THE DOUBLE SUBJECT CONSTRUCTION IN ROMANIAN. NOTES ON THE SYNTAX OF THE SUBJECT ALEXANDRA CORNILESCU 1. Preliminaries. Aim and outline of the paper In the more recent pre-minimalist descriptions of Romanian we are acquainted with, one finds the following presentations regarding the Romanian subject: a) The subject is generated VP-internally and may remain there, since the Nom(inative) Case may be assigned under government by the verb raising to Inflection. Alternatively, the subject, which has been assigned case in Spec VP, moves to SpecF, which is a Topic position in Romanian, rather than an argumental, or L-related position. This is the analysis presented in Dobrovie-Sorin (1987, 1994). b) A different view is defended in Motapanyane (1989). Working with a more richly articulated functional structure of the Romanian clause, she considers Romanian clauses to be AgrSPs and claims that there are two argumental subject positions Spec VP and SpecAgrS. Case is assigned under government by the verb in Tense or in SpecAgrSP by Spec-Head Agreement. Evidence is provided that SpecAgr S cannot be an A' position and must be an A position. Therefore, both authors concur that Case is assigned postverbally under government by the verb that has raised to 1° or T\ but they differ in the interpretation of the preverbal subject position, considered either a position of topicalization or an argumental Case-related position. Several important studies on Romance (Roberts (1991), Cardinaletti and Roberts (1991)) have argued for the existence of two AgrSPs, both of them preverbal, at various stages in the evolution of various Romance languages (e.g. Old French). Recently Cardinaletti (1996) argues that in UG there may be two AgrSPs, both preverbal in Italian. The two positions are specialized. The lower Agr2SP is Case-related and it is also the position where ij>-features are checked* securing Subject-Verb agreement. In pro-drop languages like Italian, also;the position which licenses pro. >. The higher preverbal subject position AgrjSP is also argumental and, in Italian or French, it accomodates strong pronouns and strong DPs, while the ..weak Nom pronouns remain in the lower Agr2SP. It is also the position which •. -RRL,. XLII, 3 -4, p. 101-147, Bucarest, 1997 102 Alexandra Cornilescu 2 hosts prominent nominate like the Dative Experiencer of Psych verbs. In other words, Cardinaletti concludes, this position is that of the semantic “subject of experience” or “subject of predication”. Given such descriptions for Romance, it is desirable to try to establish the position and properties of the AgrS projection(s) in Romanian, as well as the relative position of this projection with respect to the verb. This amounts to attempting to define the syntactic properties of the subject, as it occurs preverbally or postverbally. This is the problem dealt with in this paper. Evidence will be presented supporting the following claims: a) There is only one AgrSP in Romanian, which is nevertheless a postverbal position. Consequently, both Spec VP and SpecAgrSP are postverbal positions in Romanian. b) With very limited exceptions, Nom Case is regularly checked in SpecAgrSP, which is the Nom Case position of Romanian. c) The preverbal subject position can roughly be described as a topic, rather than an argumental or L-related, position. Our analysis confirms the findings of Dobrovie Sorin (op.cit.), with different arguments. More specifically, the preverbal subject is left dislocated or focalized. d) This description of the Romanian subject is supported by various morphologic and syntactic facts. In section 2, we briefly inspect the morphology of the Romanian verb, outlining the functional structure of the Romanian finite clause. Romanian sentences will be described as Mood Phrases and it will be seen that, apparently, in Romanian, the Verb raises to M°, though, arguably, not higher to C°. In section 3, we indicate that there are two postverbal subject positions: Spec VP, a thematic position, and AgrS, a postverbal position to the left of Spec VP. In section 4, we describe a double subject construction in Romanian, which can be taken to provide crucial evidence that AgrSP is postverbal in Romanian and that SpecAgrSP is the major Nom checking position of Romanian. In section 5, we prove that the preverbal subject occupies an A' position (topic or focus), so that there is only one AgrS position in Romanian. 2. On Inflection and Verb Movement in Romanian 2.1. Functional Categories of the Romanian Verb This investigation cannot proceed without first defining the skeleton of the Romanian clause, in as much as it is relevant for the syntax of the subject. Like other researchers on Romance and on Balkan languages (Rivero (1994), Alexiadou (1994), a.o.), we accept the Split Inflection hypothesis (Pollock, 1988, Chomsky, 1991), postulating at least the verbal categories of 3 The Double Subject Construction in Romanian 103 Mood, Tense, Aspect for the Romanian clause. Syntactic evidence in favour of such a position for Romanian is to be found in Rivero (1994), and comes especially from her important study of Verb Movement in the Balkan languages. While Rivero (1994) chiefly relies on syntactic facts for assuming the existence of a M(ood)P and T(ense)P, in fact, a cursory glance at the morphology of the Romanian verb may also provides evidence for the categories of Mood, Tense and Aspect, even if one ignores the agreement features of person and number. 2.1.1. It is tempting to initially consider only finite and non-finite inflectional forms ignoring the analytical tenses and moods. In Romanian, for the verb merge (go), these are as shown in-(l): (1) Indicative The Present MERG (1st person singular) The Imperfect (Past) MERGEAM (1st person singular) The Simple Past MERSEI (1st person singular) The Past Perfect MERSESEM (1st person singular) S ubj unctive The Present S A MERG Infinitive A MERGE Gerunziu MERGAND Past Participle MERS Like other Balkan Languages, Romanian disposes of Mood markers, namely the particles sd for the Subjunctive and a for the Infinitive. The examination of the forms listed above evinces the existence of two different stems in the paradigm of this verb: MERG and MERS. Since this variation is systematic for (regular) verbs of the 3rd conjugation and since it does not seem to be a case of phonological conditioning (for instance, both MERG/MERS occur in front of / E / (el merge (He goes), el merse (He went)), it is natural to assume that this opposition is the overt manifestation of one of the functional categories in the lower part of the functional system of the clause, presumably a manifestation of Aspect. We will say that the Aspect system distinguishes a non-perfect verbal stem: MERG from the verb merge (go), which appears in the Present, Imperfect, Infinitive and Gerunziu and a perfect verb stem: MERS, from the same verb merge (go), appearing in the Simple Perfect, Past Perfect and Past (Perfect) Participle. Just as in other Romance languages like French, one might say that this opposition holds between the [+ continuous] imperfective forms, and the [- continuous] forms. An intuitively satisfactory result is that the Present, the Imperfect, the Gerunziu, and the Infinitive fall together as continuous (exhibiting the stem MERG), while the other stem (MERS) appears with obviously perfect forms: (2) tu MERGi (Present) tu MERSe§i (Simple Perfect) tu MERGeai (Imperfect) tu MERSese§i (Past Perfect) a MERGe (Infinitive) MERS (Past Participle) MERGand (Gerunziu) 104 Alexandra Cornilescu 4 The Tense Projection will accomodate the features [+/- Present, +/- Past (Anteriority)]; specifically: Tense —> +/- Present and - Present —> +/- Past (Anteriority). The Present and the Imperfect wil differ in terms of +/- Present, while the Simple Perfect and the Past Perfect will differ in terms of +/- Past (Anteriority), an opposition morphologically marked by SE; El veni (He came) vs. El veniSE (He had come). Thus, even if the overt realization of Tense and Agreement is normally synchretic, the Tense projection is independently filled by SE. Ignoring subject agreement features, one gets the following characterization of the four inflectional forms of the Indicative: (3) Present MERG + indicative + present - perfect Simple Past MERSEI + indicative - present - past(-anterior) + perfect Imperfect MERGEAM - indicative - present - perfect Past Perfect MERSESEM - indicative - present + past (+anterior) + perfect Notice that the Imperfect suitably contrasts both with the Present, through the feature [+/- Present] and with the Past Perfect, through the feature [+/- Past (Anteriority)]. The agreement features of Person and Number will differentiate between any of these forms and the non-finite forms. The order of these functional projections, indicated in (4), is assumed to be the same crosslinguistically, pending evidence to the contrary, though in some languages some of these functional categories may not be projected at all (cf. Alexiadou (1994), Cinque (1993), Giusti (1992), Cardinaletti (1996) a.o. and, for a different view, see Ouhalla (1992)): (4) MoodP > TenseP > AspP That this analysis might be on the right track is also suggested by the fact that Baker’s Mirror Principle would be observed, if a derivational, rather than a checking analysis of forms like mersesem (1st pers, Past Perfect) were adopted: (5) [[[[ merv°] + s(e) Asp °] + seT°] mAgrS°] 2.1.2. Moreover, verb forms should always be taken as positively specified for a particular Mood, even when there is no mood particle. For instance, the Subjunctive (present) is formed with the Mood particle sa followed by an inflected form, which is different from that of the indicative, at least in the third person, so that the Indicative/ Subjunctive opposition is overtly marked even when the Mood Marker sa is absent. The Mood feature is overfly checked in Romanian, since the verb raises to M°. 5 The Double Subject Construction in Romanian 105 (6) Ion cite§tE cartea. [Indicative] Ion reads the book. Ion SA citeascA cartea. [Subjunctive] Ion should read the book. (7) a. Pregedintele traieqte. [Indicative] The president lives. b. Trdiascdpre§edintele! [Subjunctive] (Long) live the president. (8) traiegte traiasca +indicative +subjunctive +present +present -perfect -perfect + 3rd person 3rd person Rivero (1994) proves that in the Subjunctive, the Mood head must always be lexically filled, either by the particle sa, or by the inflected verb which skips the clitic phrase and filles the empty M° head by Long Head Movement (LHM). (9) a. Cre$tinii sa-l slaveasca pe Dumnezeu. The Christians SA him (cl.) glorify (subj) God. “Let the christians glorify God!” b.Cre§tinii sldveasca-lpe Dumnezeu! The Christians glorify (subj) him (cl.) God! The existence of Long Head Movement suggests that Romanian verbs must overtly raise to MP to check their mood feature. We will assume (and then offer further evidence) that verbs always raise to M° in overt syntax. We also provisionally accept that Case is checked in AgrPs, rather than in the Spec of verbal inflectional categories (TP or AspPs) (see Alexiadou (1994), Cardinaletti (1996)). Since Romanian is a pro-drop language, the inflectional system of the finite verb presents rich person and number marking, therefore, clear evidence for an AgrSP. At the same time, the morphology of the Romanian Subjunctive, in examples like (5, 6, 9) above, clearly indicates the position of the AgrSP, below the MP, if Baker’s Mirror Principle is observed: \ (10) MP > AgrS > TP > AspP. Summing up what we have said so far, we conclude that the morphology of the Romanian verb clearly indicates that the AgrS projections is lower than the MP, although it does precede the TP and the AspP. Since the verb raises to MP, AgrSP is postverbal. 106 Alexandra Cornilescu 6 2.1.3. The Cl(itic) Phrase is presumably between MP and AgrS, as suggested by forms like (9), analysed in (11): (11) [sa M°] + [L c,°] + [slaveasca AgrS0]. The clitics are case features of the verb (Borer (1984)). The raising verb will take them along as it moves through the clitic phrase to check its case features. The clitics (pronominal or adverbial) and the verb form a cluster which behaves like one syntactic unit (see Dobrovie-Sorin 1987, 1994). If the M° position, normally filled by mood particles, is empty and the verb skips the C1P, undergoing Long Head Movement to raise to M°, the verb in M° no longer properly governs its trace in AgrS0; the clitic is forced to adjoin to the verb in M°, forming a unit which now properly governs / licenses the trace of the verb in AgrS0 (12c) (12) a. Ion sa-l slaveasca. Ion SA him(cl) glorify (Subjunctive). b. Ion 11 slave§te. Ion him (cl.) glorifies. (Indicative, Present) c. Ion slaveasca-l. Ion glorify (Subj) him (cl) ‘"Let Ion glorify him!” At the same time, as shown by examples (13, 14), the distribution of the Subjunctive Present, a complex form (mood particle + inflected form), is the same as that of the Indicative Present, a simple form, as far as the subject and the adjuncts are concerned. Adverbs and the subject either precede or follow the complex subjunctive form. This indicates that both the complex form, which finally forms a complex X° constituent, and the simple (simpler) verb form occupy the same position, M°: (13) a. Ion maine sa mearga acolo. Ion tomorrow SA go (subj) there. b. Ion maine merge acolo. Ion tomorrow goes there. (14) a. Maine sa mearga Ion acolo. Tomorrow S A go (subj) Ion there. b. Maine merge Ion acolo. Tomorrow goes Ion there. All these very general considerations support the view that the Verb always overtly raises to the MP. which is always projected. 