtll~r. Ma-th~n
diticizes onto -$.14
ma ... abadan -'never' (cf. 'Ne ... jamais. '}
:ma ... vala -·ne1t~er ... nor·.
14 The acco\mt of the Benoni can, perhaps, be œrried over to past
participle~in French as discussed in Kayne (1967). His account of French Past Part1c1ple agreement rests on the hypotheses that the participle bas an 1gr node and a subject. Under the viev developed in this paper, ve
•ight say tbat in a sentence such as (6), repeated belov. the auxiliary verb takes a Person• coaple:aent vith a veak (dependent) Tense node.
(6}
Je sais coBbien de tablesils ont repeintes.
1 kM'W' 00... many of tables-f-Pl they have repa1 nted-f-Pl
2
r----·---~
~tions to oo added:
• PRO in the Benoni (cf. 6 oolow.)
• Analysis of the negation particle ein in Hebrew which is restricted to the
&mo ni.
4.3 The Tensed Forms.
In the tensed verbal forms, Tensefl is strong
so
V-Raising is obligatory. Since V moves to NumoorO (as in the Benoni.) it is free to move to Tensefl and further on to PersonO. Th us, the verb in a tensed sentence in Hebrew is adjoined to PersonO at S-structure V-Raising to PersonO is diagrammed in (25).'I knov hov many tables they repainted.' (Kayne (1987))
Tbus, contrary to so•e vievs vhich holà tbat the participial phrase is a bare VP (refs.) or a reduced clause, one could clai» that it is a full clause; yet one vith a veak tense. As in the Hebrew Benoni, the
participial verb raises to NuaberO and the weak Ten~e and Per~on affixes lover onto it.
26
(25)
4.4
Infinitives.The non-finit& forms in H~br~w consist of a v~rb form to wbich th~ pr~position
like prefix l-in its various, phonologically determined surface realizations, is attached, (26).
(26) a. Raci-ti li-Sœor ?al ha-xacilim.
'w"8nt- PAST -S-1 to-guard on the-eggphmts 'I wanted to guard the eggplants.'
b. Raci-ti te-?exol xadlim.
...-ont-PAST -S-1 to-eet tJNplant~
'I wanted to eat eggplants.'
2
The non-finite forms are both [-TenseJ and Iack agreement. They should be analyzoo, then, as failing to raioo even to GenderO. Given the clause structure proposed in ( 1 7), one might suppose that the non -fini te verb remains in -situ in VP at S-Structure. Follow.ing standard analyses of infinitives in English, the prefix
t-
is base-generated in TenseO and then lowen~d and adioined
to
V in VP.The infinitive prefix does indood 1ower from a base position in TenseO. but
to
a verb Wbich at S-structure is not inside the minimal VP in wb.ich it was generated.My thesis is that the infinitives in volve V -Raising as well, though to a node wb.ich lies between GenderO and the minimal VP. The relevant aspects of the derivation of the infinitive is schematized in (27).
(27) PerS<(n· (=IP)
·,,
~~r~·
Persôn Teh. se.
Tei'lse'
"··
Tense Nùrnper"
1-
28
Consider the tact that an adverb may intervene between an infini ti val verb and its complement. If
v
were in VP at s-structure, the only possible way to derive the sentences in (2ô) other than by raising the V over the adverb would beto
move the object NP rigbtwards and adjoin itto
VP on the rigbt. White this strategy cannot be ruted out in principle (especially in a language like Hebrew which adroits a relatively free word or der )1 the presence of subcategorizoo PP's to the right of the objed NP in these examples, argues strongly against su ch a move.With the object restricted to remain inside VP in (2 ô), th en, V -Raising applies and V is movoo up and adjoinoo
to
the node labelloofo
.15(2ô)
a.
Carix H-zkor la-sim tamid ?et ha-sefer ba-makom axarey m!J$t-H-S-3 to- remembcr to-ptrt elw!p eœ the-boole in-the p1GCe efter ha-Simus.the-U$C
'One must remember
to
al "Ways put the book in place back afteru~.'
b. La-tet be-me?ora kaze xadlim le-a.xot-xa be-matana, ze-hu to-give on-occasion ~ucb eggplants to-sister-your for- present, it-1$
ma 7ase :xasar takdim.
oct vitflout precedent
To giv~
y
our sister eggplants as a present on such an occasion , that's an unprecedenœd act.'15 As in (20) above, placing the adverb to the left of the verb is also
acceptable. See rootnote 10 for soae discussion.
