* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Working paper Reference - Archive ouverte UNIGE
Old Irish grammar wikipedia , lookup
Old English grammar wikipedia , lookup
Swedish grammar wikipedia , lookup
Antisymmetry wikipedia , lookup
Macedonian grammar wikipedia , lookup
Esperanto grammar wikipedia , lookup
Navajo grammar wikipedia , lookup
Udmurt grammar wikipedia , lookup
Scottish Gaelic grammar wikipedia , lookup
Arabic grammar wikipedia , lookup
Chinese grammar wikipedia , lookup
Modern Greek grammar wikipedia , lookup
English clause syntax wikipedia , lookup
Turkish grammar wikipedia , lookup
Italian grammar wikipedia , lookup
Ancient Greek grammar wikipedia , lookup
Kannada grammar wikipedia , lookup
Georgian grammar wikipedia , lookup
Modern Hebrew grammar wikipedia , lookup
Polish grammar wikipedia , lookup
Lexical semantics wikipedia , lookup
Latin syntax wikipedia , lookup
Portuguese grammar wikipedia , lookup
Spanish grammar wikipedia , lookup
Icelandic grammar wikipedia , lookup
Yiddish grammar wikipedia , lookup
Working Paper Semitic clitics SHLONSKY, Ur Reference SHLONSKY, Ur. Semitic clitics. Geneva : 1994 Available at: http://archive-ouverte.unige.ch/unige:83667 Disclaimer: layout of this document may differ from the published version. Semitic Clitics 1 Ur Shlonsky shlonsky@uni2a. uni ge. ch 1. INTRODUCTION This paper is concerned with the enclitic pronouns in Semitic (Arabie and Hebrew}, illustrated by the examples in (1) with the clitic underlined.2 (1} a. kaan bixa)'Yl:-ha 'He was sewing it' Palestinian Arabie b. (he) was sewlng-3FS ~asaan-ha bitxayyf 'Because she sews the dress ... c. because-3FS sews the dress•.. tmunot-eha tluyot ~al ha-kir. pictute-3FS hang on the-wall xasavnu ~al-cha. 'Her pictures hang on the wall' Hebrew d. (we) thought about-3FS 'We thought about her.' Hebrew "' 1-fistyaan ... , Palestinian Arabie An adequate analysis of these clitics must account for the fact that they are manifested on ali lexical categories (i.e. on V in (la}, on P in (lb,d) and on Nin (le).) The analysis must also explain why Semitic clitics are without exception enclitics and why they manifest no overt distinctions of Case. These three properties and others to be presented below, sharply distinguish Semitic clitics from Romance ones, as can be immediately discerned by comparing the expressions in (1) above with the French example in (2) below. (2) Elle l'a cousu. She it-has sewn 11 'She has sewn it.' French and generally Romance clitics appear only on verbs, they manifest Case distinctions and they are either proclitics or enclitics (in French they are proclitics on tensed verbs and negative imperatives and enclitics on other verbal forms). 2. PROPERTIES OF SEMITIC CLITICS The French object clitic in (2) is attached to the auxiliary while the Semitic one in (1) to the main verb. With sorne exceptions (e.g., Val d'Otain Franco-Provençal, see Roberts (1993a}, Brazilian Portuguese (Bianchi and Figueiredo Silva 1993), Romance clitics are attached to the auxiliary in compound tenses. This is never the case in Sernitic. Semitic clitics are invariably attached to the main verb in periphrastic constructions. Insofar as the generalization underlying clitic placement in French or Italian is that the clitics are attached to the highest verbal head in the clause, i.e., the auxiliary in compound tenses, the main verb in simple ones, the generalization holding for Semitic should be stated as in (3). (3) Clitics are always attached to the closest c-commanding head. xo Insofar as clitics are categories, then the chain connecting the clitic to its base position appears to violate the minimality clause of the ECP in Romance, but not in Semitic. The Romance case can be dealt with by assuming that clitics undergo head movement only as a final step. They are generated and move as XPs to sorne intermediate Spec position, perhaps Spec/AgrO, from where the head of this XP incorpora tes to the higher head 1 This contribution is a slightly expanded and corrected version of my presentation at the 1993 meeting of the Linguislic Society ofAmerica. The paper has benefitted from comments and discussions with numerous people. 1 am particularly grateful to O. Cinque, J. Ouhalla, M-A Friedemarm, T. Siloni and the students in my Semitic Syntax course in 1993/94. 2 The Palestinian data are based on one variety or sub-dialect spoken in Northem Palestine. The Hebrew data reflect my own judgements. Throughout this paper, 1 abstract away from the fact that clitics on tensed verbs in Hebrew are stylistically marked. Grarnmaticality judgements concerning them are firrn nonetheless. GenGenP (199) - Shlonsky Semitic C/itics GenGenP (1994)- Shlonsky Semitic C/itics (5a,b) illustrate the dative altemation. In the first sentence in this pair, there is a direct object and an indirect (prepositional dative) one. The second, dative-shifted sentence, illustrates the double-object variant. The second pair of sentences i.e.,(5c,d), illustrate pronominalization of one of the complements. In (5c), the theme is cliticized on the verb and in (5d) the causee, i.e., the shifted object. The crucial data is found in the third pair in (5). When both complements are pronominalized, only one can show up as a clitic on the verb, the other one must find another host. In the first member of this pair, (5e), the direct object is cliticized on the verb while the indirect one on the preposition. (5f) shows that cliticizing both complements on the verb in any order is unacceptable. Put differently, only the prepositional dative construction can be the source for pronominalization of both complements. If clitic clusters are disallowed in CA, th en (5e) is the only option when both objects are pronominalized because only in the prepositional dative