7 The Double Subject Construction in Romanian 107 2.2. V-Raising to Mood. Evidence from Adverb Positions Further evidence regarding V-Movement in Romanian comes from a cross-linguistic examination of the position of adverbs. Since Pollock (1988), their position has been considered important in establishing how far various heads raise in different languages. The general idea is that adverbs need to check their inherent features either in the Spec position of an appropriate head or in a position of adjunction to a semantically suitable phrase (cf. Cinque (1993b)). In important work on adverbs, Alexiadou (1994) opts for the first solution and states the general principle that (at least some) adverbs are generated in specifier (rather than adjunct) positions, obeying the following licensing condition: (15) “Generalized Licensing Criterion (i) An [+F] head must be in a Spec-Head relation with a [+F] XP. (ii) An [+F] XP must be in a Spec-Head relation with a [+F] head. If [+F] corresponds to -features (gender, number, person), then the position is an A position, when [+F] corresponds to operator features, wh/neg/topic, the specifier is an A' position”. I wish to remain neutral on whether adverbs are specifiers or are adjoined, although provisionally adopting the first position (adverbs as specifiers). In either case, however, adverbs may be used to ascertain the positions to which the verb raises, on the assumption that adverb positions are fixed since adverbs must agree with heads/projections which are generated in a rigid word-order themselves. We will first examine aspectual and time adverbs. 2.2.1. Aspectual adverbs (regulat (regularly), zilnic (daily), mereu (always), iar (again), a.o.) are generated in the Spec of an AspP, or at least may raise to their licensing position overtly. Time adverbs (acum (now), atunci (then), cur and (soon)) are generated lower in the VP(16b), but may check their feature in SpecTP. The examples below suggest that the preferred order is indeed TP > AspP: (16) a. Ion vine acum zilnic la mine. John comes now daily to me. b. Ion vine acum la mine zilnic. John comes now to me daily. c. ? Ion vine zilnic acum la mine. Ion comes daily now to me. Notice now the examples below where the Time and Aspect adverb precede subcategorized Prepositional Objects, which are surely in the VP and have no reason to raise. This word order pattern is expected if the Time / Aspect adverbs are in the Specs of the respective functional projections and the verb has raised past them to MP, as hypothesized above. 108 Alexandra Cornilescu (17) a. Ion apeleaza acum zilnic la parintii lui. Ion resorts now daily to his parents, b. Ion se gande§te acum rareori la parintii lui. Ion thinks now seldom of his parents. Negative adverbs (e.g. deloc (at all)) will be generated in the Spec of the NegP, directly below MP, while Manner adverbs are apparently licensed in the Spec of a Voice Phrase, lower than the AspP (cf. Alexiadou (1994)). Provisionally accepting a biclausal analysis of compound tenses (cf. Dobrovie Sorin (1987, 1994), Alexiadou (1994)), it is easy to notice the following striking contrast between French on the one hand, and Romanian and Italian on the other, with respect to the position of the Past Participle. (18) French Les enfants n ’ont [pas du tout bien ripondu a la mattresse]. Italian Bambini non hanno [detto mica piu bene alia maestra]. Romanian Copiii n-au [raspuns deloc bine profesoarei]. “Children have not responded at all well to the teacher”. The contrast shows that in Italian and Romanian the Past Participle in its own clause raises to MP, past the adverb of manner and the negative adverb, preceding them in the overt structure, while in French the verb apparently raises only to the VoiceP (cf. Alexiadou), being preceded by the two adverbs. It may be concluded that the verb always overtly raises to the highest inflectional projection in the area of the morpho-syntactic features, the position M°. 2.2.2. At the same time, the more articulate clause structure adopted here for Romanian allows the observation that in Romanian, although the verb always raises to MP, whether the MP is or is not independently filled, it never continues to C°. There is no 1° -to C°. A similar remark was made for Spanish in Suner (1994). The following may be regarded as arguments against raising to C°: a) In Romanian there are no main clause / subordinate clause asymmetries (cf. also Isac (1996)). b) There is a class of preverbal adverbs, presumably in SpecMP or higher, which always precede the finite verb. (19) a. Ion abia il a§teapta pe Petru. Ion hardly him (cl) awaits Petru. “Ion can hardly wait for Petru.” (20) b. *Ion tl a§teaptd abia pe Petru. Ion him(cl) waits hardly for Petru. “Ion can hardly wait for Petru.” 9 The Double Subject Construction in Romanian 109 What matters is the position the V occupies when a Wh phrase undergoes WH-Movement, accepting that the position to which the wh-phrase moves is SpecCP. The examples show that even in the interrogative sentence, the respective adverbs still precede the verb. This proves that the verb has not changed its position. (21) a. Pe cine abia a§teapta Ion? Whom hardly awaits Ion? “Who can Ion hardly wait for?” b. *Pe cine a§teaptd abia Ion? Whom awaits hardly Ion? “Who can Ion hardly wait for?” Of course, if the verb does not raise to C° past the preverbal subject, one will have to find a different account for the often noticed impossibility of a preverbal subject in questions (cf. Baciu (1996): *Ce Ion cite§te? (What Ion reads?)) In conclusion, the verb always raises to M°, but not further, so that the Romanian finite / non finite sentence is uniformly (at least) an MP. This is a satisfactory descriptive result. 3„ Postverbal Subject Positions In this section I attempt to establish the claim that there i&a postverbal AgrS position in Romanian, using syntactic evidence. As already seen, verb morphology gives a hint that AgrSP is lower than MP in Romanian, but Baker’s Mirror Principle, on which our hypothesis has so far rested, is too strong and has been shown not to work in many instances. Therefore, distributional evidence is necessary to strengthen the claim. The examination of postverbal subject sentences will prove that there are two postverbal subject positions: Spec VP and a second argumental position to the left of SpecVP, but still postverbal, a position which has the properties of SpecAgrSP. 3.1. The Postverbal Subject in SpecVP It has often been convincingly shown that the subject may remain in SpecVP in Romanian and, moreover, that it may be case-marked under government in this position by the verb raising to Inflection (cf. Dobrovie-Sorin (1987, 1994)). In the minimalist framework, maintaining a* difference between the thematic positions in the Spec of lexical categories and Case positions in the Spec of functional categories, SpecVP will not be a Case position. Alternative analyses are available. The simplest is that the subject in SpecVP may check its Case raising to a subject Case position at LF. Since movement 110 Alexandra Cornilescu 10 operations, even at LF, are costly and must be triggered, a different analysis is also available. Given that it is a pro-drop language, Romanian has the inexpensive possibility of expletive insertion. The subject might check its Case through coindexation with pro, an expletive licensed by the AgrS" head. When the subject stays in SpecVP, the surface order is V(X)SO, with the verb raising to the higher functional projections past the subject in SpecVP. The subject may precede subcategorized constituents, the direct object or a prepositional object: (22) Numai rareori $tiau top elevii toatfi lecpa. Only seldom knew all the pupils the whole lesson. (23) tntotdeauna apelau mulp dintre ei la bunicii lor dupa ajutor. Always resorted many of them to their grandparents for help. There is no strict adjacency requirement between the verb and the subject, so that adverbs can intervene between the verb and the subject. (24) Vara dormeau deseori top copiii pe terasa. In summer, slept often all the children on the terrace. Romanian is, in this respect, different from Celtic languages. Significantly for our discussion, Roberts (1994) (apud Alexiadou (1994)) concludes that in Welsh the postverbal Su is located in SpecAgrSP, because nothing can intervene between the verb, presumably in an intermediate projection, between C° and AgrS0, and the subject, so that all adverbials appear in final position. 3.2. The Postverbal Subject in the Postverbal SpecAgrSP Position Apparently, in Romanian there is a second higher postverbal subject position than SpecVP. Thus, Time / Aspect adverbials may intervene between the postverbal subject and the direct object (examples (25)) or prepositional object (example (26)). which are clearly in the VP (or even higher). (25) a. Vara, faceau top copiii zilnic probleme la matematica. In summer, solved all the children daily maths problems. b. Aproape farci gre§ invatau copiii pe atunci cdte o limb a straina. Almost without exception learned all the children at that time a foreign language. c. Din fericire, primeau top copiii in vremea aceea regulat cadouri frumoase de Craciun. Fortunately, received all the children at the time regularly nice presents for Christmas. 11 The Double Subject Construction in Romanian 111 (26) In tinerete apelau top nepopi cu regularitate la unchiul lor bogat dupa ajutor. In their youth resorted all the nephews regularly to their rich uncle for help. There are good reasons to believe that this postverbal subject position in Romanian is SpecAgrSP. A comparison of Romanian and Italian is helpful, because certain properties characteristic only of preverbal subjects in SpecAgrSP in Italian appear to be true of postverbal subjects as well in Romanian. One such property is Quantifier Floating (QF). Cardinaletti (1996) establishes that in Italian, QF is compatible only with preverbal subjects, as in (27b): the preverbal subject in SpecAgrSP licenses the quantifier left behind in SpecVP. An inverted subject in Italian is not compatible with FQ (cf. (28b)). In Romanian, not only preverbal, but also postverbal subjects and FQ may cooccur. The Subject is postverbal yet, like in Italian it has to be to the left of the Quantifier, in a c-commanding position, which can be assumed to be AgrSP, as in (30c), (but not in (30b)). (27) a. Tutti i soldati sono andati via. “All the soldiers have gone away.” b. / soldati sono tutti andati via. (28) a. Sono andati via tutti i soldati. b. *Sono tutti andati via i soldati. (29) a. Top soldatii au plecat de aici. b. Soldatii au plecat top de aici. (30) a. Au plecat de aici toti soldatii. b. *Au plecat toti de aici soldatii. c. Au plecat soldapi de aici toti. Here are more examples, where the postverbal subject is separated from the floated quantifier by Time/ Aspect / Manner adverbials. Thus, in (3 lb), the Aspect adjunct tn fiecare vara (every summer) intervenes between the subject and the quantifier, in (32b), the verb is unergative (a rade (laugh)), the postverbal subject precedes the Time adverb atunci (then), in SpecTP, while the floated quantifier amandoi (both) precedes a subcategorized PP {de cele intamplate (at what had happened)); in (33b), the verb is transitive, the postverbal subject, ei (they) is followed by the Aspect Adverb niciodata (never), preceding the floated Q (top (all)) and the direct object. 112 Alexandra Cornilescu 12 (31) a. In vremea aceea plecau in flecare vard amdndoi pdrinpi lui in strdinatate. At that time, left every summer both of his parents abroad. “At that time both of his parents went abroad.” b. In vremea aceea plecau pdrinpi lui in flecare vard amdndoi in strdindtate. At that time left his parents every summer both abroad. (32) a. Grozav mai rdsesera atunci ei amdndoi de cele intdmplate. Terribly had laughed then they both at what had happened. “Both of them had terribly laughed at what had happened.” b. Grozav mai raseserd ei atunci amdndoi de cele intdmplate. Terribly had laughed they then both at what had happened. (33) a. N-au uitat niciodatd ei toti toatd lecpa. Haven’t forgotten never they all the whole lesson. “They never have all forgotten the whole lesson.” b. N-au uitat ei niciodatd top toatd lecpa. “Haven’t forgotten they never all the whole lesson.” We have assumed that the verb raises to MP in Romanian. Since the subject appears in postverbal position, yet it precedes the adverbs in Spec TP and Spec AspP and the floated Q, it can only be in some AgrP, say AgrSP. This distributional fact agrees with the morphology of the Romanian verb, which indicated the same functional order MP > Agr SP > TP > AspP. In section 4 below, we shall prove that the postverbal SpecAgrSP position is a / the Nominative Case position in Romanian. 