2
4. 5 The Node
fQ.and
the Syntactic Aspect of Semitic Morphology.McCartlly ( 19ô2), follo\'IÏng a long tradition of Semitic grammarians, argued tllat the consonantal (verbal) root in ~mi tic must oo represented distinctly from botll its vocalic me lod y and tlle skeltal tier to Which it is linked. He showed that universal principles which govern tlle llnking of segments to skeltons can account for word formation processes in Arabie and Hebrew.
Note, now, that the infinitive form consists of a consonantal root linked to a particular skelton and melody; it is not a bare verb. Suppose tllat in Semitic, only tlle consonantal root is generat:OO under V in VP at D-structure.t6
An infinitive form sucb as li-zkor ('to rememoor') is derivoo as follows: First, tlle verbal root z.k.r is moved to
fo
where it adjoins to morpbemes consisting of 16 This iiaediately raises the issue of whether derivational morphology is 'syntactic' or 'lexical'. The proble• arises most acutely in the cases discussed in Borer (196?) involving, aaong others, adjectival and verbal passives. The cases under discussion in this paper. hovever, are unproble.a.tically 'syntactic'. By arguing tbat the bare consonantal root is generated under V I am not co.-itting myself to the claim that pre- linked verbs aay not, in principle, be generated there. There are, indeed, .any rurther questions concerning the place or the morphological component in the gra..ar and its interaction vith the syntacticcomponenent. These issues will be discussed in a later chapter where I analyze passives in Arabie in terms of V-Raising to a passive melody node.
30
a skeltal tier and a melody, ( CCVC and~ respectively). The infinitive prefix
l-
(whicb we can assume to
oo
pre-linked to a skelton) now lowers from TenseO and adjoins to V inf
0.If tlle verb generated under V at D-structure consists of nothing more than tlle consonantal stem, tllen all verbs, finite, Benoni as well as the infinitives involve an
~xtra
step of raising tof
0, a node located~low tll~
various agreement and tense nodes.More generally, once it is accepted tllat tlle ·v~rbal complex· (i.e., the main and auxiliary verbs and the inflectional morpbemes wbich are attach~d to them) consists of a seri~s of XP's, it is not unreasonable to assume that UG places no principloo limit on their numoor. In oth~r words, it is mor~ costly, and bence grammatically marke<!, to rule out
f' '
than to admit it into the verbal complex of Semitic.t?This concludes our initial discussion of Semitic, and spectfieally of Hebrew verbal inflection. 1 turn now to the tl:te tlleory of Pro-Drop wtùch, 1 oolieve, can
oo
simplified by tlle model of agreo&ment presented above.
1? Pe~etsky (torthco•ing) lakes essentially the
saae
point.5 The Pro-Drop Parameter
5.1 The Pro-Drop Problem.
Studies of llli?.-drop in languages With agreement (i.e. not Chinese) all concur around the observation that there is a direct connection ootween inftectional richness and the lkensing of NQ_. At the same time, the notion 'inflectional richness· is generally characterized in no more than vague terms.
In his 19ô6 study, Rizzi mates an expHcit proposai- since then adoptOO by a number of other authors- regarding the licensing conditions for R(Q.. Paraphrasing his proposal, ru:Q. must be both formally licensed (governed and case-marked) wbich it is in Jtalian, but not in English and assigned content by coinde:xation With the governing head.
In Italjan, Jnf1 governs subject PIQ and assigns 1t grammatical features; in English
it
!ailsto
govern NQ. (due to the negative setting of the NQ.-drop parameter), white in mixed languages Hke Hebrew, PIQ.iS formatly Hcensed, i.e., governed by Infl but the degreeto
Wbich it can oo assigned features varies in accordance with the richness of lnfl. Th us, with respectto
languages like Hebrew, the richness of agreement plays a crucial rote in the distribution of RIQ,-drop.Rizzïs tlleory of N.Q.-drop has two conceptually unappealing components: First, tlle stipulation that in sorne languages Infl governs and Case-marks ru:Q.. white in
others it doos not. Se<;ondly, the reHance on an unclear notion of ricbness of inflection, (which his theory shares With most others.)