construction is there a second host for the second clitic. A final property distinguishing Semitic clitics from Romance ones to be noted at this point, is that Romance clitics have a morphological affinity with determiners (see e.g. Cardinaletti (forthcoming)). No such affinity can be observed in Semitic. To conclude this descriptive section, 1 list the relevant properties of Semitic clitics discussed above, noting that for each and every one of them, Romance clitics have the opposite property. to which the verb is adjoined, perhaps AgrS 0 (viz. e.g Kayne (1987) Sportiche (1990) and much current work). Clearly, this sort of derivation is unavailable in Semitic. A second difference evident from the comparison of (1) and (2) is that Romance clitics may appear to the left of their hosts (proclisis) while in Semitic they are, as noted above, without exception enclitics. In Romance, the choice between proclisis and enclisis is determined by the tense and mood of the verb. No such pattern can be found in Semitic. A third important difference is that while Romance clitics appear on verbs and auxiliaries, Semitic clitics appear on ali lexical and sorne of the functional categories, as in (1) above or in the Palestinian Arabie paradigm in (4) below. (4) a. b. Verb + Object Noun + Possessor: fhimt 1-m\'alme. fhimt-ha. (/) understood the teacher (/) tmderstood her beet 1-m\'alme beet-ha hou se the-teacher her-house 'the teacher's hou se' C. Preposition + Object: min 1-m\'alme min-ha fromthe-teacher from-her d. Complementizer + Subject: ?innu 1-m\'alme ?in-ha rhatthe-teacher thar-she e. Quantifier + DP: kull-m\'almaat kul-hin al/ the-teachers ali-them A fifth property of Semitic clitics which distinguishes them from Romance ones is that there are no clitic clusters in Semitic. The relevant case involves clitics in the double-object construction. 3 Arabie causative verbs and a small number of non-causative verbs rather regularly alternate between a double accusative and an accusative dative complement alignment (see Hazout (1991), Hoyt (1989), Mouchaweh (1986)). Consider the following paradigm from Cairene Arabie (Kenstowicz and Wahba (1980)). (5) a. the teacher CAUs-understand the-lesson ta the-girl b. ?il mudarris fahhim 1-bint 1-dars. c. ?il mudarris fahhim-!! li 1-bint. d. ?il mudarris fahhim-ha 1-dars. e. ?il mudarris fahhim-!! laa-ha. 1 the-teacher CA Us-wuierstand-fter the-lesson the-teacher CA Us-undeJ:,tand-!J. /a-lter ~· ?il mudarris fahhim-ha-!!/!!-ha. the-teacher CA us-understand-her-!J.IJ!.-her 3 Fassi Fehri ( 1993, ch 3, section 1.2) documents counter-examplcs to this claim, from Classical Arabie: ?a~Jaytu-Jm-hu (i) (/) gave-2MS-3MS (ii) «;;arbu-ka-hu beating-2MS-3MS 'l gave you it.' 'your beating him' 1 bdi eve thal the status of these examples is controversial. The Classical Arabie double object construction is endowed with other peculiarities discussed in Ouhalla (1994) (see also Mouchaweh (1986). For example, conlrary lo English and many if not ali ofthe modem Arabie dialects where the order aoAI)BENEFICIAR y" THEME is imposed in this construction, there is a preference in Classical Arabie for the reverse order, as in (iii). (iv): (iii) ?a~Jaytu ?al-kitaab-a Zayd-an. (iv) ?a~!aytu-hu ?ai-Jaalib-a. (/) gave the book-A CC Zayd-A CC {/) gave-3MS the-studeni-ACC Lit. '1 gaw the book Zayd.' Lit. '1 gave it the student.' TIIE INCORPORATION ANALYSIS a. V' v the-teacher CAUS-understand-!1. ro-the-girl * 3. (7) Cairene Arabie the-/cacher CAUS-tJnderstand the-girl the-lesson f. Properties of Semitic clitics a. They occur on the right of the ir host, never on the left. b. They are always attached to the closes! c-commanding head. c. They appear on ail lexical categories and on certain functional ones. d. They do not manifest case distinctions. e. They never cluster. i.e. a single clitic per host. f They bear no morphological resemblance to nominal determinees. One thread running through generative research on Semitic has been the view that the enclitic pronouns are affixed to their governor under linear adjacency. Thus an object clitic on a verb is affixed in the configuration in (7a) and a prepositional object is affixed to a preposition in the configuration (7b). ?il mudarris fahhim 1-dars li 1-bint. 'The teacher explnined the lesson to the girl.' (6) p clusters are not pennitted in Semilic. A more delai led study of the synlax of the double-object construction in Classical Arabie is required before the and slatus of (i) and (ii) can be dctennined. - 2- DP Broselow (1976) argued that the clitic is right-adjoined to its host. Fassi Fehri (1993) espouses a similar view, based on Baker's theory of Incorporation. This approach is consistent with the locality of cliticization i.e., the choice of host. In severa! respects, however, it is insufficient. First, it fails to explain why Romance clitics, which are also presumably x<> categories, do not give rise to a consistent pattern of local affixation to their governor. The difference between Romance and Semitic would have to reduce to something like the following. While in both language families, clitics are incorporated heads, Incorporation in Romance can only take place once the clitic qua XP, has moved out of its base position, in particular to a position immediately governed by the highest verb in the clause. Semitic clitics cannot avail themselves of this possibility. This would in tuen lead one to say that the difference between the two language families has to do with the Iicensing conditions on pronouns. Romance clitics are presumably licensed only once incorporated into AgrS, while Semitic pronouns can be Iicensed by affixation directly onto the category governing them. It is not clear why this should be the case, whether it constitutes a primitive difference