3.3. Interpreting the VOS Order in Romanian Before turning to that problem, we should mention what might look like another postverbal subject position in Romanian. In the sentences considered so far, the subject had consistently preceded the object(s), in VXSO or VSXO word orders. However, the VOS structure is also available in Romanian. (34) a. Au caqtigat premii prestigioase urmdtorii studenp: Have won prestigious prizes the following students. “The following students have won prestigious prizes.” b. §i-au cumpdrat vile luxoase top noii demnitari. For themselves (cl) have bought luxurious villas all the new officials. “All the new officials have bought themselves luxurious villas.” (35) a. A spart u$a tdncii un hot foarte priceput. Has broken down the door of the bank a very skilled burglar. “A very skilled burglar has broken down the door of the bank.” 13 The Double Subject Construction in Romanian 113 b. A telefonat Mariei ION, nu Petre. Has telephoned to Maria Ion, not Petre. “ION, not Petre has telephoned to Maria.” The examples look like the Inversion structure of Italian or Spanish. The traditional analysis of VOS in these languages involves right adjunction of the Subject to the VP or base generation in this position (Rizzi (1982), Burzio (1986)). As first remarked by Dobrovie Sorin (1994) and Motapanyane (1989), the postverbal subject is not VP adjoined in these Romanian structures. In fact, there is evidence that even suggests that the subject cannot be adjoined to the VP. This hypothesis would predict that the subject can appear after an object clause. But this position is awakward or even completely unacceptable in Romanian: (36) a. ? * I-a spus lui Ion cd vremea va fi frumoasd Petru. To him(cl) has said to Ion that the weather will be fine Petru. “Petru has told Ion that the weather will be fine.” b. Petru i-a spus lui Ion cd vremea va fl frumoasd. Petru to him(cl) has said to Ion that the weather will be fine. If one considers the interpretation of the VOS sentences above, it becomes clear that in this order the V and the O are part of the presupposition of the sentence, while the subject is focalized, it introduces new information, and may even be a contrastive focus (as in 35b). A more plausible analysis of the VOS order in Romanian would involve Object Movement (scrambling) of the object past the subject in Spec VP. The object raises to Spec AgrO. As known, scrambling may create binding relations. In support of such an analysis, notice that the subject can, indeed, be bound by an object quantifier only in the (V)OS order (sentences (37a), (38a), (39a) below), though not in the SO order (sentences (37b), (38b), (39b) below). Compare: (37) a. Ce i-a cumpdrat fiecdrui copil tatal lui de Craciun? What to him(cl) has bought to each child his father for Christmas. “What did his father buy for every child for Christmas?” b. *Tatal lui i-a cumpdrat fiecdrui copil bomboane de Crdciun. His father to him(cl) has bought to each child sweets for Christmas. “His father has bought sweets for every child for Christmas.” (38) a. Unde l-a chemat pefiecare copilprofesorul lui? Where him(cl) has called each child his teacher. “Where did his teacher call every child?” b. *Profesorul lui (l)-a chematpe fiecare elev la cancelarie. His teacher (him)(cl) has called every child to the staff room. “His teacher called every child to the staff room.” 114 Alexandra Cornilescu 14 (39) a. Ce-a tnvatat-o pe flecare fetita mama ei? What has taught her(cl) each little girl (Acc) her mother? “What has her mother taught each little girl?” b. *Mama ei a mvcitat-o bunele maniere pe flecare fetita. Her mother has taught good manners each little girl. Since in the scrambled VOS order the subject is the most deeply embedded constituent of the whole IP, it will be maximally stressed, focalized, as desired (cf. Cinque (1993)). In Romanian, Object Movement may affect any DP, not only strong or specific DPs as in Dutch or German (see examples in (34) above). The Romanian counterpart of Object Movement in Germanic is Clitic Doubling. Object Movement and Clitic Doubling are independent structures. Apparently, movement of the object is triggered in order for the subject to get maximal stress in the sentence, i.e., to become focal (a point also defended in Zubizaretta (1993)). In such cases, movement continues to be driven by the need to check certain features of constituents, but, against the requirements of Greed, a constituent a (the direct object) moves only to allow the satisfaction of some property of another constituent b (in this case, the [+Focus] feature of the subject). A principle along the following lines seems to be in operation (cf. Alexadiou (1994)): (40) Move a although a has weak features to satisfy a property of P. Otherwise, a violation at PF will occur. In the same way, Lasnik (1995) had discussed cases where movement though still driven morphologically, benefited the target of movement, rather than the mover itself’s (his “Principle of Enlightend Self-Interest”). To the extent that these analyses are well-founded, they suggest that a weakening of Greed is called for. It may be concluded then that in the SOV order, the subject continues to be in SpecVP, so that there are only two postverbal subject positions: SpecVP and Spec AgrSP. In the next section, we prove that AgrSP is a / the Nom Case position, as well as the position which licenses pro in null-subject sentences in Romanian. 4. The Double Subject Construction Important evidence regarding the assignment of Nom Case in Romanian comes from “double subject constructions” (DSC from now on). 4.1. The Data. The Traditional View Romanian is not usually described as a clitic subject language. However, there is a particular double subject construction, with a pronoun doubling a lexical DP, where it can be argued that the pronominal subject is in fact a weak 15 The Double Subject Construction in Romanian 115 pronoun and a phonologic clitic. We will refer to it as the clitic subject. Romanian then becomes comparable with other clitic subject languages, like certain dialects of Italian or French. The interest of this construction for our topic is that, if, as generally agreed in the literature (cf. Cardinaletti and Starke (1994)), weak pronouns and clitics do occupy their case-checking position before Spell-Out, since the pronouns in this construction are Nom weak pronouns, their position simultaneously indicates the Agr S and the Nom Case position(s) in the Romanian clause. Romanian contrasts sharply with all of the northern Italian dialects or with French, in which the clitic is preverbal, since in Romanian the clitic is always postverbal. 4.1.1. The double subject construction that we are investigating belongs to the “langage familier et populaire” (Byck (1937)), marking a certain illocutionary attitude of the speaker: threat, promise, reassurance. Linguists have spoken of an “emploi affectif du pronom personnel en roumain” (Byck, op.cit.). The pronominal clitic subject must immediately follow the verb. The lexical subject is usually postverbal (examples (41)), but it may as well be preverbal (42). Consider the following attested literary examples: (41) a. Vazand el dracul cd n-are ce face... Seeing he the devil that he has not what to do... “The Devil seeing that he doesn’t have what to do...” b. Dar o sa vie ea o vreme cand o sa se gaseasca cineva sa scrie despre vitejiile romdnilor. But will come it a time when will be found someone to write about the heroic deeds of the Romanians. “But a time will come when there is going to be someone who would write about the heroic deeds of the Romanians.” c. S-a trece ea §i asta. Will pass it this too. “This too will pass.” d. Ne cam codeam noi fetele. Rather hesitated we the girls. “We the girls rather hesitated.” e. S-a gdsi el ac §i de cojocul lui. Itself(cl) will find it a needle also for his sheepskin. “A remedy will be found against him as well.” (42) a. Mama $tie ea ce face. Mother knows she what(she) is doing. “Mother knows what she is doing.” 116 Alexandra Cornilescu 16 b. Ion vine el mai tdrziu. Ion comes he later. “Ion will come later.” An interesting fact is the existence of two “dialects”. In one of them, that of standard familiar Romanian, illustrated in all the examples above, the pronoun agrees in gender, person and number with the lexical subject. In the second, non-standard, variety, the pronominal subject is invariably the 3-d person singular masculine form EL, irrespective of the gender and number of the lexical subject. The verb naturally agrees with the clitic, not with the lexical subject. Here are attested literary examples: (43) a. Are sa-l certe elpreoteasa pe popa. Is going to him(cl.Acc) snub HE(Nom. Masc. Sg.) the priest’s wife (Nom. fem.) the priest. (Acc) “The priest’s wife is surely going to snub him.” b. Ne-a venit el apa la moara. To us has come HE(Nom. Masc. Sg.) the water (Nom. Fem. Sg.) to our mill. “The water has come to our mill.” c. Las ’ c-a pdtit-o el hotii. Be sure that has(sg) got it HE(Nom. Masc. Sg.) the thieves (Nom. Masc. PI.) “You may be sure that the thieves will get it!” Throughout the decades, the construction has caught the attention of reputed Romanian linguists, who have been intrigued by the possibility of doubling the subject, an uncommon possibility for Romanian, a language which may or must double the postverbal direct (44) and indirect Object (45), but not the postverbal or preverbal subject (46). Compare: (44) a. L-am vazut pe copil. Him (cl) (I) have seen the child. “I have seen the child.” b. Am vazut copilul. (I) have seen the child. (45) (I-) am dat copilului bomboanele. (To him) (cl) (I) have given to the child the sweets. “I have given the child sweets.” (46) a. In curte se jucau copiii Ini Petre. — ■ In the yard were playing the children of Petre. “Petre’s children were playing in the yard.” 17 The Double Subject Construction in Romanian 117 b. Copiii lui Petre se jucau in curte. The children of Petre were playing in the yard. Petre’s children were playing in the yard. Moreover, even in left dislocated structures, where doubling of either (preverbal) object is obligatory, the subject is not doubled; see, for instance, sentence (49) where the subject precedes other left dislocated constituents without being doubled: (47) a. Pe Ion l-am vazut in curtea §colii. PE lon(Acc) *(him)(I) have seen in the schoolyard. “I have seen Ion in the schoolyard.” b. Copilul l-am vazut in curtea §colii. The child *(him)(I) have seen in the schoolyard. “I have seen the child in the schoolyard.” (48) Copilului i-am dat homhoanele. To the child (Dat) *(to him)(cl) (I) have given the sweets. “I have given sweets to the child.” (49) Ion de la Paris numai carti a cumpdrat. Ion from Paris only books has bought. “Ion has bought only books from Paris.” Several important properties of the DSC have been noticed by Romanian linguists. Philippide (1929, apud Byck) considers that the pronoun is pleonastic, and that the pronoun and the lexical subject make up one constituent, the lexical subject being an apposition. He also correctly points out that in this construction the pronoun is always postverbal: “On emploie parfois le pronom personnel a cote du substantif, en apposition. Pour le sujet, cet emploi pleonastique existe seulement la ou le verbe precede”. Against Philippide (1929), Byck (1937) claims that, although initially the noun may have been an apposition to the pronoun, at the current stage the pronoun forms one constituent with the verb, not the noun. He emphasizes that the pronoun is unstressed and forms a phonetic unit with the preceding verb, not with the following noun. Moreover, the pronoun always agrees with the verb, and only sometimes with the noun (see dialectal examples in (43)). It may certainly be added that the lexical subject may be separated from the pronoun, when it is preverbal, or when there is an adverb between the pronoun and the lexical subject; therefore they cannot be said to form one constituent. (50) a. Fratele meu vine el mai tarziu. My brother comes he later. “My brother will come later.” 118 Alexandra Cornilescu 18 b. Las ’ci5 vin ei mdine copiii, n-ai grijd. Be sure that come they tomorrow the children, don’t worry. “The children will come tomorrow, don’t worry.” Most linguists agree that, originating in an appositional structure, the repetition of the subject has become a means of rendering the speaker’s affectivity and emotional intensity. Traditional grammarians have thus established that the pronoun forms a phonologic and perhaps also syntactic unit with the verb, and that the pronoun is an expletive. 4.1.2. Before turning to other properties of the DSC, it is important that it should not be mixed up with other situations where there is a pronominal copy of the subject, accompanied by the presuppositional adverb p (also) as in $i el (he too) or by the emphatic reflexive pronoun (el insu$i). There are important distributional differences between these structures: 1. In the DSC, the clitic pronoun must be adjacent to the verb (cf. (51a), (52a), vs. (5 lb), (52b), while the other two combinations may be separated from the verb by an intervening constituent, such as the adverb mdine (tomorrow) (cf. (53b), (54b). In the DSC, the adverb may not appear between the clitic and the verb. (51) a. Vine el tata maine. Comes he father tomorrow. “Father will come tomorrow.” b. *Vine mdine el tata. Comes tomorrow he father. c. Vine el mdine tata. Comes he tomorrow father. (52) a. Tata vine el mdine. Father comes he tomorrow, b. *Tata vine mdine el. Father comes tomorrow he. (53) a. Tata vine .p el mdine. Father comes he too tomorrow. “Father too will come tomorrow.” b. Tata vine mdine $i el. Father comes tomorrow he too. (54) a. Tata vine el insu§i mdine. Father comes he himself tomorrow. “Father himself will come tomorrow.” 19 The Double Subject Construction in Romanian 119 b. Tata vine maine el insup. Father comes tomorrow he himself. 2. In the second place, in the DSC, the subject is either postverbal or preverbal. In contrast, the other two combinations appear only with a preverbal subject, a postverbal subject causing ungrammaticality. There is a clear difference between example (55a) and examples (55b, c) below: (55) a. Vine el tata mai tarziu. Comes he father later. b. *Vine $i el tata mai tarziu. Comes he too father later. c. Wine el insu.p tata mai tarziu. Comes he himself father later. 