The tlleory of agreement deveJoped above
o~nsup an alternative view of
RIQ.- drop in which the RrQ.-drop parameter reduces entiretyto
the theories of movement and government. The results of this view, lead usto
concur with Adams ( 19ô7: 13)to
the effect that," ... there is no NQ. -drop parameter perse, but rather a descriptive
generatization reducible
to
other properties, namely, directional government and feature identification:In this section, 1 have little
to
addto
Adams· discussion of government directionality.18 Rather, my intention is to darify the notion of 'featureidentification· and derive it, too, from the theories of movement and government
5. 2 The Hebrev
Pro-Drop Paradign.
Argument null subjects in Hebrew are tully licenced only in the first and second persons, past and future. They are ungrammatical With Benoni verbs. The case of null subjects witll tllird person past and future inflection is more complex They are impossible as subjects of root clauses and steadily improve in direct correlation
18 But see section 6 on Arabie vord order vhere this notion coaes to play a crucial role.
33
With tlw degroo
to
which they canoo
identified contextually or in environmentssu~rficially r~mbling th ose of control. 1 de fer discussion of this case to (5. 4).
5.3 The 'Pro Condition' -Lexical Government: Rich Inflection
= Moye a.
My basic claim is that the only condition
to
wbich ru:2. is subject is (29), itself an instantiation of the lexical government clause of the ECP.(29) Condition on Pro
Pro (in (Spec/Person·], i.e. subject of IP, must
oo
lexically-governed.Under the assumptions of Cbomsky ( 19ô6b), PersonO is lexicalized (and thus made capable of L-marking Tense ·) when V raises and adjoins to it. When V moves to PersonO, the newly-formed lexical head Will also lexically-govern its specifier, .Q!:Q.. under a standard definition of government based on m-command. If
v
does not get up to PersonO, the element in Spec/Person· will notoo
governed and cannot be PIQ. or, for that matter, any empty category excluding PR0.19 Pro is lexically-governed only When the Person node is filled, i.e. wben V adjoinstoit.
1 have argued that in the Benoni forms, V doos not reach PersonO. Consequently PIQ. in the clausal subje<::t position will be ungoverned. This explains why PIQ. is
19 If Spec/Person' is unqoverned in the Benoni, various questions arise v.r.t to the clausal subject position. How is noainative Case assigned etc ... I discuss this and other implications in the tollowing cbapters.
not Hcensed in the Benoni or, for that matter, in any environment in which V is not adjoined to PersonO at S-structure.
In the past and future JSl and 2nd person forms, V doos, ind~, move
to
a transparent or strong PersonO node vmere it can go vernru.
in subje<::t position.Th us, PIQ. is licenced in the context of a past and future tense verb.
In English, ru:2. is unavailable oocause there is no V -movement at s-structure.
Hence, Spec/Person·, is ungoverned and caanot be occupied by an empty category.
It must
oo
filled by a lexical NP.2oTh us, we can derive Rizzfs ( 19ô6) descriptive generalization that argument R!Q.
must be identified by features of person and number without making re<::ourse to special me<::hanisms of feature identification. A language with rich infle<::tion and bence one wbere :ru:.2. is attested, is a language where V can adjoin PersonO and hence go vern ru:2. at $-structure.
20 One consequence to this analysis is tbat .NP :aove:aent in !nglish to Spec/Person' is not Case-driven, but rather forced by the fact that an e.c. is not allowed in the clausal subject position. See Shlonsky (1989c) for further argu:aents to this effect. based on the theory of inherent case developed 1n Belletti (1988}.
5. 4 The Third Person Past and Future Forms.
Suppow that the third person forms are characœrized
by
a W&ak PersonOnode.
Thus, V in th~ forms adjoins TenseO but fails to move up to PersonO, as in (30).
Pro drop is unavailable sin ce .RrQ. in Spec/Person · is not lexically -governed.
(30) Person· (=IP)
Perusal of the verbal paradigm in ( 14) sho-wos that such an analysis can be partiaHy motivated on the basis of the third person past forms. Tbese forms differ from the first and second person ones in lacking an overt manifestation of person morphology.
Moreover, the third person masculine (singutar and plural) forms in Hebrew future and past are homophonous with the impersonal forms wtlich pattern like the Benoni in Hcensing non -argument nul! subjects, as shown in Shlonsky ( 19ô9a).
