The Internal Structure of Absolute Small Clauses

The Interna1 Structure of Absolute Small Clauses Luis Mpez Cornell University This paper is an investigation into the functionai shucture of Absolute Small Clauses (ASCs), with data based on Spanish. ASCs are composed of a predicate and an NP and they are adjoined to a main clause without a mediating complementizer. It is shown that theu structure is not 'speciai' but a proper subset of main clause structure. ASCs have some interesting restrictions: (i) Individual Level (IL) predicates are disailowed. (ii) deep subjects are disailowed, and (iii) the order NP+predicate is disailowed. The analysis of (i)-(iii) relies on recent proposais regarding clause structure, in particular, (i) Diesing's Mapping Hypothesis and her structurai analysis of the IndividuallStage Level distinction and (ii) what Koizumi has termed the Split VP Hypothesis, according to which some functionai categories are generated between the lower and the upper VP shells. Additionaily, the question of how the case of the NP is checked is addressed, and it is suggested that the feature composition of the head Aspect in ASCs is minimaily different from that of main clauses. A discussion of Itaiian ASCs ensues and it is argued that they have the same structure as the Spanish ones. 1. Introduction The empirical scope of this paper is Spanish Absolute Small Clauses (henceforth ASCs). ASCs are adjuncts composed of a predicate -usually, but not necessarily, a past participleand a noun phrase, which is optional in many cases. In (la) the predicate that heads the ASC is a participle, in (Ib,c) it is an adjective, in (lc,d) it is an adverb (examples from Hernanz (1991)): Catalun Working Papers in Linguistics (Cat WPL) 4.1 '(1994): 45-92 Universitat Autbnoma de Barcelona (1) a. Desaparecidas las joyas, llamaron a la policia. disappeared the jewels called3pl to the police b. Limpias las mas, 10s soldados salieron a la calle. cleaned the weapons the soldiers went-out to the street c. Tenso el gesto, el gato nos observaba desde la silla. tense the gesture the cat us looked-at from the chair d. Asi las cosas, uno no sabe qué pensar. like-this the things one not knows what believe e. Lejos 10s buques, la poblacidn isleña pudo retornar a sus casas. far the bats the population of-island could return to their homes Here I focus mainly on participial ASCs, though the analysis presented extends to the other types. Using the examples in (2), I sketch the distinctive characteristics of this construction. ASCs are unaccusative or transitive: (2) a. Desaparecidas las joyas, llamaron a la policia. disappeared the jewels called-3pl to the police b. Rotos 10s vinculos familiares por la irresponsabilidad de 10s padres, 10s broken the bonds familiar by the irresponsibility of the parents the adolescentes se sumen en la depresi6n. adolescents SE sink into the depression In both cases, the NP can be phonetically empty. If so, it has to be controlled by some constituent in the main clause, usuaily the subject. I assume this empty category to bepro: (2) c. Sitiada proi durante varios meses, la ciudadi tuvo que rendirse. besieged for several months the city had that surrender (2) d. declarada prq en estado de emergencia por tercera vez en pocos meses, el declared in state of emergency for third time in few months the gobierno decidi6 enviar refuerzos militares a @a ciudad de A~~cuc~o]~ govement decided send reinforcements military to the city of Ayacucho When the predicate is transitive, it admits a by-phrase: (2) e. Destruido el litomi por la mancha de petr6le0, 10s pescadores emigraron. destroyed the shore by the spot of oi1 the fishermen emigrated The NP must be a 'deep' object, that is, the subject of an unaccusative verb or the object of a transitive verb. 'Deep' subjects are disallowed (Perlmutter (1978), Belletti (1981)): (2) f. *Trabajado Juan, se fue de paseo. worked Juan SE went-3sg of walk g. *Destruido el general, mir6 la ciudad con orgullo. destroyed the general looked-3sg the city with pride The order participle+NP is obligatory. The sequence NP+participle is ungrammatical: (2) h. *Los vínculos familiares rotos, 10s adolescentes se sumen en la depresi6n. the bonds familiar broken the adolescents SE sink into the depression , Finally, the NP receives nominative Case:' I use a pronominal NP to show the morphological Case difference. While ASCs with pronouns sound somewhat unnaturai, there is a definite contrast between (2i) and (2j). (2) i. Desmayado yo, la reunión no pudo comenzar. fainted I the meeting not could begin j. *Desmayado mi, la reunión no pudo comenzar. fainted me the meeting not could begin Notice that ASCs do not show any trace of a tense morpheme. Since current theory (see for instance Chomsky (1993)) links nominative Case to finite Tense, the availability of nominative in ASCs is a phenomenon that needs to be accounted for. Another characteristic of this type of constructions -noticed by Hernanz and