3. The other two combinations may appear with any type of lexical subject; they are not sensitive to quantifiers, while the DSC construction excludes bare quantifier subjects, whether they are preverbal or postverbal: (56) a. Vine el Ion / unchiul Ion/ stapanul. Comes he Ion/ uncle Ion/ the master, b. *Vine el un copil. Comes he a child. Wine el fie care copil. Comes he every child. (57) a. Fiecare copil vine §i el cu speranfa. Each child comes he too with hope, b. Un copil vine §i el cu speranta cd va fi ajutat. A child comes he too with the hope that he will be helped. (58) a. Nici un copil nu vine el tnsu§i la tine sa-ti cearti ajutor. No child comes he himself to you to ask your help, b. Orice copil vine el tnsu§i la un profesor sa ceara ajutor. Any child comes he himself to a teacher to ask for help. 4. Finally, the distributional contrast is also apparent in the fact that either of the two phrases may itself appear in the DSC, though they themselves cannot cooccur. (59) a. Ion vine el maine §i el, n-ai grija. Ion comes he tomorrow he too, don’t worry. “Ton too will come tomorrow, don’t worry.” 120 Alexandra Cornilescu 20 b. Ion vine el negrept el insup, n-ai grija. Ion conies he unfailingly he himself, don’t worry. “Ion will unfailingly come himself, don’t worry.” (60) *?/om vine (el) negrept p el el insup. Ion comes (he) unfailingly he too he himself. Whatever their correct analysis may be, the combinations p el (he too), el insup (he himself) are clearly phrasal units independent from the verb, while the pronoun in the DSC depends on the verb. These phrasal units are probably distributed in some adjoined (adverbial) positions (perhaps adjoined to SpecVP). 4.2. The Pronominal Subject in the DSC Two initial questions to settle regard the argumental/non-argumental status of the pronoun, and also its phrasal status, i.e. is the clitic subject a head X° category or an XP category, in other words, is it a syntactic or merely a phonologic clitic? 4.2.1. As to the first question, the Romanian pronoun appears to be an expletive, devoid of reference. This is why it may co-occur with a lexical subject, in the same way in which an object clitic occurs with a lexical object in clitic doubling constructions. In languages (or in situations) where the clitic is argumental, it does not normally appear with a lexical subject, since one of the two nominals will be devoid of a 0 role. From this point of view, the Romanian clitic differs from its French counterpart and is similar to the subject clitics from certain Italian dialects discussed in Safir (1985), Rizzi (1986) or Poletto (1991): (61) French II mange. Trentino El magna. Romanian Manama el. (with appropriate intonation) (62) French *Jean il mange. Jean mange. (63) Trentino El Gianni el magna. *El Gianni magna. (64) Romanian Vine el tata. Vine tata. Tata vine el mai tdrziu. 21 The Double Subject Construction in Romanian 121 Moreover, as we already said, in dialectal speech, the clitic has the invariable 3d person masculine form, el, in which case it must be followed by the lexical subject, if the intended meaning is to be expressed. This is clearly an indication that the pronoun is expletive. (65) Ne-a venit el apa la moara. To us has come HE(Nom. Masc. Sg.) the water(Nom. Fem. Sg.) to our mill. “The water has come to our mill.” As to the second question raised above, there is evidence that the pronominal subject is still a DP, not a D, in Romanian. In the typology of Cardinaletti and Starke (1994), it should better be viewed as a weak pronoun, rather than a clitic, comparable to the French il, or the Italian egli/ esso. A comparison with other Romance language is again helpful. Rizzi (1986) has established the existence of two types of subject clitics in Romance. On the one hand, there are head clitics, placed in the verbal agreement head, AgrS0, adjoined to the verbal morphologic features, in a configuration like (66a) below, and there are DP clitics (weak pronouns) which occupy specifier positions (SpecVP, SpecAgrSP), as in (66b) below, cliticizing on the verb in the phonological component: A head clitic part of AgrS0 is often a sort of morphologic specification, which is (always) expressed on the head of AgrSI) independently of the element that is realized in SpecAgrSP, which can be a null subject or a phonetically realized DP. The clitic performs morpho-syntactic functions and is often not involved in referential relations. Thus, there are dialects, where the clitic is the assigner of Nominative Case to the preverbal subject, by means of Spec-Head Agreement. Poletto (1991) discusses dialects like Trentino/ Fiorentino where the clitic is an obligatory spell-out of Inflection, always accompanying the preverbal subject. (67) La Maria la magna. The Maria she eats. *La Maria magna. It may be that the preverbal and postverbal subject check case with different heads. The subject clitic must appear with the preverbal subject, which cl TP cl AgrS0 122 Alexandra Cornilescu 22 checks case with Agr, but cannot appear with the postverbal subject, which checks case in a different manner. This is the case of the Modern Trentino variety (cf. Poletto 1991:43), illustrated bellow: (68) Trentino La Maria la riva. The Maria she arrives. *La Maria riva. (69) a. *La riva la Maria. She arrives the Mary. b. Riva la Maria. Arrives the Mary. In many northern Italian dialects (e.g., Trentino, Fiorentino, Veneto), the clitic subject is also the head capable to license and assign case to pro, performing the morpho-syntactic functions of strong Agr. Pro is licensed by the clitic head in Agr through Spec-Head Agreement, in contrast with French, where the clitic is in SpecAgrSP. These Italian dialects are, therefore, pro-drop languages, unlike French. The only difference with standard Italian is that the strong Agr is realized in phonetic form not only in concrete verbal morphology, but also in its abstract syntactic position, by a sort of reduplication of agreement: (70) a. Pro el magna. b. *- magna. Binding and quantificational facts may also help in distinguishing between clitic and full DPs, if we accept a uniformity condition regarding the phrasal status of chain members (cf. Progovac (1993), Dobrovie-Sorin (to appear)). French instructively contrasts with certain Italian dialects along these lines. French has subject clitics in SpecAgrSP at S-structure and cliticization to V in the phonological component. For some speakers a lexical subject can occur left dislocated with a subject clitic, even without any perceptible intonation break, but a bare quantifier (BQ) is impossible in this environment: (72) a. Jean n 'a rien dit. “Jean has said nothing.” b. Jean il n 'a rien dit. Jean he has said nothing. (73) a. Personne n ’a rien dit. Nobody has said anything. b. *Personne(J il n ’a rien dit. Nobody he has said nothing. 23 The Double Subject Construction in Romanian 123 Rizzi (1986) explains that in the correctly formed quantifier sentences, the preverbal subject must be in SpecAgrSP. It cannot be a left dislocated element since BQs cannot be left dislocated leaving (clitic) pronouns behind. This restriction on BQs is expressed below (cf. Rizzi (1986:395)): (74) A pronoun cannot be lexically bound by a non-lexically restricted quantifier. As Rizzi explains, this statement essentially amounts to claiming that pronouns cannot function as primary variables. They sometimes acquire variable status only parasitically, through binding from licit primary variables. When a pronoun is directly bound by a BQ, an ill-formed chain will result. As known, a QP must move to an A' position at LF, leaving a variable as its trace. On the assumption that in French the clitic occupies the subject position (cf. (66b) above), the ungrammatically of Personnel) il n ’a rien dit follows at once. At LF, BQs must be in A' operator position. If we accept that the BQ already is in an A' position, then the BQ binds the pronoun in subject position in direct violation of (74). If the BQ were to further raise by QR from its left dislocated position, it would leave behind a variable in an A', instead of an A position. Italian dialects like Fiorentino or Torinese sharply differ from French. In these dialects, the subject clitic is obligatory even when the subject DP is a BQ. This is significant since BQs can appear in subject (=argumental) position (SpecAgrSP), but not in a left dislocated position, as seen above. Examples (71) are due to Rizzi (1986:356): (71) Torinese a. Gnun I ’a dit gnent Nobody he(cl) has said nothing Fiorentino b. Nessuno I'ha detto nulla Nobody he(cl) has said nothing The Italian examples are fine, since the clitic is adjoined to Agr° under the Agr° head, in configuration (66a), while the BQ is in the argumental subject position SpecAgrSP. Intuitively, in this case, the clitic is merely an element of verb morphology. The BQ undergoes QR, leaving behind a (licit) variable in the argumental case-marked subject position. As to the clitic, one may follow Rizzi and say that the clitic is bound from a licit primary variable, acquiring variable status, or, preferably, following Dobrovie-Sorin (1997), one may assume that chains should be uniform as to phrase structure status, so that only the clitic trace, i.e. the subject position itself enters a well formed chain of the form (QPi5 tj). However, the situation is not so clear cut, since Poletto (1991:42) cites the Modem Trentino variety where the clitics are, demonstrably, heads, but doubling of the clitic subject by a BQ is, nevertheless, prohibited. 124 Alexandra Cornilescu 24 (75) Trentino a. *Nisun el vien qua. Nobody he comes here, b. INisunvien qua. Nobody comes here. Interpreting such examples, Poletto explains that in this dialect, the clitic is a Nom Case assigner. As such it will be coindexed with the preverbal subject and participate in chains involving the subject position. When the quantifier undergoes QR, the clitic will be coindexed with the subject trace and will count as the member of a chain having the form (QPi, vbli5 clj), in violation of the uniformity condition on chains. The head clitic moreover counts as A-binding the variable left by the raising quantifier, a variable which should be A-bound. The clitic enters a chain having the form (QPi, vbli5 clj), distinct from the chain corresponding to examples (71) above, where only the clitic trace position (i.e. the subject position itself) entered the chain. The possibility of doubling a Bare Quantifier by a subject clitic surely shows that the BQ is in subject position and the clitic is a head, but, as suggested by the Trentino facts, lack of it does not always mean that the clitic is a DP in specifier position. A second reliable test in deciding on the clitic D° vs. DP status of a pronoun is that of coordination. Clitic subjects are repeated when there is verb coordination, as shown by the Trentino examples below (from Rizzi, 1986: 402-403). In contrast, in French the verbal heads are coordinated leaving the subject in SpecAgrSP: (76) French Elle chante et danse. She sings and dances. Trentino a. *La canta e balla. She(cl) sings and dances, b. La canta e la balla. She(cl) sings and she(cl) dances. 4.2.2. Let us examine the Romanian facts now: To begin with, the clitic surely plays no part in Case checking. It appears equally felicitously with preverbal or postverbal subjects. This is not unexpected, since the DSC represents a feature of familiar, affective speech, which could hardly have assumed any morpho-syntactic role in the economy of Romanian. Romanian subject clitics are optional and they behave like full DPs, not heads. Notice first that subject clitics do not have to be repeated under verb coordination (cf. (77a)), although, of course, they may. In contrast, object clitics, which are adjoined to the verbal head (in an agreement projection), being syntactic clitics, must be repeated in verb coordination structures: 25 The Double Subject Construction in Romanian 125 (77) a. Las ’ cd vine el tata §i te Invata minte. Be sure that comes HE father and you(cl) teaches a lesson. “Be sure that father will come and teach you a lesson.” b. Las ’cd vine el tata §i te Invata el minte. Be sure that comes HE father and you teaches HE a lesson. c. Tata vine el §i te invafd (el) minte. Father comes HE and teaches (HE) you a lesson! (78) a. *L-am vazut $i am pldcut. (I) HIM(cl) have seen and have liked. “I have seen him and liked him.” b. L-am vazut $i l-am pldcut. (I) HIM(cl) have seen and HIM(cl) have liked. This is an indication that the pronominal subject is in SpecAgrSP. Applied to Romanian, the binding and quantification facts mentioned above strengthen this result. Like its counterparts in French or Trentino, the Romanian pronoun is sensitive to the type of lexical subject which it doubles. All the attested examples we dispose of have a definite referential postverbal subject (see examples (41)—(43)). What is theoretically significant is that, like in Modern Trentino or French, BQs are excluded in postverbal as well as, even more strongly, in preverbal position in the DSC: (79) a. **Are sa se mire elflecare de una ca asta. Is going to wonder he everyone about a thing like this. “Everyone is going to wonder at such a thing.” b. *Nu vine el nimeni pe vremea asta, n-ai grijd. Isn’t coming he nobody in such a weather, don’t worry. “Nobody will be coming in such a weather.” c. *Cumpard el un elev cartea pdna la urmd. Buys he a pupil the book in the end. “Some pupil will buy the book in the end.” (80) a. **Fiecare are sa se mire el de una ca asta. Everyone is going to wonder he at a thing like this. “Everyone is going to wonder at a thing like this.” b. *Nimeni nu vine el pe vremea asta. Nobody not comes he in such a weather. “Nobody will be coming in such a weather.” c. *Un elev cumpdra el cartea, n-ai grijd. A pupil buys he the book in the end, don’t worry. “Some pupil will buy the book in the end, don’t worry.” 