between pronouns in the two language fa mi lies, or whether it is due to the setting of some parameter. A further difficulty with the approach to Semitic cliticization in terms of local affixati01Jincorporation is that it provides no account for why the clitic appears always on the right and never on the left of its host. It could be that, contra Kayne (1989, 1993), there is a fundamental (perhaps parametric) difference between Semitic and Romance in the direction of incorporation of heads to heads such that in Semitic, head movement al ways yields right adjunction. If this were the case, wc would expect othee types of incorporation to follow the same pattern. However, to the degree that one can fi nd examples of overt incorporation in Semitic, left adjunction rather than right adjunction seems to be the mie. A particularly illuminating example is provided by a construction named Participle Inversion (Pl), studied in Borer (forthcoming) and illustrated in (8b) below. Given this and other peculiarities of the Classical Arabie construction. il seems hasty lo regard (i) and (ii) as counter-examples to the claim thal clilic relev~nce P' b. DP -3- GenGenP ( 1994) - Shlonsky S'emilie C/itics (8) a. GenGenP ( 199) - Shlonsky Semitic C/itics Dani haya lofer smalot. Da ni was-MS sew-f>ART dresses b. Dani lofer haya smalot. Dani sew-f>ART was-Ms dresses b. Hcbrew ktivat-o ?et ha-ma?amar hirgiza ?et Miryam. wriling-he lerrhe anic/e angered lei Mi1yam 'Dani was sewing dresses.' 'Da ni was sewing dresses.' In llli construction, a participle nppcars to lhc !cft of Lhe auxiliary. Borer (forthcoming; see also Doron (1983)) ?rgues thal the participle incorporates to Ille a ux.iliaty, forming an inseparable unit with it. That incorporation is mdeed involved in (8b) is evidenccd by (9) and ( 10). In (9a) an adverb may intervene between the auxiliary and tlle parûciple when the order is !luxiJiary A parliciple, but as (9b) shows, the adverb may not intervene when the arder is parûciplc "auxiliary. The arder of constituents in (13) strongly favars a Noun-Raising account, according to which a D-structure representation where the agent of writing is the specifier of the NP is transformee! into a head-initial one by raising of the No across it to some functional projection F 0 {D 0 or perhaps Agr 0 (Ritter, (forthcoming), Siloni (1994)). In the derived S-structure representation, the agent argument is not a sister to N. Rather, it is exceptionally governed by it, as in (14). FP (14) (9) a. Dani haya tamid tofer smalot. * b. Dani was-MS a/ways sew-PART dresses F' 'Da ni was al ways &wing dresses . ' F Dani lofer tamid haya smalot. Dani sew-PART a/ways was-MS dresses 'Da ni was always sewing dresses.' ktivati beyalduto haya Dani tofer smalot. in-yoruh-his was Da ni sew-PART dresses b. * 'ln his youlh, Dani used to sew dresses.' beyalduto tofer Dani haya smalot. in-yourh-his sew-PART Dani was dresses 'ln his youth, Dani used to sew dresse s. ' The conclusion thal Participle Inversion involves incorporation seems quite reasonable. The relevance of this .construction to the foregoing discussion is Ùlal insofar as the particip le raises to adjoin to the auxiliary, it adjoins on the !cft of lhe auxiliary and not on Ille righl. 4 lf one main tains thaL clitics are incorporated to their heads, then one tr:usL perforee compl ica le the clescript ion of Semitic incorporation by adding an exception clause affecting the dtrccuon of the attaohment of clitics (or converse ly, of inve rtcd participles).s A further sel of difficullies wit11 lhis family of proposa is is signalled by an observation of Brosclow's (Broselow 1976) thal while postverbal direct objects ca n and must be encliticized onto the verb, postverbal subjects cannat. Consider the following Hebrew case. When triggered by the presence of a Topic, Hebrew verbs may precede their · subjects. (Il a) illustra tes inversion of a non-pronominal subject and (1 1b) of a pronominal one. (11) a. be-sofey savu)a roxev Dani )al gama!. on weekends b. rides Dani on-came/ Hebrew 'On weekends, Dani rides a came!.' be-sofey savu)a roxev hu )al gama!. on weekends rides he on-came/ 1ides-JMS on-came/ DP 4. IS SUBJECT-AGREEMENT A CLITIC? Fassi-Fehri (1993) argues that subject agreement is an incorporated pronoun. According to this view, in a VSO language such as Standard Arabie, the subject pronoun is incorporated to the verb, exactly like an abject pronoun. There are severa! difficulties with this view. Consider the following Hebrew case. The non-clitic negative head .?eyn cao be followed by a subject pronoun, as in (l5a) . .?eyn can also be followed by a suffix, identical in fonn to a clitic, (l5b) 'On weekends, he rides a camel.' a. ?eyn ?ani ?ohev xacilim. NEG 1/ikes eggplams b. be-sofey savu)a roxv-Q )al gama!. on weekends N The structural relationship between a verb and a postverbal subject is probably identical to the one schematized in (14), with the appropriate change of category labels. Thus, there seems to be no obvious structural constraint blocking incorporation of a (postverbal) subject onto a verb. The paradigm in (13) also shows thal there is no ban as such against subject clitics (where subject is taken to be the D-structure specifier of the host). This is further reinforced by ( 4d) above, where a clausal subject is encliticized onto a complementizer. The ban on subject clitics holds only for verbs. This, in turn, strongly suggests thal we pursue an explanation of this ban in tcrms of a complementary distribution of verbal agreement and subject clitics. One might argue that subject agreement is itself a clitic, just like say, an abject clitic. Conversely, one might argue thal ali clitics are manifestations of agreement. The following section dcvelops an argument against the first approach. (15) * N' DP t·1 The following example shows that the postverbal pronoun cannat be cliticized onto the verb. (12) NP Dan Similarly, a postver?ai subject can occur between the auxiliary and the participle when they are in thal ordcr, but not when the order ts reversed. (10) a. 