by Stump (1985) for English, is that the participle must be a Stage Level (SL) predicate, it cannot be an Individual Level (IL) predicate (see Carlson (1977) for a description of the distinction): (3) . a. *Sabidas muchas matemgticas, Carlos pudo entrar en la universidad. known many mathematics Carlos could enter in the university b. *Hablado espaiiol, Susana consiguió un buen trabajo. spoken Spanish Susana got a good job In this paper I will argue for an analysis of ASCs that answers the following questions: Q1. what the functional structure of ASCs is; Q2. why ASCs must be SL predicates; 43. why the NP must be a 'deep' object; Q4. why the order V+NP is obligatory; Q5. how the NP receives nominative Case. Following Belletti's, De Miguel's and Hernanz's insights, I show that ASCs are composed of the lower part of an ordinary clause. Furthermore I claim that ASCs provide strong evidence for the view that some functional structure has to be projected between the two VP shells (in the sense of Larson (1988)), as has been argued by Travis (1991), Bowers (1992), Koizumi (1993), Zagona (1993) and Collins and Thráinsson (1993). Following Koizumi, I call this the split-VP hypothesis. If (4) represents an ordinary clause, ASCs consist of only the VE, the inner shell, and the functional projections associated with it (FP=functional projections): ASCs I call the functional projections projected by V2 'inner' functional categories, and the functional projections projected by Vl, 'outer' categories. The fundamental hypothesis that underlies my discussion is that the functional hierarchy is fixed by UG, variation limited to the feature composition of the functional categories (Borer (1983)). This has two consequences for an analysis of ASCs. First, the structure of Spanish ASCs has to be a proper subset of the grarnmar of Spanish clauses -i.e., the structure of ASCs cannot be 'speciai' with respect to ordinary clauses. Since structure is not subject to intralinguistic variation, 'subgrammars' (Addis (1993)) must be defined in other tems. I will argue that the feature composition of the functional category Aspect in ASCs is minimally different from that of main clauses. Secondly, ASCs in other languages should have the same structure, hence the same properties as ASCs in Spanish -except word order, which depends not on hierarchy but on the strength or richness of features, as generally assumed (Pollock (1989), and many others). In this respect, I believe that the analysis that I present here extends to ASCs in English and Italian. This paper is organized as follows: in section 2, I present the structure that I propose for ASCs, answering Q1. In section 3, I explain why ASCs must be SL predicates. In section 4, I show how the split VP hypothesis accounts for the absence of externa1 arguments in ASCs. In section 5, I discuss the word order problem. In section 6, I present an account for the licensing of the NP, with which I answer Q5. Section 7 summarizes my results so far. In section 8, I discuss Italian ASCs and finally, in section 9, I argue contra Belletti (1990), De Miguel (1990) and Hernanz (1991) that there is no CP in ASCs, in keeping with the hypothesis that the structure of ASCs is properly contained in the structure of ordinary clauses. 2. The Structure of ASCs 2.1. No 'Outer' TP in ASCs Belletti (1990), De Miguel(1990) and Hernanz (1991) argue that there is no Tense Phrase in ASCs. I agree that there is no 'outer' Tense in ASCs. Participles in Spanish do not support any tense morphemes: when participles are constituents of full clauses, tense features need the support of an auxiliary. ASCs are ungrammatical with an auxiliary with finite tense, as shown in (5). It is possible to have a gerundive auxiliary with a participle forming a sentential adjunct, as in (Sb), but then we obtain a different type of construction, as we shall see in section 2.5. (5) a. *Ha disuelto el sulfuro en agua, pierde sus propiedades corrosivas. has dissolved the sulphur in water loses its properties corrossive b. Habiendo disuelto el sulfuro en agua, el quimico sonri6 a la audiencia. having dissolved the sulphur in water the chemist smiled to the audience Moreover, as I mentioned at the beginning of this paper, adjectives and adverbs can also head an ASC, and neither of these categories are assumed to project a TP. On the other hand, De Miguel (1992) claims that TP is the highest node in ASCs. This TP is described as an abstract morpheme but as a syntactically relevant category. Her motivation is that ASCs contain the temporal information that the event described in the ASC is anterior with respect to the event taking place in the main clause. She thus characterizes the head of TP as containing the feature [+anterior]. This feature would be responsible for the variety of adverbial functions vis-a-vis the main clause that an ASCs may have, since all of them are to be understood as temporally anterior. For instance, in (6a) the ASC is to