Th us, (31) can only have the impersonal interpretation in (a) and is
ungrammatical under the persona! interpretation, (b). The similarity ~twoon the impersonal
use
of the third person and the Bênonican
oo seên by comparing (31) with (32).2121 The third person future for•s are •ore co•plex. They sbare vith their past tense variants the aabiguity vith the i.personal form; yet they aanife~t an overt prefix. Ii=_for the .a~culine for.. ti- for the feainine form. One aight speculate tbat these are not person prefixes.
but rather. gender prefixes. (which are perhaps truncated when precede by the person prefixes in the first and second person forms. )
Note that a fora such as [ti-l.ad] can aean either 'she will stUdy' or 'you (H-S) will study'. as shown in (i). Yet vhen the subject pronoun is null. only the second person singular interpretation is available. (ii).
(i) hi ti-lmad sifrut.
a ta
sbe 'Will study literature IJOU
(ii) R!Q_ti-l.ad sitrut.
*she vm study merature IJOU
Suppose tbat the prefix [ti-] is a person .arker in the second person singular for• and a gender .arker in the third person feainine fona.
Only in the tor•er can the verb raise to PersonO and govern ~- In the latter, it reache~ up to ·~trong' ten~e but i~ blocked from adjoining to a 'wealc' pen~on node.
(31)
(32)
Hizminu ?et Dani le-rnesibat xacilim.
i f'IVlte- PAST- H- Pl ace Deni to- party eggplsnts a.
b.
Dani was invitoo to an eggplant party.
*They invited Dani to an eggplant party.
(Tamid) mazmin-irn ?et Dani le-mesibot xacilirn.
(Al ways) i f'IVlte- Bft«)NI- M- PL ace Da ni to- parUes eggplents 'Dani is (always) invitéd to eggplant parties.·
However, ru:Q.-drop with third person infloction is not almys ungrammatical.
Borer ( 19ôô) ct
res
exampl~ such as (33>. which show tllat RIQ. can be llcensed in a control-environment. The subject of the future masculine singular tllird person verb yacliax must be coindexed with the matriX subject. In (33b) it mustoo
coindexed witll the object. Cructally, in neither sentence can it be free.a. Dani hivtiax Ie-Rutij oo-yacliaXi/*j/*k ba-bxinot.
l)(,n ftfflmi~-P~1-t1-S\t~ Ru\i \hfl\-t1-S-~~-fU1in-\bt-cxom
'Dan promised Ru ti that he will succeed in the exam b. Dani hivtiax le-Rutij Se-tacliaX*i/j/*k ba-bxinot
nore generally, these data support the viev tbat there isn't a one to one aapping betveen aorpheaes and agreeaent features, (Anderson (1982)).
Indeed, 1 see no vay to account for the facts in (i} and (ii} other than to assuae that the syntactic representation of agreeaent isn't aapped isoaorphically onto the morphological one.
38
Dao promise-PAST -M-S to Ruti toot-f -S-succeed-FUT in-the-exam
'Dan promised Ruti that she will succeed in the exam:
Borer analyz~ the Agr node in (33) as anaphorically bound to the matrix. Be the details of the the analysis what they may. what is crucial is that su ch control is ungrammatical with the Benoni forms, as Borer herself no~.
Moreover, we find in Hebrew cases of subject ru:2. in contexts wbere its reference can be recovered from the discourse, as in the fragment given in (34).
Again, this is possible only with tensed verbs. Pro with Benoni forms is quite infelicitous, even in narrative contexts.
(34) .. Dan baxan be-?ein-av ?etdeletmaxsan ha-xacilim. Pito'm. WQ
hiSmi?a eeaka ve hexella-rue le-?evra. Ka?avor sniyot sfurot kvar ~
baya bünim, Sax\lv ?al ha-?rema ke-sikor...
.. .Dan examined \ollith his eyes the door to the eggplant store-room.
Sudden1y, he let out a scream and t>egan torun towards it. After se veral seconds, he -was already inside, laying on the pile like a drunkard ...
The difference between the Benoni and the third person forms can be dealt with as folloW'S. In the tensed tllird person forms. V reaches TenseO and faits to raise to PersonO bocause PersonO is weak. In (33) and (34). the Person node is
'strengtllened' by sorne form of non standard binding or control. wbich supplies it with features through coindexation \ollith an appropriate ante<;edent. Thus
3
st.nmgtlled, PersonO now attracts V -Raising and V moves from TenseO and adjoins to PersonO, coming tlms to govern its spocifier and licensing R!Q..
ln the &moni. to recall, it -was a ~ak Tenoo node wtücb blockoo movement to PersonO by the HMC. Thus, even if non-standard binding/control can strengthen PersonO so that it could attract
V -Raising, as in the tensed forms. the
weakTenseO
node would still constitute a barrierro
such movemenl V -Raising would Ulen be blocked and 2!Q could not~ governoo in Spec/P&rson·.aa5.5 Veak Nodes and NUùl Nodes.