126 Alexandra Cornilescu 26 27 The Double Subject Construction in Romanian Principle (74) above is operative in Romanian as well in examples like (80), similar to the French (73b). In examples (79), the postverbal BQsubject will raise to an A' position at LF, giving rise to a chain of type (BQi? cl„ varj where the variable is again wrongly A-bound by the clitic. Putting together the test of coordination and the impossibility of doubling by a BQ, we are driven to conclude that the clitic subject sits in SpecAgrSP. But now comparison of the Romanian examples with any of the French or Italian ones immediately shows a striking difference. The Romanian clitic is systematically postverbal. In fact the clitic can never be preverbal in the DSC. (81) a. **El vine tata, n-ai grija. He comes father, don’t worry. b. Vine el tata, n-ai grijd. Comes he father, don’t worry. c. Tata vine el, n-ai grija. Father comes he, don’t worry. This paradigm is in fact strong evidence for a postverbal Agr S position in Romanian. The conclusion that Romanian disposes of a postverbal AgrSP seems hard to disprove. The adjacency of the pronoun to the verb is a direct effect of its structural position directly below the verb cluster, a position that makes possible phonological cliticization. At the same time, when the verb raises by Long Head Movement, the subject pronoun as well as the object clitics are left behind: (82) Sldveasca-l ei, proqtii cat or vrea. Glorify(Subj) him(cl) they (cl) the fools, as long as they want. “Let the fools glorify him as long as they want.” A second tentative result, suggested by the similarity of the French examples with Left Dislocation and the Romanian DSC with preverbal subjects, is that the preverbal subject position may well not (always) be an L-related argument position in Romanian. This problem is discussed in section 5 below. 4.3. Problems of Case If the pronoun is a specifier, a full DP, then there is the problem of how Nom Case is assigned to the two DPs. Neither of them can get inherent partitive case, since both of them are definite and partitive case is restricted to indefinites (Belletti (1988)). Also, since both DPs are phonetically realized, Case cannot be transmitted through chain formation, as the case is in (expletive pro, DP) chains. The two Nominative cases are presumably checked in different positions. In the discussion of Case, it is profitable to remember that there exists a subdialect of Romanian where the clitic is invariably marked for the 3rd person singular, masculine, possibly having different (p features from the following lexical subject. When this happens the verb agrees with the clitic, not with the lexical subject. The lack of agreement between the full lexical subject and the verb shows that Agreement is not involved in checking the case of the lexical subject. It is likely that the latter checks its Nom feature against a different head. (83) a. c. Las ’ c-a pdtit-o el hotii. Be sure that has(sg) got it HE(Nom.Masc.Sg.) the thieves (Nom.Masc.PI.). “You may be sure that the thieves will get it!” b. Are sa-l certe el preoteasa pe popd. Is going to him(cl. Acc) snub HE(Norri.Masc.Sg.) the priest’s wife (Nom.Fem.) the priest(Acc). “The priest’s wife is surely going to snub him.” The Romanian facts are parallel with those of Renaissance Veneto (RV), discussed in Poletto (1991). except that the clitic is preverbal. In this dialect a lexical subject that has undergone free inversion cooccurs with a subject clitic in preverbal position. The verb agrees with the clitic, not with the postverbal subject: (84) a. RV El viene quel suo fraelo. He (cl) comes that his brother, b. R Vine el fratele acela al lui. Comes he(cl) that brother of his. “That brother of his will come.” (85) L 'e sta suspesso le prediche al Si or Geronimo. Cl is been suspended the sermons to Mr. Geronimo. The subject clitic was, at the time, a full DP (cf. Poletto, op.cit.) and got Nom Case by agreement, as the verbal morphology indicates. How did the full DP subject get Case? To solve similar problems, Roberts (1991) assumed that there were two means of assigning Nom Case in UG, in two distinct configurations: (86) a. Agr assigns Case through Spec-Head Agreement, b. T assigns Case through Government. Such a parameter of Nom Case assignment could, according to Roberts, explain the difference between languages like French and Welsh. In French, the subject is preverbal and triggers morphological agreement in person and number with the verb. French employs the option expressed in (86a). In Welsh, the subject is postverbal and does not trigger morphologic agreement in person and lumber with the verb. Welsh avails itself of option (86b). Roberts further assumes that, in Romance languages, the subject can be ireverbal, as well as postverbal, because both options can be selected. Nevertheless, inguages like standard Italian always show morphological agreement of person 128 Alexandra Cornilescu 28 and number, even when the subject is postverbal, while Welsh never does. Presumably, in Welsh, Agr is never active in Nom Case assignment, while it always is in Romance. To account for this difference in the behaviour of the postverbal subject, Roberts (1991) formulates a rule of cosuperscripting between the heads of AgrP and TP, which applies in Romance, but not in Welsh. The cosuperscripting rule determines the sharing of the morphological agreement features of person and number, so that the verb and the postverbal subject agree in person and number in Romance. (87) Coindex Agr and T. A rule like (87) is then active in Romance, because both Agr and T are able to assign Nom, but it will fail to apply in Welsh. Adapting Robert’s ideas to current theory, we might claim that Nom Case may be checked either in SpecAgrSP, or in SpecTP, with independent properties of the language determining which option is taken. Let us come to the facts of Renaissance Veneto and dialectal Romanian, where the postverbal subject does not agree with the verb, while the clitic does. The configuration of the Romanian sentences under discussion is as follows (88) ...[AgrSp clitic subject [ V° + Agr0 [TPDP2tv...[vPt2[vtv... }]]] Apparently, in Renaissance Veneto and in dialectal Romanian, just as in other Romance languages, both mechanisms of case assignment can be exploited in principle, and must be exploited in certain situations like the DSC. The clitic subject necessarily occupies the position SpecAgrSP, since a weak pronoun or a clitic must reach its case checking position by Spell-Out. A second reason for the clitic to be in SpecAgrSP rather than in a lower position is that the clitic is expletive and SpecAgrSP is a non-thematic position. Consequently the clitic must check its case by agreement with the verbal agreement head. The postverbal subject checks case by overt or covert movement to SpecTP. Cosuperscripting of Agr and T does not apply, so the postverbal subject does not agree with the verb. The double mechanism of case checking may account for the frequent lack of agreement of the (postverbal) subject with the verb in substandard Romanian (e.g. Se vinde masline ((It) sells olives)). In standard Italian/Romanian, unlike the Welsh or the Romanian dialect discussed above, there is agreement with a postverbal subject. This may be interpreted as showing that standard Romanian/ Italian disposes of only one Nom checking position, namely, SpecAgrSP. Postverbal subjects in SpecVP may check case covertly at LF, as proposed in Cardinaletti (1996). Alternatively, expletive pro may be inserted in SpecAgrSP, forming a chain with the postverbal subject (cf. Alexiadou, 1994). In the latter case, Nom Case would be uniformly checked before Spell-Out. There is standardly a unique Nom checking position, namely SpecAgrSP. 29 The Double Subject Construction in Romanian 129 As to the DSC, we may claim that exceptionally, through cosuperscripting of Agreement with Tense, the case-checking potential of Tense is “activated”, i.e. the case/ person/ number/ gender features of Agreement are transmitted to its complement, Tense, allowing a second nominal, the lexical subject in the DSC, to check its Nom feature overtly or covertly. The preferred word order V+ expletive pronoun - lexical subject, typical of the DSC, is precisely the order MP > AgrSP > TP postulated in our analysis. Alternatively the lexical subject may stay in Spec VP and check Case at LF: (89) a. Nu vin ei copiii maine. (Do) not come they the children tomorrow. (90) b. Vin ei maine copiii, n-ai grijal Come they tomorrow the children. 4.4. Conclusions on the DSC • 1. The pronoun in the DSC is an expletive weak one, cliticizing on the verb at phonological form, so there is a subject clitic in Romanian as well. 2. In the DSC there are two full DPs in the Nom Case, which must check case in different positions. The lexical postverbal subject checks case in SpecTP, moving into this position overtly or covertly. In the present analysis, the SpecTP position is made use of only in the DSC. In contrast, other researchers (e.g. Motapanyane (1989)) had argued that all and only postverbal subjects get case from the verb in Tense under government, while preverbal subject checked case in a preverbal AgrS position. Dobrovie-Sorin had also argued for a unitary mechanism of Case assignement: Nom was always assigned in Spec VP by the verb raised to Inflection. 3. Since weak forms and clitics occupy their case-checking position before Spell-Out, the clitic occupies the position SpecAgrSP before Spell-Out. In fact the clitic is generated in that position, since it does not need any 0 role. 4. The clitic always agrees with the verb, even when it does not agree with the full subject. This shows that the clitic checks its case by Spec-Head Agreement with the verb in AgrS0. 5. Since this position accomodates and assigns case to expletives like the invariant EL in the DSC, it will also be the position which licenses (referential and expletive) pro. Pro behaves like a weak pronoun and must move into its Case-checking position before Spell-Out. 6.Outside the DSC, when the clitic is not present, the case of the postverbal subject is uniformly checked in SpecAgrSP (cf. also Cardinaletti (1996)). This idea may be implemented in two ways: a) An expletive pro in SpecAgrSP forms a chain with the postverbal subject, b) Alternatively, the postverbal subject checks its case covertly at LF. 130 Alexandra Cornilescu 30 Diesing’s important insight that semantically weak subjects are in VP (or in the AspP according to Borer (1994)) while strong subjects move out of the VP into SpecF in English is easily expressed in our analysis as a difference between subjects that check case covertly (weak subjects) and subjects that move overtly (strong subjects). 5. On the nature of the preverbal subject position in Romanian In the previous sections, we have tried to prove that Romanian has a postverbal AgrSP below MP, and that SpecAgrSP is the Nom Checking position, as well as the position which licenses pro in Romanian. In this section, we will try to establish the claim that this is the only AgrS position of Romanian. At stake is the nature of the preverbal subject position. Two views have so far been expressed on this matter, as already announced: a) The preverbal subject position is a topic, left dislocation position (cf. Dobrovie-Sorin, (1987), (1994). b) The preverbal subject position is argumental and it is a Nom checking position for preverbal subjects. This second view is expressed in Montapanyane (1989). 5.1. Previous results The arguments adduced to defend the A, or the A', nature of the preverbal positions are briefly summarized in this paragraph. Dobrovie-Sorin (1987) argues that the subject, which has got Case from the governing Inflection, scrambles to Sped', a topicalization position in Romanian, more exactly a position of left dislocation. The main argument, very well-sustained by the data, is that several XPs may be topicalized in Romanian, preceding or following the topicalized subject: (91) a. Ieri Ion drept in mijlocul curtii facea plajS. Yesterday, Ion right in the middle of the yard was taking the sun. “Yesterday, Ion was sunning himself right in the middle of the yard.” b. Drept in mjilocul curtii Ion fdcea ieri plaja. Right in the middle of the yard Ion was taking yesterday the sun. c. Ion ieri drept in mijlocul curtii facea plajQ. Ion yesterday right in the middle of the yard was taking the sun. (92) a. Ion tot flori i-a adus Mariei. Ion also flowers to her (cl) has brought to Mary. “Ion has also brought flowers to Mary.” b. Mariei Ion tot flori i-a adus. To Mary Ion also flowers to her(cl) has brought. 