'His writing of the article angered Miryam' Hebrew '1 don'! like eggplants.' ?eyn-(!!)i ?ohev xacilim NEG ]MS likes eggplanrs '1 don'! like eggplants.' 'On weekends, he rides a ca mel.' ~ne might argue thal the ungrammaticality of (12) is due to structural reasons, namely, thal clitics must be stste~s to ~heir host and a subject is never a sister to a verb. This argument is belied, however, by severa! constderatwns. Under the incorporation approach, (l5b) is a variant of (15a) with the post-negation pronoun incorporated into it. This, however, falls short of explaining the following paradigm, which shows that when the clitic corresponds to a third persan pronoun, the sentence is ungrammatical without an overt subject (contrast (16b) with (l6c)). Clitics in Semitic, as we have seen, occur on nouns. In particular, they occur on derived nominals, as in (13b). (16) a. (13) a. ktivat Dan ?et ha-ma?amar hirgiza ?et Miryam. wriring Dan lerrhe-anic/e angered let Milyam Hcbrew b. 'Dan's writing of the article angered Miryam' ?eyn hu ?ohev xacilim. * NEG he likes eggplams ?eyn-o ?ohev xacilim NEG c. Hebrew 'He doesn't like eggplants .' Jms likes eggplanrs hu ?eyn-o ?ohev xacilim he NEG 3ms likes eggplant 4 Borer (forthcoming) argues that Parlic1ple Inversion is a Iso subject to a lowering derivation, in which the auxiliary Jowers and adjoins to the participle. We put aside discussion ofthis alternative. The paradigm in (16) strongly suggests that the suffix on the negative particle .?eyn should not be analyzed as an 5 Conceivably. clitics could adjoin to the lefl of their host• but subject lo reordering by the ru les of Semitic morphology. 1 know of no indcpendent motivation for su ch a nlle. incorporated pronoun, but rather as an inflectional marker, a manifestation of subjcct agreement. This is so since -4- -5- GcnGcnP ( 199) - Shlonsky .S'emilie C/ilics GcnGcnP (1994)- Shlonsky s· . . ' C/11/ 1IC ct·11/CS . 5. it pallerns wirh ali otller instances of tlùrd . . . r7fererni~I nuiJ subjcct With pro unliccnsed person lltfl~ctlon Hebre~ in failing to appropriately identify a Jhe sem • Pro drop propcrt.ics of Hcbrew . an overt subJect must be rcalJZed, as in (16c). berween an inflectional affi·· aJld fl · (see e.g. Borer ( 1983) 0 986)) th us make it possible to distinguish ,, • Il •ncorporated prono Tl Il paltem charactcrisLic of the VSO SV . . . un . 10 eyn construction gives rise to the agreement . 0 alternauon m Sta da d A b' prevcrb<li subjects. However, sincc referenûal null . n . r ~ IC where verbs fully agree only with m ~tandard Arabie, one is at pain to decide whe subjects ar~ licensed Ill the company of third person inflection subjcct orwhether il contains aJl inco d ther theve~ 10 c.g.(l7) bears an inflection which 1icenses a null ' rpora 1e postverbal subJcct. 111 ?akal-uu 1-ta~aam. (17) ale-3MPL the-food Sin j:,ru lar 1 2m 2f 3m 3f ttY- t- Stem suuf Plural Suffi x suuf suuf suuf suuf bitxayyt-Q-S (20) Palestinian Arabie 'She does not sew it.' 3F-IMPERF sews-J.Ml-NEG Similarly, in Hebrew Participle Inversion, an object clitic must be carried along, since AgrOP in Hebrew (and by extension, in Semitic) is located above VP and below the base position of the auxiliary (Friedemann and Siloni (1993)). Dani haya tofer-.!!!!!. (21) a. lnjleetional paradigm of M ('see') in Palestinian Arabie. Prefix ?a- Assume that Semitic clitics are Agr0 elements and consider first objcct clitics. Suppose that in Semitic, Agr0° may contain an affix. Thus, when the verb raises out of VP, e.g. over negation, (20), il must adjoin to AgrOO on its path upwards. Slandard Arabie 'They ale lhe food., The marker of subjeet agreement marker on Jc . . . . howevcr. subject agrecmeJl! · ·~ S . . . . C?yn IS umquc m Lhat H fonnally resembles a clitic. Typically . · enuuc rs e:·:pressed by a d' . . ' whrch , to recall, appe:tr always 10 Lhc r ig h~ f LI • para Jgm morphologtcally di stinct from thal of clities othcr hnnd, is both su ffixa l <IJid prefhal o . le. stem and do n?t vary Case-\visc. Subj ect agreement on tlle 111 crb. ( 18) provides the p;w1digm of s;lbj~ ~arymg ac_corda nce WJ~h . Llle Tense and Aspect specification of the clitics. c <~greement Ul the Palcsttman Arabie imperfect fom1 and ( 19) thar of (18) AN AL TE RNA TIVE 1 2m 2f 3m 3f -i Prefix Stem nsuuf tsuuf tsuuf 1'suuf vsuuf Suffi x -u -u -u -u (19) Clitie Forms in Palestinian Arabie SinguJar 1 -(n)i 2m -ak 2f -ik 3m -o 3f -ha Plural 1 2m 2f 3m 3f -na -ko -ko -hin -hin Hebrew 'Da ni was sewing lhem.' Dani was-MS sew-PART-Ihem b. Dani tofer-am haya. 'Dani was sewing lhem.' Dani sew-PART-Ihem was-MS c. * Dani tofer haya-am. 'Dani was sewing lhem.' Dani sew-PART was-Ms-them The position of the clitic on the right of the verb, i.e., the fact that clitics are invariably enclitics (property (6a)) is a consequence of the fact thal the verb incorporates into Agr0°, and not vice versa. The strict locality on clitic attachment, i.e. the fact that abject clitics and Semitic clitics in general are manifested on their 'closest' host, (6b), is a consequence of the Head Movement Constraint. The contents of AgrOO must appear on the main verb in periphrastic constmctions and not on the auxiliary because AgrO c-commands the main verb and nol the auxiliary. One major difference bctween Romance and Semitic clitics thus emerges: Romance clitics are XPs at D-Stmcture while Semitic clitics are affixal (Agr) heads at alllevels. A problem arises since object clitics lie on the outside of subject agreement. This leads me to