be read as causal, whereas in (6b) it is conditional (examples from De Migue1(1992:65), who attributes the first one to Alcina and Blecua and the second one to iázaro Carreter). Both causality and conditionality are thus encapsulated in the feature [+anterior]: (6) a. Dada su poca seriedad, optamos por no renovarle el contrato. given her little seriousness choose-lpl for not renew3sgDat the contract b. Gastada esa agua, no podremos ni beber. wasted that water not can-FUT-lpl neither drink Since the TP that De Miguel argues for cannot be identified with the overt Tense of main clauses that has a ckar morphological correlate, I suggest that it is an 'inner' Tense, not an 'outer' one. The claim that there is a functional node Tense interna1 to the VP has been put fonvard by Travis (1991) and Collins and Thr6insson (1993). Thus, De Miguel's argument for a TP in ASCs would provide further evidence for this hypothesis. 2.2. No NegP in ASCs Belletti (1990), De Miguel (1990)and Hernanz (1991) correctly point out that there is no NegP in ASCs either: (7) a. * No disuelto el sulfuro en agua ... not dissolved the sulphur in water b. * No disuelto, el sulfuro ... not dissolved the sulphur Following these authors, I attribute the ungrammaticality of (7) to the absence of a NegP. De Miguel (1992) claims that the prohibition against negation in ASCs is of a semantic, nota syntactic nature, derived from its temporal nature. However, compare the participial absolute of (8a) with the gerundive of (8b): (8) a. *No disuelto el sulfuro de la manera adecuada, Juan hubo de not dissolved the sulphur of the way right Juan had of posponer la conclusi6n del experimento hasta el dia siguiente. postpone the conclusion of-the experiment - until the day following b. No habiendo disuelto el sulfuro de la manera adecuada, Juan hubo de not having dissolved the sulphur of the way right Juan had to posponer la conclusi6n del experimento hasta el dia siguiente. postpone the conclusion of-the experiment until the day following (8a) is ungrammatical, in contrast with the grammatical (8b). I cannot find any semantic feature that distinguishes (8a) from (8b) and that can be held responsible for the ungrammaticality of (8a).2 I conclude that lack of the necessary structure causes the ungrammaticality of (8a). Notice that the positive counterpart of (8b) has a different interpretation: (i) Habiendo disuelto el sulfuro, Juan prosigui6 con el experimento. having dissolved the sulphur Juan went-on with the experiment 2.3. No CP in ASCs Belletti (1990), De Miguel (1990) and Hernanz (1991) claim that there is a CP in ASCs. Since this point has numerous ramifications, I postpone it to section 9, where the hypothesis of a CP in ASCs is discussed in detail. It will be shown that there is no convincing evidence for a CP in ASCs. 2.4. AgrOP Participles and adjectives in ASCs agree with the NP. This is shown in (9), where the vowel suffix -a indicates feminine and -s indicates plural: (9) a. Acabada la cena, Jesús se dirigi6 a 10s ap6stoles. finished-fem the-fem dinner Jesús SE addressed to the apostles 'Dinner finished, Jesus addressed the apostles.' b. Acabadas las tareas, 10s niños salieron a la calle. finished-fem-pl the-fem-pl homework the children went-out to the street 'Homework finished, the children went out.' Following Kayne's (1985, 1989) analysis for participial agreement in French, which involves an AgrP, I conclude that the participle of ASCs projects an AgrP. Now we are in a position to answer Q1: the functional structure of ASCs reaches as far as AgrOP and maybe to some TP. ~ - Whereas (i) can have both a temporal and a causal interpretation, (8b) can only have a causal one (see Rigau (1991) for a similar phenomenon with Catalan infinitives, also cited in De Miguel (1992)). It could be argued that negation makes rhe adjunct clause in (8b) obligatorily causal, and that ASCs do not have this interpretation -however, (6a) shows that ASCs can have a causal interpretation. The question is why (8a) cannot be grammatical with a causal interpretation. being [+perfective]. In participial absolutes, Asp is morphologically realized on the participle. See also Carstens and Kinyalolo (1989) on AspP and Travis (1991) and Zagona (1993) on the existence of an 'inner' aspect. Finally, see Collins and Thráinsson (1993) on an 'inner' Tense, connected, as I mentioned in section 2.1, with De Miguel's (1992) proposal for a TP in ASCs. The structure that I propose for ASCs is the one contained in the box of (16): VP 1 n NP V' ASC v A T AgrOP n s pec A gr' n Agr AspP n ASP VP 2 n NP V' I v - (16) gives us a direct answer to 43: 'deep' subjects are generated in the upper VP1 shell and this is not presentin ASCs. Also, it accounts for the licensing of the by-phrase and hence the passive appearance that transitive ASCs have, as can be seen in example (Ze), repeated here: (2) e. Destruido el litoral por la mancha de petróleo, 10s pescadores emigraron. destroyed the shore