31 The Double Subject Construction in Romanian 131 c. Flori Mariei ION i-a adus (nu Petre). Flowers to Mary ION to her(cl) has brought (not Petre). We find the evidence compelling. Motapanyane (1989) brings proof in favour of the view that SpecAgrS is an argumental preverbal position in Romanian. Briefly, she claims that there are preverbal constituents which cannot be considered to be left dislocated, but which nevertheless may function as preverbal subjects in Romanian. One example is that of BQs. She argues that BQs must occupy A-positions at S-structure, so as to form well-formed chains when they raise at LF. Therefore, at least for such constituents, there has to be an available preverbal subject A-position. To prove her point, Montapanyane (1989) compares the behaviour of Objects and subjects. She notices that strong direct objects can be emphasized by being placed between the subject and the predicate, and finds an important contrast between sentences like (94), where a definite direct object has been scrambled, and sentences like (95), where the scrambed direct object is a BQ: (93) a. Ion pe top ii invitase la masa. Ion all of them(Acc) them(cl) had invited at dinner. “Ion has invited all of them at dinner.” b. Ion pe soldati ii admira, nu pe profesori. Ion PE soldiers (Acc) them(cl) admired, not the teachers. “Ion admired the soldiers, not the teachers.” c. Pe soldati Ion ii admira. PE soldiers (Acc) Ion them(cl) admired. (94) a. Profesorul a invitat pe cineva la masa. The professor has invited PE someone(Acc) at dinner. “The professor has invited someone at dinner.” b. *Profesorulpe cineva a invitat la masa. The professor PE someone(Acc) has invited at dinner. c. *Pe cineva profesorul a invitat la mas ft. PE someone (Acc) the professor has invited at dinner. She explains (1989: 22-23) that (94 b, c) are ungrammatical since a BQ cannot scramble out of an argumental position, because it would end up in an A' position at the level of S-structure. Given tfcat, at LF, the BQ must further raise to an IP adjunction position to bind its variable, if the position out of which it raises by QR is itself A', an ill-formed chain will result, where the variable the BQ direcly binds is in an A'-position, instead of an A-position. To avoid this undesirable result, BQs in argumental positions must stay in this position up to LF; this makes scrambling impossible. This is the very same restriction discussed by Rizzi (1986) and above, preventing the left dislocation of BQs. 132 Alexandra Cornilescu 32 33 The Double Subject Construction in Romanian 133 both However, Montapanyane notices, in Romanian BQs function as preverbal and postverbal subjects. (95) a. Cineva a batut la u§a. Someone has knocked on the door, b. A bdtut la ufja cineva. Has knocked on the door someone. “Someone has knocked on the door.” By parity of reasoning, the BQ subject should also find itself in an argumental position at S-structure; this leads to the conclusion that the two subject positions hosting the BQs in the examples above must be argumental Case positions. Consequently, SpecAgrS is a Case position for Montapanyane: “Par consequent, le sujet ne monte pas en ‘scrambling’ en SpecAgr, mais pour recevoir le cas assigne a cette position.” (1989:25) This elegant argument cannot really stand, I believe. In the first place, empirically, the examples presented above are not convincing, since, in spite of (94 b, c), there surely are instances where an emphatic (focalized) direct object BQ felicitously appears between the subject and the predicate. Examples of type (94b) become acceptable if cineva (someone), harder to emphasize, but still a possible focus, as (96c) shows is replaced by nimeni (nobody) (96a) or oricine (anybody) (96b). An emphatic quantified direct object may, certainly, appear between the subject and the predicate, in the same position as in (94b), while examples like (97) show that it can also felicitously precede the preverbal *>ihipict in the same position as in (94c): a. Profesorul PE NIMENI n-a invitat la el. The professor NOBODY he has invited to him. “The professor has invited NOBODY to his place.” b. Profesorul PE ORICINE ajuta cum putea. The professor ANYBODY helped as he could. “The professor helped ANYBODY as he could.” La urma urmei chiar §i Maria PE CINEVA tot va trebui sa invite " 'T-/-VMT7 „+;„ . The view that we defend is that in Romanian there is only one postverbal SpecAgrSP position, which is the standard Nom checking position, at least in finite clauses. The preverbal subject position(s) is/are not L-related, so that the preverbal subject appears to have (operator) A' properties. 5.2. The Left Periphery As announced, a brief review of some of the more recent results regarding the structure of the left periphery of the clause is welcome at this point since we believe that the Romanian preverbal subject always occupies positions in that area. In a synthetic presentation of this problem Rizzi (1995), proposes that the C system, actually the CP > IP(operator) area may be analysed into a number of positions, which check different non L-related features of sentences, defining a force-finiteness system with the following configuration: (98) (Force)... (Topic)... (Focus)...(Fin IP) subject, in (96) The Fin(iteness) system is the old IP, therefore the M-projection for Romanian. The highest projection is the old CP, supplying differences of illocutionary force, through features like [imperative], [declarative] etc. For the present analysis, what matters most is the Top(ic) Fo(cus) articulation of the sentence. The Topic system interprets a sentence as to its Topic/Comment structure, while the Focus system interprets its Focus/Presupposition articulation. Although the two articulations are formally similar in many languages, there are clear, inter-pretational differences, sometimes correlated with distributional differences, allowing one to distinguish between Topic/Focus (Rizzi (1995), Zubizaretta (1993)). The basic syntactic difference between the higher operator projections and the lower morpho-syntactic projections is that the specifier of the former is an A1 position, checking an operator feature, while the specifier of the latter is an L-related A position. A constituent endowed with topic or focus features must end up in a Spec-Head configuration with a Top or Foe head, containing the respective feature, to satisfy the Top/ Foe Criteria. In Romance, topics may involve resumptive clitics. A typical topicalization structure in Romance, also available in Romanian is the Clitic Left Dislocation h ar si Maria LX/vc-r/i yu "“'T Structure (CLLD), whose syntax is discussed in Cinque (1990) for Romance c. La urma urrn<*c 1 qQ]ylEONE still will have„ and Dobrovie-Sorin (1994) for Romanian. A topicalized director indirect object After all, even ary ^ave invite SO is always clitic doubled in Romanian. “After all, even ary The following are the most salient Topic/ Focus differences. rFVA chiar Maria §tie la matem matics a) Focus is quantifieational, topic is not. This is the main difference: (97) a thWQ even Mary knows at ma e ^^between them partitioning the class of A1 dependencies into those that involve a hich wenow turn, has prove quantifier that binds a variable and those that involve non-quantificational c f'nndlv more recent researc to w Spell-Out is not’m binding, i.e. binding of a null constant (nc), a DP trace, as with CLLD. the occurrence' of a BQ in an A’ position b theoretically disturbing either. 134 Alexandra Cornilescu 34 35 .The Double Subject Construction in (99) a. IL TUO L1BRO ho comprato t. (Focus) YOUR BOOK (I) have bought. b. *IL TUO LIBRO lo ho comprato t. YOUR BOOK it (I) have bought. (100) a. IL TUO LIBRO lo ho comprato. (Topic) YOUR BOOK it (I) have bought. b. *IL TUO LIBRO ho comprato. YOUR BOOK (I) have bought. The Focus legitimately binds a variable in (99a), but not in (99b), where neither the clitic nor its trace is a variable. Under classical assumptions on the typology of empty categories, (100b) is also ruled out right away; the topicalized element is not quantificational. Therefore the ec in object position has no legitimate status; it cannot be a variable, as there is no quantifier to bind it. It is in fact a clitic trace, so that structure (100a), with the empty category bound by a clitic is well-formed. Consider now some Romanian examples: (101) a. CARTEA TA am cumparat-o. (topic) YOUR BOOK (I) have bought it (cl). b. * CARTEA TA am cumparat. YOUR BOOK (I) have bought it. c. CARTEA TA am cumparat-o, mi pe a lui. (non-quantificational focus) * CARTEA TA am cumparat. (102) a. NIMIC n-am cumparat. (focus) *NIMIC nu l-am cumparat. (103) a. NI§TE CARfl am cumparat, nu ni§te caiete. (focus) SOME BOOKS (I) have bought, not some copybooks, b. *NI$TE CARJ1 le-am cumparat, nu ni§te caiete. SOME BOOKS them (cl) (I) have bought, not some copybooks. In Italian, Focus and Topic are formally associated not only with distinct intonation, but also with different types of chains, involving variables in the case of Focus, but DP-traces in the case of Topic. In other words, Focus is always quantificational. This is not always the case in Romanian, where the form of the chain, specifically doubling by a clitic, depends more on the inherent properties of the moved DP, and less on the Topic/ Focus interpretation. Thus fronted definite direct/ indirect objects are invariably doubled even when they are contrastively stressed (focalized), as shown by example (101c) in contrast with the Italian (99b). What matters, however, for the present discussion is that a BQ in a focus configuration will always be marked by the absence of the resumptive clitic in Romanian as well. Examples (102, 103) involve quantificational Focus, behaving like the Italian examples in (99). ____________135 Thus, the focalized elements in (102) and (103) are genuine quantifiers and must bind syntactic variables. This happens in (102a, 103a) but not in (102b, 103b), in which potential bindees are the clitic and its trace, neither of which qualifies as a syntactic variable i.e. (a non-pronominal empty X-max category in an A position). We retain that in Romanian, BQs clearly distinguish between Topic and Focus: they are excluded in the CLLD structure, i.e. that cannot be topics, but they easily allow focalization. b) While A' dependencies are all sensitive to strong crossover, weak crossover may be used as a second test to distinguish between the two classes of A' dependencies. A topic never gives rise to any weak cross-over effects. Such effects are detectable with a focus, however. (104) a. Un copil, nu-l, abandoneaza maicd-sa,. (Topic) A child(Acc) not him(cl) abandons his mother. “His mother doesn’t abandon a child.” b. ? *UN COPIL, nu abandoneaza maicd-sa,. (Focus) A child(Acc) not abandons his mother. (105) a. Pe oricare copil, il, rasfata mama lui,. (Topic) Any child(Acc) him(cl) spoils his mother. “His mother spoils any child.” b. *PE ORICINEj rasfata maica-sa,. /mama lui,. ANY CHILD (Acc) spoils his mother. “His mother spoils any child.” A further illustration of this difference is the contrast between questions and appositive RCs. Questions, which (may) involve genuine quantifiers, may be sensitive to weak cross-over effects, in contrast with appositive relative clauses (in 108), which merely involve anaphoric pronouns. (106) a. ? *Pe cine, iubeqte mama lui, cu adevarat? b. ? *WhOj does hiSj mother really like? (107) *Ce copil, ajula mama lui, la traduceri? What childj helps his, mother at his translations? “What child does his mother help with his translations?” (108) Ion, pe care, maicd-sa, tl iube$te cu adevdrat, este un copil talentat. John, whoj his, mother really likes t, is a talented child. c) Uniqueness further distinguishes between Focus and Topic. A clause may contain as many topics as are consistent with its topicalizable arguments, as in example (109a) below; on the other hand, there is a unique focus. A focus 136 Alexandra Cornilescu 36 37 _The Double Subject Construction in Romanian 137 and one or more topics can be combined in the same structure. In that case, the focal constituent can be both preceded and followed by topic constituents, as in example (109b) below: (109) a. Mariei cu banii acegtia florile nu i le pop cumpdra. For Mary with this money the flowers (you) not her(cl) them(cl) can buy. “You cannot buy the flowers for Mary with this money.” b. Lui Ion GRE$EALA ASTA laproces nu i-o iartd nimeni, sunt sigur. To Ion THIS MIST A KE at the trial nobody not to him(cl) it(Acc.cl) is going to forgive, I’m sure. “Nobody is going to forgive Ion this mistake at the trial, I’m sure.” 5.3. Our View We may now return to the preverbal subject position in Romanian. Since Case and 0 features are checked in AgrS or lower in TP, it is difficult to see why another L-related position would be needed. There are however several logical possibilities which would warrant one more preverbal AgrSP. First, it is still conceivable that certain subjects may not cheek case in the lower AgrS position, simply because this position is empty, as is the case in non-finite clauses. We will not deal with non-fmite clauses in this paper (See Dobrovie-Sorin (1987, 1994), Cornilescu (to appear)). In the same line of thought, a very strong case for a preverbal AgrSP structure would be the existence of situations where Nom Case could only be checked in this preverbal position. For instance, the preverbal subject might get Case in that position from some case assigner in a higher functional projection, such as C°. A situation of this type is discussed in Rizzi (1982, ch3, ch4), who convincingly shows that in a class of gerundial complements, Nom Case is checked by an auxiliary that has raised to C° (see Rizzi’s Aux -to Comp rule). In Italian examples like (110) a full subject can in fact be found in a position internal to the sentence (i.e. to the IP), comparable to the position of