Before concluding this section, let us briefly a~nd to the more general problem of bow to distinguish ~ak agr~mentnodes (e.g., the person node in the Hebrew third person) from nodes wbicb are phonetically null yet syntactically strong. ~.g .• th~ g~nd~r nod~ in the Hebrew first p&:roon, \>lhicb atlo\IJ'S movement through it up to PersonO or the gender node in the non-partidpial verbal forms in Romance.
I believe that the language learner can deduce that the gender node in Hebrew and Romance is strong. albeit ooing morpbologicaUy nu11, OOœuoo s/be bas
22 we i:aply tha: t TenseO, 1.mlike PersonO. cannot be bound or controlled.
Tbis .akes intuitive sense if binding and control are properties of referring elements.
40
evidence that V bas raised to PersonO. Given the HMC, V must have gone through GenderO and NumoorO first
On the other band, the morphological poverty of the Hebrew third person past form willtead the language learner to conclude that it is weak and not mer ely unspectfioo. This is so becauoo the other person nodes are aU phonetically non- null and, unlike the case of the gender node, there is no other evidence for V- movement to PersonO.
A spoculation we can entertain at this point is that the language learner assumes that V -Raising app1ies all the -way up to PersonO in the unmarked case.
The unmarked cases ooing, on the one band, languages with full agreement such as Italian and on the other band, languages 'Without any agrooment wbatsoover, e.g.
Chinese. Ot follo-ws that both languages hav~ RrQ.-drop for the same reason:
Spec/Person· is lexically-governed at S-structure.23) Wben a language has partial agreement (i.e. English, French, Hebr~w) it is marked and positive evidence must
be gathered to determine the extent to which V raises.
In this section, I have tried
to
show that the analysis of agr~ment develop&d in this paper can account for the range of null subject data in Hebrew 'Without rocouroo to RrQ_-specific rules. The connection ootwoon rich inflection and the distribution of RrQ. was madero
follow entirely from the thoory of government;Like NP-trace and variables, RrQ_must
oo
lexically-governoo.23 But see Aol.ID and Li (1989) vhere i t is explicitly argued tha:t V does not raise to Intl at S-structure.
4
~to~added:
~istribution
ofnon-argum~nt
R!Q..6 Agreement and Word Order Variation in Arabie_
6.1 vs-vord Order and the Position of the Clausal SUbject.
~tion 1 descri~d and illustrated th~ basic word order paradigm in Arabie. To reœll, wben the verb cornes bef ore the subjoct, it is inflocted only for gender and wben the subject cornes bef ore the verb, the verb bears full, i.e .• gender, num~r
and person agreement.
Th us, in the framework of the theory developed in the previous se<:tions, the verb in V -initial clauses, moves only up to GenderO, While it moves all the way up to PersonO in the subject-initial clauses.
If the verb mo v es only up to GenderO in V -initial clauses, it must be the case that the subjoct is in a position lower than Gender· at S-structure. Under this approach, th~n, the VSO-ho<Xi of Arabie is due not to a V-Raising transformation Which rn oves V to C, as argued, for example by Sproat ( 1985) for Welsh or Shlonsky ( 1989b) for Arabie. Rather, VSO order results from base-generation of the subject in VP and subsequent raising of V, as argued in Demirdacbe ( 1989).
42
I will assume that the subjoct is base generat:ed in S~/V·. As a first hypotl:tesis, then, consider the foltowing derivation of V -initial clauses in Arabie, (35).
(3 5) Perso~(
7\o~
Pefson Tehse"
'ié~·
,. / ~,"Ten$e Nhmber"
The main question Which arises here is Wh y tl:te verb falls to move higher than GenderO. One cannot simply say that V faits to raise higher than GenderO because th other inflectional nodes are weak. First, this is not entirely true, since
Tense/AspectO is actuatty strong in indicative root sentences such as those in ( 1) and (2) above. In more general terms, such aline of reasoning -works for
French/English (cf. Section 2), but its application to Arabie begs the question. This is so since full inflection is available once the subject is fronted, so the nodes are not inhenmtly weak as they presumably are in English. Rather, their weakness in Arabie should be viewed
as a morphological reflex of
theimpossibility of
Vérb raising.Under Cbomsky's Least Effort Guideline, which is central to his analysis of V- Raising in English and French, V -Raising is obligatory unless impossible. This means that unless something prevents V from raising. it must raise .. since
otherwioo the inflectional features willlower and at LF the entire amalgalm will be raisoo again. Such a derivation involves more steps and is hence more complex and less valued tllan one wbere tlle verb raises directly and adioins
to
tlle relevant infloctionalxOs.