abandon the assumption that AgrS 0 in Semitic contains an overt affix. Rather than taking subject agreement morphology on a verb to constitute the affixal contents of AgrS 0 , I suggest that we consider it to be base-generated on the verb. Movement of V to AgrS 0 is motivated by the need to check the appropriate features, as in Chomsky (1992). The different properties of subject agreement and clitics discussed in the previous section, can be correlated with a difference between lexical affixation and syntactic affixation. Proceeding, let us suppose that whenever an affix appears in one of these Agr heads, a DP bearing the appropriate features appears in its Spec. Take this DP to be a referential pro, as in the schema in (22) below, where X is a member of the class of clitic-bearing heads (cf. Sportiche ( 1992)). (22) AgrP DP pro Agr' Agr XP clitic Addilionally, subject ag ree ment is full (i.e. it inco orates ers . forms; Ule latter are i.nflected on lv for number andrp d pct·o~, number and gender) only m the non-participial . 1 . . gen er. ltJcs however can o .. 1 . , ccur on particlp es, Ill fact, on a ll 1exJ~1 categones and they a lways incorpora te a full set of ~-fea;ures Al~ngslde the syntacti c and morphological considerations which . . cast doubt on the VIew that subject agreement is an mcorporated pronoun thcre is a comparative an 1 l . l be eq uated with clitics. 'Berber ; 11011-Se 't· 1 ge w uc l further suggests that subject agreement should not nu Ic anguage shares a numb f · ·· • Senutic. In particular, subjcct agreement in Berber formali re bi er ~ . morphological properties with otber hand, are second-position clitics attracted t tl 1 . y sem es the SemitJc pattern. Berber clitics, on the a re stacked (s_ec Ouh<lfl a (1988) fo~ di~cussion.). o le ughest overt head in the clause, to the right ofwhich they lf agreemeut IS not an incorporatcd pronoun then perha . ps mcorporated pronouns constitute a form of agreement In lhe following section I 'lrglle il' "lll . . l'.k 1 • < · ... us 1s 1 e y to be the case. 6 · ~'tud 6 A second f.lrnily of analyses of Sernitic clitics, represenled by Borer's 1983 o .. . IJ 1 ( ) y f Hebrew ciJtlc doublmg should he mcntioncd at this pomt. Bor~r reasons !hat since clitics (in part· 1 Jeu ar on ae >rew nouns) can be doublcd the clitic d same base position. She proposes thal Ule cl! . b . ' • 1 Je IS ase-generated on ils host where it r. an the doubled NP cannot he laken to occupy UJe With the NP it doubles. The structure she propos r. r ., . . . . ~ust, or reasons thal need not concem us hcrc. be co-indexed XP es or cille constructiOns Js gJven m (1) below, where X=N in Hebrew. (i) X+cl X' x Let us further assume, with Chomsky (1992), that AgrSP, AgrOP etc. are mere heuristic labels. The content of Agr0 is identical for ali Agr heads. This derives the absence of Case distinctions on clitics, (6d). One expects Case distinctions to show up on nominals, i.e. on XPs, and Standard Arabie indeed manifests a robust system of morphological Case on DPs but not on clitics. The fact that clitics appear on ali lexical and some of the functional categories, (6c), should now be taken to indicate that ali lexical categories have Agr projections, so that a PP or a CP are actually dominated by Agr phrases. This rai ses a serious problem for selection. If a verb selects a proposition, i.e. a CP, how can selection be satisfied if CP is dominated by AgrP, just as a VP or a PP? Suppose that Agr phrases are not visible for selection but are contributed by the syntactic component to enable feature-checking to be carried-out in a Spec-head configuration (see also Kayne (1993)). This allows us to capture the intuition that contrary to other functional projections, such as TP, DP, AspP or NegP, AgrP plays no coherent semantic rote (viz. also Speas (1991)). Tilis approach raises similar problems to Uwse mentioned in connection with the Incof1Joralion !heory of clitics. Among olhcr things, U1e basegeneration approach provides no exp lana lion for the locality of cliticization. TI1atthere cannol be an inherent link belween the possibility of clitic doubling and the locality of clilics is evidenced by the fact thal at !east some Romance languages manifest both clitic doubling and clitics on auxiliaries. Like the lncof1Joration approach, the base-generation lheory of clitics providcs no clue as lo why Semitic clitics arc enclitics and never YP proclitics. -6- -7- GcnGenP ( 1994)- Shlonsky SemilicC/itics GenGcnP ( 199) - Shlonsky Semilic C/itics Chomsky (1 991) for exam 1 · · . . . P e, cruc1ally reqmres that AgrP deJete m LF. Whether AgrP 1s inscrtcd or dcleted is essentially a tl invisible to all~~~(-~~~~rnal ques_tion. 1 s~Iall merely assume, that AgrP is, in sorne sense to be made precise, Ct'. p Y_ syntactlc operations. Thus when e.g. , co selects a proposition, it selects a TP AgrSP bcing irrele va n s.Or select.Jon. • The formai similar'ty b t .. 1 ca tego 0 e wee.n cl1t1~s and determiners in Romance suggests that they belong to one and the same Serniti;'·(~~It:e 1~ D • (Cardmalettr (~ortl~c.om i ng)). By a similar tokcn, the absence of any such si rniJarity in · s JUSt what one e>.:pects tf clr tJcs arc AgrO clements and detenniners be long 10 tJle eatego 0 o T . · b ul~mng .10 the absence of clitic cl uslcrs, (6c), note thal this does not follow from anytJi ing yet The ryb · f o 1 ~l.lbject and abject cli tics showing up on a verb simply follows from tJte hypothesis tllat. A Sa dscncc o cont;t1n an affi ·:x • so that a verb base-gcnera ted w1th . subject . agreement can move through A r0 · k gr h oes ·not agreement affix and move further up to AgrS. g • pre up t e abJect The relevant paradigm to examine is one where a verb has more than a sin 1 · · . by the double~bject con~truction, manifested in numerous Arabie dialects. g e object. Such a test case IS provJded ~ecall the Cauene Arable paradigm in (5). (5f) in particular illustrates th . . Vlew that Arabie clitics are Agr heads allows us to fonnulate s' . 