by the spot of oi1 the fishermen emigrated The external 0-role in an ordinary clause is assigned by V1, the upper verb. In a passive construction, the passive morphology suppresses the external 0-role (Baker et al. (1989), among others), which in our tems means that V1 loses its ability to license it. Then the external 0-role may reappear as an oblique. In ASCs the upper V1 is not inserted, so the external 0-role is not licensed by a predicate. It follows that ASCs have no 'deep' subjects and, as a consequence, the agent role can reappear as an oblique. Since the external argument is not generated, an ASC is effectively identical to a truncated passive construction, the result of suppressing an external argument and of not generating it in the structure being the same. Addis (1993) and Franco (1993) argue that Spanish ASCs are passive constructions. Given the analysis in (16), saying that ASCs are passives is redundant because it would entail suppressing something that was never there in the first place.' In order to confirm that (16) is needed to account for the absence of 'deep' subjects in ASCs, let us try to see if 43 could be answered from a structure in which AgrOP is projected above VP1, as in (12a). Let us assume (12a') as the structure of the clause. In (12a') there is no functional structure between the two VP shells. The curly brackets represent the structure of ASCs as argued for by Belletti (1990) and Hemanz (1991) (I omit any reference to a CP, TP or AspP, to simplify): 7 A passive formed with a flip verb (i.e. a verb in which the patient gets nominative Case and the experiencer gets dative) is iU-formed, as shown in (i). As pointed out by an anonymous reviewer, ASCs formed with a flip verb are also ill formed, as shown in (ii): (i) *Lm claveles son gustados por Juan. the carnations are liked by Juan (ii) *Gustados 10s claveles por Juan ... liked the carnations by Juan I do not have an explanation for the ungrammaticaliity of (i) and (ii). In any case, the fact that flip verbs are unacceptable in both types of constructions does not entail that one is a sub-type of the other. As I explain in the main text, suppressing a 8-role and not generating a 0-role are distinct processes that give rise to constructions with the same propehes. Assuming the structure of ASCs represented in (12a1), we can predict that the subjects of transitive verbs are not licensed because there is only one Case position - [Spec,AgrOP]- and two arguments. [Spec,AgrOP] is taken by the object, and the subject does not have a place where it checks its case and $-features and cannot be licensed. However, (12a') does not rule out unergative predicates: since they do not have an object, nothing prevents their only argument from being licensed in [Spec,AgrOP]. Nothing explains the ungrammaticality of (2f). As Collins (p.c.) points out, Kayne's (1993) proposal could be relevant at this point. He suggests that unergative predicates contain an empty complement that raises to [Spec,AgrOP] to check features. Kayne proposes this empty object. to make sure that the subject of an unergative verb does not stop at [Spec,AgrOP] and triggers agreement on the participle in a typical perfective construction in Romance. With the structure proposed in (16) this problem does not arise. The subject of an unergative predicate is generated in [Spec,VPl], that is, at a position higher than [Spec,AgrOP], therefore the subject of an unergative cannot check features in [Spec,AgrOP] unless it lowers, a movement disallowed for principled reasons. Moreover, the claim that languages like English and French can license an empty object needs elaboration, since so far all the available evidence has pointed out in the opposite direction (see, for instance, Rizzi (1986)). I conclude therefore, that Kayne's idea is not a viable alternative to the analysis defended in this paper.8 8 Incidentaily, Kayne's proposal wntradicts the analysis of unergatives developed by Hale and Keyser (1991). Hale and Keyser propose ihat unergatives contain an inwrporated object which, following cment assuinptions on ihe nature of incorporation, entails that it mot check features in [Spec,A@P]. Belletti (1990) assumes a structure like (12a') and, acknowledging the problem presented, she offers a solution. However, her solution is based on assumptions different from mine. As I mentioned above, she argues that there is a CP dominating AgrP in ASCs. Further, she claims that C contains a Case feature. When the unergative participle, also containing a Case feature, raises to C (a movement that she argues is necessary to license the NP) a Case conflict arises. The assumption of a CP violates one of the basic hypotheses of this paper, namely, the notion that the structure of ASCs is a subset of the structure of ordinary clauses. More on this in section 8. Moreover, Belletti must assume that unergative verbs have an accusative Case to assign. In current conceptions of Case Theory -as in Chomsky (1993)-, Case is regarded as a feature that must be checked. Unchecked features give rise to crashing derivations. Assuming this, if unergative verbs have a Case feature that is not checked, then the derivations of unergative verbs will never converge. 