the sub ject in English. Notice that a BQ subject (qualcuno “someone”) is grammatical in (110) (due to Cardinaletti (1996: 10)) although BQs cannot be topicalized and have to occupy argumental positions. (110) Avendo Gianni /egli /qualcuno telefonato a Maria... Having Gianni/ he/ someone telephoned to Mary... Although identifying a Nom position as a position governed from C° is surely correct, this line of reasoning cannot apply to Romanian, since in Romanian there is no V°-to-C°, so there will be no way of identifying a Nom position as one governed by the V in C°. Secondly, it may be and has been argued that, even in finite clauses, a preverbal SpecAgrSP position is needed to accomodate those preverbal elements, for instance BQs, which must be in A-positions before Spell-Out, to avoid the formation of ill-formed chains. This is the argument offered by Motapanyane (1989), presented above. However, the mere presence of BQs as preverbal subjects is not sufficient reason to speak of a preverbal AgrSP. The operator area of the clause does contain a position that accomodates quantificational elements and out of which quantifiers need not raise at LF; this is the focus position. BQ subjects might thus occupy the Focus position when they are preverbal, and perhaps they always do. On the other hand, following Cardinaletti (1996), we might hypothesize that there are two AgrSP positions, one of them postverbal, and one preverbal. But this option would imply finding evidence that the preverbal position is argumental. Only then could it be accepted that there is a second preverbal AgrSP in Romanian. The present analysis has, in fact, been unable to find evidence for a preverbal case/argumental subject position in Romanian. On the contrary the facts we have examined suggest that there is no preverbal argumental subject position in Romanian. The preverbal subject occupies one of the positions in the left periphery of the sentence, namely it is either Topic or Focus. Two types of arguments are presented: a) At least in certain contexts, like that of the DSC, the preverbal subject can only be described as a Topic. Even though it is directly preverbal, apparently like a subject in a preverbal AgrSP, it has only Topic properties. Moreover, constituents which cannot be topicalized cannot be preverbal subjects either, clearly because there is no AgrSP to accomodate them. b) Since there is no preverbal AgrSP, Romanian possesses only one preverbal position for quantifiers, the focus position. Consequently there are distributional differences between languages like English, French, Italian, etc. which have two preverbal positions for BQs, namely SpecFocP and SpecAgrSP and Romanian where only one position, Focus, is accessible to BQs. 5.4. The Preverbal Position in the DSC Ordinarily, the preverbal subject is correlated with traces in SpecAgrSP and Spec VP, and these positions may sometimes contribute to the properties of the resulting chain. The DSC construction has the advantage of throwing light on the properties of the preverbal position in contrast with the postverbal AgrSP position which is occupied by the pronoun. The characteristic chain of the DSC (preverbal subject + (weak/clitic) Jronominal subject + DP-trace) could in theory be either an A or an A' chain, lepending on its head. 138 Alexandra Cornilescu 38 If the preverbal subject position were argumental, i.e., a preverbal Agr,SP were available, any subject could appear in this construction, including BQs. A BQ in SpecAgriSP before Spell-Out would raise at LF, say to check a focus feature, correctly leaving behind a variable in the higher preverbal SpecAgr,SP (see (111)). This variable is licit, since it is A'-bound by the raised BQ, and it cannot be A-bound by the lower pronoun in SpecAgr2SP; in fact, it will c-command, and can licitly bind, the pronoun in the lower Agr2SP, a pronoun which becomes a secondary variable (cf Rizzi (1986)). The chain would be as in (111) below at LF: (111) [focpBQ, FOC0+ [Focus][AgrlPt, [MpV0+Agr#...[Agr2sp clitic subject [wti [t„... On the other hand, if, as I belive, there is no preverbal AgrS, a BQ should be excluded from the DSC. Before Spell-Out, the BQ would already be in an A' position where it could check its operator feature. But then, in the DSC, the BQ cannot bind a variable, since in the chain in (112) below neither the clitic in SpecAgrSP, nor its trace in SpecVP qualifies as a variable. (112) [FocPBQ ifoco [Focus] [mp V° + Agr0... [Agrsp clitic subject [vp ti [tv... The (non)-acceptability of preverbal subject BQs is a vital test for deciding between (111) and (112). But, as we have already mentioned, BQs are completely ungrammatical in the DSC. This proves that there is no SpecAgrSP to host them. Here are examples: the structure underlying them is (112). (113) a. *Nimeni nu vine el la seminar pe vremea asta. Nobody not comes he to the seminar in such a weather. “Nobody will come to the seminar in such a weather.” b. * Cineva nu spune el a§a o pros tie. Somebody not says he such nonsense. “Somebody will not say such nonsense.” The preverbal subject position in the DSC is not argumental; it is a non- quantificational A' position, therefore a Topic, left dislocation position. The chain (preverbal subject + (weak/ clitic) pronominal subject + DP-trace) has to be interpreted as an instance of the CLLD construction. We would, incidentally. not like to take sides on the issue of whether the left dislocated element is base- < —Cinaue (1990)) or involves non 139 The Double Subject Construction in Romanian Several properties of the construction confirm that the DSC with preverbal subject illustrates the Topic/ Comment articulation of the sentence, the Topic being left-dislocated. ivxi. uioiUUdlCU. a) First, the lexica! subject in the DSC does not allow contrastive stress preverbally or postverbally. This means that, in this construction, the lexical subject cannot be interpreted as a (non-quantificational) focus. (114) a. *TATA vine el, n-ai grija. b. *Vine el, TATA. b) Secondly, since the DSC is not quantificational, but involves topicalization, one consequence that can be detected in this construction is that there are no weak cross over effects: (115) a. Copilul, nu prea este el, iubit de mama lui,. The childj not really is he* loved by his, mother. “The child is not really loved by his mother.” b. Copilul, chiar de maica-sa, nu prea este el, iubit. The child even by his mother not really is loved. c. Copilul, nici pe maica-sa, n-o prea iube§te el,. The childj neither hisj mother (her cl.) really loves he;. “The child doesn’t really love even his mother.” c) The behaviour of questions is also revealing. As mentioned above the questioned word is an interrogative quantifier and it is naturally assumed to move to SpecFocP, the quantifier position. Consequently, a questioned subject excluded in the DSC, because its preverbal subject is a Topic, not a Focus. (116) a. *Cine vine el mai tarziu? Who comes he later? “Who will come later?” b. lion, care vine el mai tarziu, are sa-$i ia banii de la secretariat. Ion, who comes he later, is going to pick up his money from the secretariat. “Ion, who will come later, is going to pick up his money from the secretariat.” is be interpreted as au ui» - - ft dislocated element is cannot like to take sides on the issue o w e _ (1990)) or involves non On the other hand the DSC is possible if a ™ generated in a peripheral position (c . q 0n either accounted is questioned. The topicalized subiect nfth dlfferent from the quantificational WH Movement (cf Dob< 99 » pazedsbect of the DSC must be placed ab, the expletive pronoun in Sp_ecAgrSP w,«hj»h.ch ft.^ ^ ^ ^ ^ « - examples (1,7) belo _ Viiv quantificational WH iviovcu.cn. v- — , ^ ^ dislocated constituent wu-rocus position of^th^intorogativTword Ts iff ""“f he^Ced above the expletive pronoun in SpecAgrSP .c^pcVP position. Its trace u Also as expected, in the non-anantifi^Qt- i xamPles (117) below. is coindexed, licenses a DP in the arS“"^“ * ft di located DP - (weaF subject itself may be relativized in the DSCThe reM™ Constructlon' SpecVP leads to the formation of chan* of the type (Le aced ^ ^ ^ ^ ^ M e Th«dauve prono™ is probably clitic) pronominal subject + DP trace). the left dislocated pos,t.on, as in (118). (117) a. Ion cand o veni el acasQ, cine §tie. Ion when will come he home, who knows. “Who knows when Ion will come home.” b. Femeia unde s-o fi dus ea dupa ajutor, cine poate afla? The woman where will have gone she for help, who can find out? “Where will the poor woman have gone for help?” (118) Fratele meu, care §tie el ce face, te va ajuta. My brother, who knows he what (he) does, will you (cl) help. “My brother, who knows what he does, will help you.” d) In a more recent analysis of CLLD. Dedyskoptter (1994) claims that CLLD has mixed argumental and non-argumental properties. For instance, according to him, unlike wh-movement and (English) topicalization, CLLD allows floated quantifiers (= FQs). (119) French Aux enfants je leur ai tous parle. To the children I to them (cl) have all spoken. “I have spoken to all the children.” This is an argumental property, which should characterize preverbal subjects (cf. Cardinaletti (1996)). Preverbal subjects in the Romanian DSC are surely compatible with FQs. But of course in the examples below, illustrating DSCs with FQs, there is no need to assign any role to the dislocated phrase in licensing the FQ. We may simply say that the FQ has been licensed by the pronoun in SpecAgrSP, rather than by the lexical subject in LD position: (121) a. Egli e molto simpatico. He is very nice. b. * Egli a Maria [pro non le ha parlato anocra\ He to Mary not to her (cl) has spoken yet. Comparable examples are not available in Romanian. All types of pronouns may appear in subject position even followed by topicalized constituents or by parentheticals. [In Cardinaletti’s analveic ----------------- in the higher AgrSP could be followed bv oarenth 1*1 — T? - __ - ^ ^ f» w-ci uy ujpicanzed >arentheticals. [In Cardinaletti’s analysis only strong pronouns _P could be followed by parentheticals, while weak pronouns, like French, //, Italian egli/esso, could not]. (122) a. El Mariei , w ur^ftvi nu i-a vorbit incd. He to Maty not to her (cl) has spoken yet. H A c tn >' - - ' ~ b. Asta, dupa cate vdd This, -- ^ c e o prostie. This, as far as (I) can see is nonsense, c. El/ Acesta, dupa cate vdd, este un geniu. He/ This, as far as (I) can see is a genius. Significantly, the only clear instance of a (subject) weak pronoun in Romanian, namely, the pronoun occuring in the DSC, cannot appear in preverbal position, because weak pronouns cannot be dislocated. Again, there is no preverbal AgrS position which could accomodate the weak pronominal subject: (123) a. *El vine tata. He comes father, b. Vine el tata. The impossibility of a preverbal weak pronominal subject ir important piece of evidence proving the correctness of our anal Tilt31 ■fcsr^f-o ~ — --- 1 ----~ pivvciuai weak pronominal subject in the DSC is . ^0 an important piece of evidence proving the correctness of our analysis. (120) Copiii vin ei mdine toti, n ai g vvhile top is The facts examined regarding the DSC testify that the preverbal subject ' . . the LD position, ei is in SpecAg > must be a topic in this construction: constituents that may not be topicalized In (120), Copiii sits in D§c 0ffers a(BQs, weak pronouns) cannot appear as preverbai subjects in thi« mno in SpecVP. ^ ^ mav sav that the lexical subject m t e < ^ 3ecai_ _ ShT* "" t. " . lexlcal subject in the DSC offers afBQs^k^ -ay ^ __ In conclusion, one: may a u h must be interpreted as a Topic, because there .s no preverbal SpecAgrSP to host tlTenT ' “ COnstructlon> example of preverbal subject wmc DSC proves that there is & ost mem- Futhermore, the non-occurrence o su j 5.6. Bare Quantifier Subjects bo preverbal AgrSP in Romanian. As a matter of •' ' ~ 5 5 Weak pronouns cannot be topicalized as indicated that there is an A' position which caifac** Pe”phei^ ofthe clause cnhiects are left-dislocated inructure, namely Focus. This is nr mav h accomodate BQs at surface Arguing against the view that V^.