1 h e u?g~am~aticahty of clusters. The the following terms. uccrnct Y t e restnctJOn vrolated in this example in (23) There is only a single AgrP associated with VP. Consideration of (5c,d) shows that the single A r ro ·ectio . . b~ assocJated With VP can be headed by an affix corresponding to either the direct object ('ic) or tl g. ~ - ~ of Case distinctions on the Agr heads. le 111 lrect o ~ect (5d). We take this to be related to the absence Why should (23) hold? A straighlforward and princi led a · · movement in tenus of domain extention ande t ns\~r ls.provJded b~ Chomsky's (1993) approach to Ahold, i.e. , where VP is domimHed by two Agr proJ~~ti~~nasr~~e~ in ~~;~~~aa(f9o;:;~.~~ structure where (23) does not .J. (24) AgrPI A gr' Agr 1 A gr' VP V' V VP Subject V' VP IO v V' Chomsky's proposai is that A-movement of a DP k' tl mo ed to a higher head and cxlendcd il s domain ;t~l ~ 1p ~e closest C-cor~1nta l~ding Spec position if V has position arc rendered equidistant from the Large; S us 1\~o. smce the extnlclron Sll~ and the intermediute Spec crossing over tJlc base subject position 'lnd for s b' p~c. ~~~ system allows fo r ObJCCts to raise lo Spec/AgrO only one Spec can be skipped Thus if,one ob·e~t ~cc to ra~se over objects in Spec/AgrO. Crucially, however, subject w~uld be strandcd in VP. Sir~tilarly, if~ne ~~?:c~I~o ~ec/Agr2 in (24) and the oL!lcr to Spec/A.grl , the ~ ove 10 Spec/ Agr2 and the subJect 10 Spec/Agr 1, the second ObJect would be stranded. Pronominalization of bath abjects in <lirene Arabie re uire .. . shows up on the verb, indicating thal the single A rP abo~e ~ ~le ~reposi_twnal dative constru.c~io.n . On~ clitic on the preposition, indica ting that the prepositior~s a . t d Aas p etn tnggered. The other chtlc Is mamfested ~r las been ~sed and that p has raised to its respective Agr. Suclt a derivati on is fully consistent ,:i~~o~\a e lOmsky s approach, smce the two AgrPs lie within the do ma ins of Iwo distinct heads. . . A .question ari ses with respect lo the derivation of (5b) nam 1 thrs construction - which rescmblcs the famil'a E r ·, e.y the double obJect construcLJOn. The properties of For the purposes of this chapter s m, . 1< r u.g as l one rn ma ny details - merit an independent di scussion ( . u . rce rt to suggest thal the 1 NP (b 1 · · accusative in Modern Standard Arab 'c) . . f: . wo s Oll marked w1lh morphological possibili ty is thal they are LF-mov~d ~~e m s~mc asluon related to the same Spec/Agr, albeit in LF. One Altemativcly, on e objcct is raised to Spec~ A a ur~\J (perllmps ~long ~lt~ li nes suggested in Johnson ( 1991)). in Friedcmann and Si!oni (1993). gr an IC ot Jer Ob.Ject adJOiltS to it or to AgrOP in LF (as proposed 7 llnlike Larnon (1988), however, the indirect ob'ect in (24) is • . . . and Shlonsky (forthcoming). ~ , mapp~d onto a hJcrarclucally lugher position tltat the di reel object. Sec Belletti -8- If Semitic 'cli tics ' are .XO categories then there should be no principlcd ban on cl itic doubli ng, since for evcry xo, there will always be a YP (Spec) posi tion associated with it. (In fact, there arc two, Spec/Agr and the 8 positio n of the argument. ) This is quite clea rly confirmed. The followi ng paradigm illustrates the regularity of this phenomenon, abstracting away from the question of why individual languages choose only a subset of these possibilities. (see Sportiche (1 992) for a proposai along these li nes fo r Romance clitics.) (25) a. fllimt-ha la 1-m'lalme (1) tmderstood-3FS 10 the-leach er b. Pa l ~sti ni an ' 1 underslood the Arabie teacher.' beet-ha la-1-m'lalme house-3FS 10-the-teacher c. 'the teacher's house' min-ha Ia-1-m'lalme from-3Fs lo rhe-teacher d. 'from the teacher' ?in-ha 1-m'lalme ... that-3FS the-teacher 'th at the teacher ... ' Notice that (25a-c) differ from (25d) in that the former contain a preposition, (/a), preceding the doubled argument, while the latter one does not. This preposition has been viewed as a dununy Case-marker inserted once the Case typically assigned by the head of the phrase to ils the argument is absorbed by the clitic.(viz. e.g. Borer (op. cit.)). 1 think that the idea this is fundamentally correct and can be used to explain the absence of the preposition in (25d). Suppose that the clitic, i.e. the Agr head, requires coindexation with a Spec at S-structure (i .e. its nominal features are strong). The question arises asto the position occupied by the doubled NP. Is it Spec/Agr oris it the 8-position of the argument? Word order considerations give no real indication, bccause in ali of (25a-c), the head+clitic maves above the argument in question. Let us suppose, however, thal the argument does not rai se to the Spcc position, as a consequence of Procrastinate. One might reason that if the possibility exists for an expletive to fi li Spec/Agr, allowing the doubled argument to remain in-situ until LF, this possibility is always prefcrred. Now, if the featurcs of Agr arc checked against expletive pro then they will no longer be around to check those of the argument. One might now consider the dummy preposition as the head of a PP which cornes equipped with an associated Agr. This AgrP is not activated at S-structure, since that would recreate the problem thal the prepositional Agr is designed to resolve. Rather, AgrP become active in LF, triggering head movement of the preposition to AgrO and allowing the argument to move into its Spec. This account entails a prediction (not entailed by the Case absorption idea in its classical guises). If there are cases of overt movement