It follows, then, tllat V doos not raise higher than GenderO not t>ecause it optionally chooses not to, but t>ecause it is blocked from doing so. We must attempt to discover why this is the case.
As a first s~p. consid~r wny V raisoo rais~ at aH? Wb y d~ it not sim ply remain in VP?
The latter question can be answeroo by referring bac.k to the analysis of Hebrew infinitives in 4. 4 . There, it was argued that a consonantal verbal root raises in Semitic to
f
0 to link with a melody and a skeltal tier. If V mu$t move out of VP in order to amalgamate withf
0, why does it stop at GenderO and not raise all the WdY up to PersonO?My capsule answer
to
this question is as follows. In addition to amalgamating with the contents off
0, V must lexically support the node wbichcaS&
marks the subject. Under the assumption that (in VSO languages) Case is assigned44
rightwards, i.e. in the direction of canonical Government, (Sproat ( 19ô5)) it follows that V mustmove to the rightof the subject before S-structure.
As it stands, though, V governs the subject from its position in
f
0. Y et it does not remain there but obligatorily raises further to GenderO. This is t>eœuse GenderO, though notf
0, is endowed with case features. More generally, I adopt the standard assumption that case is assigned (actually, checked) throughagreement coindexing. While
f
0 con tains inflectional and derivational information crucial to the $-structure in whkh V appears. it is not, strictly speaking. an agreement node and th us plays no role in nominative Çase marking.V raises
to
adjoin a node which cancase
mark the subject. The first such node is GenderO. It need not move any further since gender agreement alreadyestabHshes the coindexing nocessary for Case chocking. Raising higher (to NumoorO. for example) stands in violation of the Minimal Effort Guideline.
Y et GenderO doos not go vern tlle subjoct of SpecN'. The subject must itself raise to a position in which it would be governoo by GenderO in the canonicat direction. Consequently, movement takes place from Spec/V'
to
Spec/r·.
asdiagrammoo in (36).24
I thus follow De»irdache (1989) in arguing that both V-Raising and HP-nove»ent take place in the derivation of Y-initial clauses.
(36)
Insofar as
v
-Raising is Caoo-driven, Wfl predict no movement of V~yond f
ifthe subject can get case by sorne other means. Interestingly, when the subject is non-adjacent to the verb, i.e., wben a subcategorized PP or an adjunct appear oot""*n the verb and the subject, the verb optionally manifests no agreement at all with its subject.
consider (37). In (37a),
v
agrees in gender with Ule subject, an instance of Ule phenomenon ~have ooen examining up till now. In (37b), ho'Wflver, the verb appears in its Ulird person masculine singular form, an agreement-less.ïmpersonal' torm, wnile its subject is feminine.
(37) a. ia?at ?i1ay-na Maryam.
come-PAST- F to-us M.
'Maryam came to us.'
b. Za7a
7ilay -na Maryam.
come-PAST to-us M.
Same as (a).
I ~lieve (37a) to ~an instance of
v·
Reanalysis of V and the PP, an exœnsion of Larson ·s ( 19ôô) analysis of English Dative Shift. I discuss dative constructions in Arabie and Hebrew in a separate chapt-er. Suffiee it to say, at Ulis point, that V+PP is reanalyzoo as a V which then raises tof
0 and to GenderO. the subject is moved to Spec/f'
where it is case marked. Modulov·
-Reanalysis, th en, (3 7a) is derived just like the cases of standard VSO clauses we have been examining.In (37b),
v·
Reanalysis does not take place (its optionality is crucial tor Larsorfs ana1ysis of the alt.ernation t>et~n datives and double objects.) The subject is not in Spec/r·
at S-strucutre. If it were, it would be case marked by GenderO and gender agreement would be manifested on V. Rather, the subject is shifted rightwards (extraposed) over the PP and adjoined to VP, as in (3ô).(3ô)