*ence of weak subject pronounshich the BQ does not need to further raise at LF^H f Cat,°I!al position out of Italian, Cardinaletti (1996) mentions nevertheless occur as preverbalay bind the subject variable in SpecAgrSP ’ the quantifier which cannot be left- chslocat,* 1(see (121a)). Sh| . . We will therefore assume that^the preverbal BO , h * subjects SpecAgrSP: the Italian egl another left-dislocated element'sition and will try to give evidence fnr th' i subjects occupy the Focus remarks that “weak pronouns cannot precede another left ^ ^ ^ ^ ^ ££ ev,chum We argue that, given that (1996:10) (cf. (121b)). ’ ere alread-v exists another focalized element 142 Alexandra Cornilescu 42 in the left periphery, a preverbal BQ subject simply cannot appear at all, either before or after this Focus, since a topic position is not accesible to BQs and Romanian does not have any preverbal AgrSP position. In the Romanian sentence, there is a unique preverbal quantifier position, the Focus one. This is not true in English, or any other languages that have a preverbal AgrSP, where one quantifier may be in SpecFocP and the other in SpecAgrSP. (124) Whom does no one love? What did someone say? 5,6,1. Relying on the uniqueness of Focus, we will examine several facts having to do with the syntax of questions, which suggest that there is no preverbal AgrSP in Romanian. We accept that the questioned constituent is a (quantificational) Focus. Evidence that the questioned constituent occupies the unique Focus position comes from the fact that a questioned constituent is incompatible with a second Focus, appearing before or after it. (125) a. *LA PARIS cine pleaca maine? TO PARIS who leaves tomorrow'? "Who leaves for Pans tomorrow?” a. *Cine LA PARIS pleaca mame? Who TO PARIS leaves tomorrow9 “Who leaves for Pans tomorrow?” Appositive relative pronouns, probably in SpecC/Force, which are not quantifiers but resumptive pronouns, arc. in contrast, compatible with focalized constituents: (126) Omul acesia, care IN VIA JA LUI n-a tnvaiat nimic, ne invatape noi ce este lingyistica. This man, who IN HIS LIFE not has learned nothing, us(cl) is teaching us what linguistics is. "This mao. who has no! learned anything in his life, is teaching us what linguistics is.” We will then take for granted the uniqueness of a Focus position, accomodating focalized constituents, including quantifiers and wh-constituents. As mentioned above, questioned constituents in Focus may be preceded by a Topic constituent, in the same TOPIC + FOCUS order found in declaratives, in examples like (96) above: (127) a. Mariei ce-i cumpdrd tata? — To Mary what to her (Dat.cl) buys father? What will father buy for Mary? 43 The Double Subject Construction in Romanian 143 b. La Paris cine pleaca? To Paris who leaves. “Who leaves for Paris?” In particular the focalized/ questioned constituent may be preceded by a topicalized subject. (128) Profesorul PE CINE prefers? The professor WHOM prefers? “Whom does the professor prefer?” The same sequence topicalized subject + focalized DO was apparent in our earlier examples (96), one of which is repeated below: (129) Profesorul PE N1MEN1 n-a invitat la el. The professor NOBODY he has invited to him. “The professor has invited NOBODY to his place.” It is essential that the subject should be a Topic in such examples because there are again restrictions on the possible types of subjects. It is again the case that BQs which cannot be topics, cannot appear in such structures either (see (131)—(133) below). Since there is only one Focus position, already occupied by the wh-word, we expect and get contrasts between acceptable cases where a Topic (LD) subject constituent precedes an interogative quantifier (130a, 132a), and unacceptable sentences where a non-interrogative quantifier subject precedes an interrogative one (131a, 133a). In every case, the subject may appear postverbally (the b examples of each pair below): (130) a. Ion pe cine prefers? Ion whom prefers? “Who does John prefer?” b. Pe cine preferS Ion? Whom prefers Ion? (131) a. * Cineva pe cine prefer^ in asemenea cazuri? Someone whom prefers in such cases? “Who does someone prefer in such cases?” b. Pe cine prefers cineva in asemenea cazuri? Whom prefers someone in such cases? (132) a. Ion cui nu-i ajutS? Ion to whom not him(cl) help? “Who doesn’t John help?” b. Cui nu-i ajutS Ion? Whom not him (cl) help Ion? 144 Alexandra Cornilescu 44 (133) a. *Nimeni cui nu-i ajuta? Nobody to whom not him (cl) help? “Whom does nobody help?” b. Cui nu-i ajuta nimeni? To whom not him (cl) help nobody? The two quantifiers compete for the same position in order to check their Focus feature, the sentence resulting in ungrammatically. In conclusion, the subject position to the left of the Focus position is not AgrSP, but a topic position. At the same time, questions also prove that there is no preverbal AgrSP position below the FocusP either. Cardinaletti (1996) mentions that in Italian a full subject, in SpecAgrSP is sometimes allowed to occur between the wh-constituent and the verb. This position is, however, not available for left-dislocated items (134b), which precede the wh-phrase in Italian (134c): (134) a. A chi Gianni/nessuno ha consegnato Vinvito? To whom Gianni/ nobody has given the invitation? b. ?A chi Vinvito Gianni I’ha connsegnato? To whom the invitation Gianni has given? c. Vinvito, a chi Gianni I’ha connsegnato? The invitation to whom Gianni l’ha connsegnato? (134c) can be represented as in (135) showing the order of the three projections in Italian: (135) [TopicP Finvito [FocusP [a chi [AgrSP Gianni Pha consegnato]]] It is highly significant that no possibility of a preverbal subject is available in the Romanian equivalents of these examples, obviously because there is no SpecAgrSP below Focus where the subject could remain. Moreover, there is no Topic position below a Wh-Focus either, as confirmed by example (136b). The only preverbal position accessible to the subject is the Topic position to the left of the Wh-Focus (cf. (137)). (136) a. *Cui Ion i-a dat invitafia? To whom Ion has given the invitation? “Who did John give the invitation?” b. *Cui invitafia i-a dat-o Ion? To whom the invitation has given Ion? (137) a. Ion cui i-a dat invitafia? — Ion to whom has given the invitation? “Who did John give the invitation?” 45 The Double Subject Construction in Romanian b. Invitafia cui i-a dat-o Ion? The invitation to whom has given John? Concluding, we may assert that the facts we have examined prove that Romanian possesses only one preverbal quantifier position, namely, Focus, whereas, two positions ought to be available if there had a preverbal argumental subject position in Romanian. We conclude that there is no AgrSP before the verb in Romanian. Preverbal BQ subjects occupy the Focus position. 6. Conclusions 6.1. We have argued on the basis of morphologic and syntactic facts that Romanian clauses are (at least) MPs. 6.2. The functional structure of the MP includes (at least) the following ordered categories: MP > C1P > AgrSP > TP > AspP.VP 6.3. There are two argumental subject position in the Romanian clause, both of them postverbal: a) the thematic position which is SpecVP; b) the Case and pro-licensing position which is SpecAgrSP. 6.4. In finite clauses case is uniformly checked in SpecAgrSP, except for the DSC, where SpecT is also involved. Depending on the intrinsic semantic properties of the DP, it will check case overtly or covertly. 6.5. The preverbal subject is in the non-argumental position of left dislocated Topic or Focus, depending on the particular [+Topic] or [+Focus] feature it must check. The position that we have reached is similar to that expressed in Alexiadou (1994) about Modern Greek. She, too, claims that in SVO clauses, the Su cannot be in SpecAgrSP and that to license a preverbal subject an additional TopicP, and we should say, FocusP is needed. The Agr verbal features do not include the topic/focus feature, so such features are checked in special operator projections. We have confirmed the analysis in Dobrovie-Sorin with different arguments. 6.6. Romanian gives more prominence to the Topic Comment and Presupposition Focus articulation of the sentence than other Romance languages. Romanian again comes out as a more Topic oriented than Subject-oriented language, a feature which is in line with its higher degree of orality often discussed in the literature. 146 Alexandra Cornilescu REFERENCES 46 47 The Double Subject Construction in Romanian d., Philosophischen Fakultat, MAHAJAN, A. (1990) The A/A -bar Disitnction and Movement Theory, PhD Thesis MIT. MOTAPANYANE, V. (1989) La position du sujet dans une langue a Vordre SVO/ VSO, Universite de Geneve. OUHALLA, (1992) Functional Categories and Parametric Variation, London, New York, Routledge. ........ ,PANA-DINDELEGAN, G. (1974) Sintaxa transformational a a grupului verbal in limba roman a, Editura Academiei. PANA-DINDELEGAN, G. (1976) Sintaxa limbii romdne, Bucharest Univeristy Press, Bucharest. POLETTO, C. (1991) “The Diachronic Developments of Subject Clitics in North Eastern Italian Dialects, Working Papers in Linguistics, University of Venice. POLLOCK, J. Y. (1988) “Verb Movement, UG and the Structure of IP”, Linguistic Inquiry, 20, 365-424. PROGOVAC, L. (1993) “Long-Distance Reflexives”, Linguistic Inquiry, 24, 755-772. RAMCHAND, G.C. (1996) “Two Subject Positions in Scottish Gaelic: The Syntax Semantics Interface”, Natural Language Semantics, 4, 165-191. RAPOSO, E. & URIAGEREKA, J. (1995) “Two Types of Small Clauses (Toward a Syntax of Theme/ Rheme Relations)’', Cardinaletti, A. & Guasti, T. (eds.). Small Clauses, Academic Press, New York, London, 179-198. RIVERO, M.L. (1994) “Clause Structure and V-Movement in the Languages of the Balkane”, in Natural Language and Linguistic Theory, 12, 63-120. RIZZI, L. (1982) Issues in Italian Syntax, Foris Dordrecht. RIZZI, L. (1986) “On the Status of Subject Clitics in Romance”, Studies in Romance Linguistics, O. Jaeggli & C. Silva-Corvalan (eds.), Foris, Dordrecht. RIZZI, L. (1986a) “Null Objects in Italian and the Theory of pro”, Linguistic Inquiry, 17, 501-557. RIZZI, L. (1990) Relativized Minimality, MIT Press, Cambridge, Mass. RIZZI, L. (1994) “The Fine Structure of the Left Periphery”, ms. Universite dc Geneve. ROBERTS, I. (1991) Verbs and Diachronic Syntax, Kluwer, Dordrecht ROBERTS, I. (1994) “Word Order”, Gissl seminar notes. ROBERTS, I. (1994 a) “Long head Movement, Case and Agreement in Romance”, Horn stein, N. Sc Lighfoot, D. (eds.) Verb ^Movement. Cambridge University Press, 207-242. . press Bucharest. SAFIR, K “Syntactic Chains”, Cambridge University Press, Cambridge. - u*'-ar Buc-harestUmversity t ms SPORTICHE, D. (1988) “A Theory of Floating Quantifiers and Its Corollaries for Constituent seminar ^ Concepts 0f Modem Gra.£» ^ Romanain Non-t mt.e - Structure”. Linguistic Inquiry, 19, 425-449. CORNILBSCU^/v. v appear) Notes on ^ su 0f Chains, STARKE, M. (1995) “On the Format for Small Clauses”, Cardinaletti, A. & Guasti, T (eds.). CORNILESCU, A. guc|lgrest. ^ ic posuions and the Nature , Small Clauses, Academic Press, New York, London, 237-260. Universi^) ^ ^ Typoiogy °J STOWELL, T. “Subjects, Specifiers and X-bar Theory”, M. Baltin and A. Kroch (eds.), DBPM-Z, V. ( _ , f tat ion, Cambsidgc, panibndgc, Mass. i Alternative Conceptions of Phrase Structure. University of Chicago Press, 232-262. Press indefinites* .'sVil i Pre^, _^,>nain These de doctoiatji b ^ ^ j^omam^nSUNER, M. (1994) “V-Movement and the Licensing of Argumental WH-Phrases in Spanish '\S ^ Vvninxe da r' Natural Language and Linguistic Theory, 12, 336-378 »«ssivi2»tioURlArwpw k t - —.......... ALEXIADOU, A. (1994) Issues in the Syntax of Adverbes, Ph. Universitat Postdam. BAKER, M. (1988) Incorporation, University of Chicago Press, Chicago. BELLETTI, A. (1988) “The Case of Unaccusatives”, Linguistic Inquiry 19,1, 1-34. BELLETTI, A. (1990) Generalized Verb Movement, Rosenberg and Sellier, Turin. BORER, H. (1984) Parametric Syntax, Foris, Dordrecht. BORER, H. (1994) “The Projection of Arguments”, in Benedicto, E. & J. Runner (eds.), Functional Projections, University of Massachusetts, 1993. BURZIO, L. (1986) Italian Syntax, Reideh Dordrecht. BURZIO, L. (1994) “Case Uniformity'”, ms. Dept, of Cognitive Science, The Johns Hopkins University. BYCK, X (1937) “I/emploi affectif du pronom.personnel en roumain”, Bulletin linguistique, 15-33. CARDINALETTI, A. (1996) “Subjects and Clause Structure”, to appear in The New Comparative Syntax, Longman. CARDINALETTI, A. & L ROBERTS (1991) “Clause Structure and X-Second”, W. Chao & G. Harrocks (eds.,). Levels of Representation, Foris, Dordrecht. s A. & M. STARKE (1994) “The Typology of Structural ihfi T.OJ CARDINALETTI, A. 1 Grammatical Classes”. H. van Riemdisjk; . r.nnaua&s Typology, Mouton, Berlin. Three vol. 8. Languag CHOMSKY, N. (198 1). lecture, CHOMSKY, N. (1991) “Some Notes on Economy of fed.), Principles and Parameters in Comparative. -r.*.» Pro&ratn* the Mil urai Deficiency. On the Clitics in the Languages of Europe, '’gy y VO . . , _ es on Government and Binding, Foris, Dordrecht. ^otes on Economy of Derivation and Representation”, " /"*—MT'F Press. R. Freidh CHOMSKY, N. CINQUE, G. (19-', ^ CINQUE, G. Linguistics, lit CINQUE, G. (1993b) id Parameters in The Minimalist Program, »*•?.? of A-Bar Dependench of Tiuase and Comp* Yr, MIT Press Cambridge. Mas \ csl u I! ^eoi v iverst^v ot 'Venice. t4F unctiotsa! Struciui Mass, Working Papers Sentence and the DP\ *r DDESING, M. (1992) DOBRO VIE-8 ORIN, C DOBROVTE-SORIN, C Mouton de Gruyter. DOBROVTE-SORIN, C. (1997) “Impersonal , icj i,11- f\9%T) Vyt ta.-ce du roumain. — (1994) Thz, Syntax oj Romanian, it; uv uuwtv..- Zomparaiive Studies in Construction in Romance and 1 of Unergatives”, to appear _...0 GUERON, J. & HOEKSTRA, T. (1995) “The Temporal Interpretation oi Cardinaletti, A. & Guasti, T. (eds.). Small Clauses, Academic Press, New Yor DE HOOP, H. (1992) Case Configurationality and Noun Phrase Interpretation, PhD Thesis, Groninge ISAC, D. (1996) “Verb Movement in Romanian”, in Revue roumaine de linguistique. KAYNE, R. (1989) “Null Subjects and Clitic Climbing”, The Null Subject Parameter, ( Jaeggli & K. Safir (eds.), Kluwer Academic Press^ Dordrecht, 239-261. KAYNE, R. (1993) “Towards a Modular Theory of Auxiliary Selection”, Studia Linguistica, 47, 3-3 KAYNE, R. (1994) The Antisymmetry of Syntax, The MIT Press, Cambridge, Mass. LASNIK (1995) “Case and Epletives”, Linguistic Inquiry. ------^igurnentai WH-Phrases in Spanish”, Natural Language and Linguistic Theory, 12, 336-378. the passivizatioURIAGEREKA, J. (1994) “Aspects of Clitic Placement in Western Romance”, Linguistic Inquiry, 25, 79-123. l?^7^ngui$tic Inquiry* tation of PredicatioiiZUBIZARETTA, M.L. (1994) “The Grammatical Representation of Topic and Focus. 1U/ - rsc\c\ aThe Temporal Interpre New York. Implications for the Structure of the Clause”, Working Pavers of th* 7^ University of Bucharest