to Spec/AgrO, i.e. where Procrastinate does not apply, thcn wc should expect cliticdoubling without a preposition since Checking of the agreement features can be propcrly implemented through Spec-Head agreement at S-structure. When the direct abject in Palestinian Arabie (and in many other dialects) is a (universal) QP lacking a lexical argument, clitic-doubling wilhout a preposition, or rather, object agreement is obligatory, as in the paradigm below. (26) a. (1) saw-31'L aii-3PL b. * suft kull-hin. c. * suft-hin kulll-wlaad. (/) saw a11-3PL (/) saw-3PL al/the-chiltiren d. Palestinian Arabie suft-hin kull-hin. '1 saw them ali.' '1 saw them ali.' '1 saw ail the children.' '1 snw ail the children.' suft kull 1-wlaad. {l) saw allrhe-children Recall from (4) above, thal Semitic Q is one of the non-lexical categories dominated by AgrP, which is triggered whenever the complement of Q is a pronoun, i.e., pro, as weil as in Q-Fioat configurations (Shlonsky (1991). Suppose, now, thal when a direct abject QP (or rather, the AgrP dominating it) does not contain a lexical complement or quantified DP, it is obligatorily moved to Spec/AgrO in the syntax, i.e., it patterns with pronouns in not being able to procrastinate movement to LF. This in turn forces the activation of AgrOO at S-structure, i.e., the manifestation of an affix in Agr0°, titus giving rise to clitic doubling as in (26a). No need for a preposition arises, since the direct object can check its features directly with the clitic in Agrüo. When however, Q contains a lexical complement DP, QP does not raise to Spee/AgrO and clitic doubling is impossible, (26c). Fronting of QPs to some middle-field position is not unique to Arabie. A similar phenomenon was noted in Kayne (1975) for French. (27) shows that a bare quantifier must surface to the left of the past-participle. One is tempted to consider the position of Ieftward-moved tout in French to be Spec/AgrO with the difference that - 9- GcnGenP (1994)- Shlonsky Semilie C/ities GenGenP (199) - Shlonsky S'emilie C/ities French Agr0° does nol conlain an alli;~; When tout · b · · (27c,d). .. IS nol are (or whcn JI •s stressed), judgemenls are rcversed, (27) a. J'ai tout vu . 1 have ali seen b.?? J'ai vu tout. 1 have seen ali c. * J'ai tous les enfants vu(s). 1 have al/the children seen (pl) d. '1 saw everylhing .' (29) '1 saw everylhing.' Suppose (with Koopman (1992)) that atonie pronouns must occur in Spec/Agr at S-S, i.e., they cannot procrastinate Checking. Given (29), pronoun Case-checking is blocked. This in turn leads to the reanalysis of pronouns as Agr heads, as proposed above. In the Romance languages, and more generally in SVO languages where (29) does not hold, the need to reanalyze pronouns as Agr heads does not arise. '1 saw ali the children' J'ai vu tous les enfants 1 have seen ail the children r· s co FINAL REMARKS Let us llO\ ask how lo best characterize the diifcr b 1 R . ·· clilics. This issue c ncapsu lates Iwo di stinct q e~Jc~ Wetl\\eeSn ~~~lance and ScmH•c undcrlying the bchavior of uenes. 1y ernJIJC lan cn ,ages Jack Ro r· · conver ely, why Romance Jacks the Semilic typ If R . . . o' o< mance-type c Jllcs and move ment, i.e. functiomll hC<Ids lacking a 1 • 1e. . ~mance clJtJcs arc D C<Jtegorics Lhat undergo headplausib ly be rel,atcd loth~ nmure of its det,c r c~ J ca res,tnctJon, Lhcn tllc absence of D movcmcnt in Semitic miglll . . ' n11ncr sys em. DefinJtc deternHncrs in Semitic are lhemselves de end J which disûnguish them from Romance ones · Th/ e 11 morphe_mes. Moreover. they have Iwo other propertics only on tlle head noun, but obligat~rily on aÏ r ils ~l:~~;.r agree Wllh ll•e h~ad noun aJ~d . lhey are manifestee! not ;walyzes this type of determiner it seems cle;.u tl 1 't . ers, even thosc wlllch a rc conjomcd, (28). Howcver one frec-st:mding functiomll head as its Ron1anc la ' cannot be detached from the noun il determines. it is not a • e countcrpan. (28) ha- ulxan ha-xum ve-ha-?arox Hebrew rh e-rable-Ms the-bmwn-Ms and rhe-long-MS 'the brown and lall tabk' 1 co nclude therefore th(lt lhc absence of Romance 1 J' . . .. D0 (sec a Iso Borer ( 1989)). ' - ype c Ill cs Ill Senut•c correlates with the non-separatabilily of Let us turn Io every student diachronica lly reanalyzed as the seco nd question . namoly the ab f S .. .. . of Semitic th·H clitics m~rpholo s.en~~ o cm~:c-typc chtJcs Ill Romance. lt is self-evident to derived from ;hem The d ' 1 ~Jca y ~esen~ e pronouns and preHy clear thal they are Ag r heads r believc. tl·•at tl ~acf:uon~c questiOn IS thcn wh 1 ntonic frec-standing pronouns arc · c li S nct 1s strongly correlated 'tl · la ngu<lges in question are a li (M !east diachron.·call ) VSO "'' 1 tl no 1er, narnely that the Semitic 1anguages.9 Y u 8 Standard Arabie indicative c . . omp appears to have an accusative Case feature which 11 · · · · munedmtely follow it (giving rise to e g . . . must dlscharge. Th1s forces e1ther sorne argument to · ., accusa 11ve-marked subJects) or a clitic to be manifo d · , ·· system contains an Affinnation PIU'a-'~ dominating cO lt. . 1 . . este on Il. Khalmh (1993) argues lllat the Arabie Comp 'fo · · . 1s qu1te p aus1ble that dmlectal At 1 · argue thal it is AftP and not CP which is d . t db a lie mam ests a smular Comp structure. One might then S omma e Y AgrP. Following Shlonsky (forthcoming) let us label this Agr AgrC uppose, moreover. that Age{: in Arabie always contains an affix. TI!is would inv ' .· . complementizer .il·nnu to b~ the head of A p TI • S • f . . . .. olve takmg the final vowel of the Palestinian Arabie . . gr · 1" pee o tlus proJect1on1s filled by a ho ·t'· JI 11 · ·. · · · lake, tollowmg Belu!Ïs (1986) to he r~l:' t' 1 p n" Ica y-nu tt-hke pronoun, as m (1) Whlch we may "'H!fli!ll Il\, (i) b~f.1kkir ?im1-u Mona ?ajat (/) think (il) that-3MS Mona came L' , . . t' 1 . Il. 1 thmk Il (to be the case) thal Mona came.' ema Ive y, AgrC may contam an afiix which agrees with the clau,al sub ' ct . . might wonder why the allix in AgrC must agree w'tl Il l' h . . . ~e . (VIZ. (2.5d)). ln !lus case Spec/AgrC is fi lied hy an expletive. One 1 1 Je su ~e~'1 w en Ils Spec 1s explet1v If · 1 · th~ only argument capable of movement lo Spe /AgrC 'tl . . • . ~- ~xp ellves need to be replaced (or adjoined to) in LF, . . c Wl Jout v10latmg (the nunnnahty 1 f) 1 E )) 'gl . cause 0 t Je CP IS tl1e subJect. Agreeing complementizers in West-Gennanic (Haegeman 0992 1 9 A pattem ofsystematic end isis is a Iso similar analysis (see (forthcoming). tl . g ages, another t:1 mlly mamfestmg VSO order. See Roberts ( 1993) and Roberts and Shlonsky (19 94 ) \ 1 · · v lere Je corre 1at10n between Semitic clitics and VSO-hood is further explor..:d. AJt ofpron~uns Spec/Agr is nol an S-stmcture Checking position in XSO configurations. '1 saw ali the children' The consequence to be drawn at this point i tl .. . . . .. the doubled DP is not •. s lat a prepositiOn IS reqUJrcd Ill clliJ C-<Ioubling s tructures only wben n an agreement configuration 'tl tl suggcsls that the consequence is more cner WJ 1 le c ll~c. at -structure. Consideration of (25d) way of checking its Case at S-stmctur: In (~Sd~a::,~?è~hatl ~~~~ prcp~~JtJon m~s.t occu~ if the doublcd DP has no raises to an AgrO c-commanding it TÎlc b' 1 1 w ,_•c 1 mam ests a clitJC, wluch J Lake to mean thal Thcre is no need to introducc a furtÎJ su ~ec • IO":e~er IS al,rcady in_ a checking configuration with AgrSO. ils fcatures in LF.8 er dummy preposJtJon. and <lU assowlled Ag rP in order to allow il to check 6. A fondamental property of VSO language is that A-specificrs arc never filled at S-structure. Take VSO arder to mean that the specifier of the verb's landing site (AgrS 0 ) is not a Checking position at S-S. Generalizing across categories, we arrive at (29) . ::Ja;:~~:~:;~~~~~h:~e:,:i:l~at:: - 10- Shlo~sk~ REFERENCES Belletti A. and U. Shlonsky (forthcoming) 'The Order of verbal Complements: A Comparative Study. NUT. Bennis H. (1986) Gaps and Dummies. Dordrecht: Foris. Bianchi V. and C. Figueiredo-Silva (1993) 'On Sorne Properties of Agreement Object in Italian and in Brazilian Portuguese'. Pisa: Manuscript, Scuola Normal Superiore, and Geneva: Université de Genève. Borer H. (1983) Parametric Syntax. Dordrecht: Foris. (1989) 'On the morphological Parallelism between Compounds and Constructs' in Booij et al (eds.) Morphology Yearbook 1. Dordrecht: Foris. (forthcoming) 'The Ups and Downs of Hebrew Verb Movement'. NLLT. Broselow, E. (1976) 'The Phonology ofEgyptian Arabie'. Ph.D. dissertation. UMass, Amherst. Cardinaletti, A. (forthcoming) "#" The Linguistic Review. Chomsky N. (1991) 'Sorne Notes on Economy of Derivation and Representation' in R. Friedin /ed.) Princip/es and Parameters in Comparative Grammar. Cambridge: The MIT Press. (1992) 'A Minimalist Program for Linguistic Theory' MIT Occasional Paper. Doron E. (1983) 'Verbless Predicates in Hebrew'. Ph.D. dissertation. Austin: University of Texas. Fassi-Fehri A. (1993) Issues in the Structure of Arabie Clauses and Words. Dordrecht: Kluwer. Friedemann M-A. and T. Siloni (1993) 'Agr0 Is Not AgrP' Geneva Generative Papers 1.1 Haegeman L. (1992) Ilwmy and Description in Generative Symax: A Case Study in West Flemish. Cambridge: Cambridge University Press. Hazout 1. (1991) 'Verbal Nouns: Theta Theoretic Studies in Hebrew and Arabie'. Ph.D. dis. UMass, Amherst. Hoyt K. (1989) 'Verb raising in Lebanese Arabie'. MIT Working Papers in Linguistics 11. Johnson K. (1991) 'Object Positions'. Natural Language and Linguistic Themy 9, 577-636. Kayne R. (1975) French Syntax. Cambridge: The MIT Press. (1989) 'Romance Clities, verb Movement and PRO'. Linguistic Inqui1y 22, 647-686. (1993) 'The Antisymmetry of Syntax'. Manuscript. New York: CUNY Graduate Center. Kenstowicz M. and W. Wahba (1980) Kenstowicz, M. and W. Wahba (1980) 'Ciitics and the Double Object Construction in Cairene Arabie', Studies in the Linguistic Sciences, 10.2. Urbana-Champagne: UIIIinois. Khalaili S. (1993) 'On the Relevance of the Split Complementizer Hypothesis to Case and Agreement in Arabie'. Manuscript. Leiden: Holland Institute of General Linguistics. Koopman H. (1992) 'The Internai and Extemal Distribution of Pronominal DPs'. Manuscript. UCLA. Larson R. (1988) 'On the double Object Construction' . Linguistic lnquiry 19, 335-391. Mouchaweh L. (1986) 'De la Syntaxe de Petites Propositions'. Thèse de doctorat. Paris: Université de Paris VII. Ouhalla J. (1988) 'The Syntax of Head-Movement: A Study of Berber'. Doctoral Dissertation. UCL (1994) 'Verb Movement and Word-Order in Arabie'. In Homstein Norbert and D. lightfoot, (eds.) Verb Movement. Cambridge: Cambridge University Press. Ritter E. (forthcoming) 'On the Syntactic Category of Pronouns and Agreement'. NUT. Roberts 1. (1993a) 'The Nature of Su~ject clitics in Franco-Provençal Valdotain'. In Belletti, A. (ed.) Dialects of ltaly. Turin: Rosenberg and Sellier. (1993b) 'Agreement, Clitics and Case in Welsh and Elsewhere'. ms. U North Wales. & U. Shlonsky (1994) 'Pronominal Enclisis in VSO Languages'. ms, U Northe Wales & UGeneva. Shlonsky U. (1991) 'Quanti fiers as Functional Heads: A Study of Quantifier Float in Hebrew'. Lingua 84. (forthcoming) 'Agreement in Camp'. The Linguistic Review. Siloni, T. (1994) 'Noun Phrases and Nominalizations'. Ph:d Dissertation. Geneva: Université de Genève. Speas, M. (1991) 'Funetional Heads and Inflectional Morphemes'. The Linguistic Review 8, 389-417. Sportiche, D. (1989) 'Movement, Agreement and Case'. Manuscript. Los Angeles: UCLA (1992) 'Ciitic Constructions'. Manuseript'. Los Angeles: UCLA - 11 -