5. Word Order 5.1. English and French In this section I tum to 44, the word order issue. As we saw in (2h), repeated here, the order participle+NP is obligatory. This contrasts with the situation in other languages: French and English participial absolutes must have the order NP + participle, as can be seen in (17): (2) h. *Los vínculos familiares rotos, 10s adolescentes se sumen en la depresi6n. the bonds familiar broken the adolescents SE sink into the depression (17) a. All things considered, the situation is not that bad. b. Dinner finished, we left for the opera. (from Kortmann (1991)) c. Ceci dit, la rkunion a pu commencer. this said the meeting has could start (17) d. Le chat parti, les souris dancent. the cat gone the mice dance A good answer to Q4 should try to encompass (2h) and (17). Let us first consider the English and French cases. Main clauses in English and French have obligatory preverbal subjects. Following Chomsky (1993), this means that their 'outer' Tense projects strong N-features. As we saw in (14) and (IS), evidence seems to indicate that the 'inner' Tense also has strong N-features that attract an object NP by spell-out in English. Therefore, the order NP+participle in English ASCs is the result of overt movement of the NP to [Spec,TP]. By hypothesis, the same analysis would extend to French. The logic of this argument leads to the conclusion that the 'inner' T in Spanish has weak N-features and that is why it does not raise. 5.2. Spanish: the ThemelRheme Distinction As I discuss in more detail in Section 9, Belletti, De Miguel and Hemanz propose that the NP moves out of the VP to [Spec,AgrP], the participle moving to a position higher than AgrP.9 The order participle+NP can be obtained with several different scenarios, given the stmcture proposed here. The NP could remain in situ and the partic~ple could raise to either Asp, Agr or T. The NP could make a short movement to [Spec,AspP] and the participle still has two possible surface positions. Finally, the scenario preferred by the cited authors IS one rn which the NP is in [Spec,AgrOP] and the participle is in a higher position. I do not have anything to add conceming the surface position of the participle. Let us assume that the predicate that heads an ASC has an extra feature, call it [absolute], that triggers overt A head T for De Miguel, C for Hemanz and Belletti. 64 movement to Asp, where the feature is checked. At this point, this feature is a mere stipulation, but in section 6. I show that it is necessary for a full account of ASCs. As for the NP, I present a piece of evidence that suggests that the NP remains in situ in Spanish. In (18a), we can see that an NP subject in an ordinary clause can strand a quantifier when it raises and in (18b) we can see that this is possible even when the NP raises to a position to the right of the verb -[Spec,TP], probably. In ASCs, even though the subject can have a universal quantifier, as in (I%), this quantifier cannot be stranded, as in (18d). The conclusion is that the NP in an ASC does not raise out of its base generated position: (18) a. Los hombres decidieron todos quedarse en casa. the men decided-3pl all stay-SE at home b. Saludaron 10s embajadores todos ai hijo del monarca. greeted3pl the ambassadors all to-the son of-the monarch c. Saludados todos 10s embajadores, el hijo del monarca sonri6 a las cámaras. greeted ai1 the ambassadors the son of-the monarch smiled to the cameras d. *Saludados 10s embajadores todos, el hijo del monarca sonri6 a las cámaras. greeted the ambassadors all the son of-the monarch smiled to the cameras Let us take a look at the position of subjects in.ordinary clauses. Ordinary clause subjects in Spanish do not need to raise to [Spec,AgrSP] in the overt syntax. Consider the following grammatical sentences: (19) a. Vino Pedro. came Pedro 'Pedro came.' b. Telefone6 Pedro. called Pedro 'Pedro cailed.' (19) c. iCompr6 Pedro las patatas? bought Pedro the potatoes 'Did Pedro buy the potatoes?' d. Las patatas las compr6 Pedro. the potatoes them bought Pedro 'Pedro bought the potatoes.' e. Fueron atacados 10s cargueros. were attacked the cargos The cargos were attacked.' It seems that Spanish subjects do not raise into [Spec,AgrSP] overtly. I propose that Spanish AgrS has weak N-features and subjects do not need to raise until LF to check their morphological features. I suggest that Spanish AgrO also has weak N-features, so the NP generated in [Spec,VP2] does not need to raise to [Spec,AgrOP] until LF. This hypothesis would explain why the NP of an ASC can stay in situ. However, the subjects of ordinary clauses can always raise. What is left to explain is: (i) what triggers NP movement in ordinary clauses, (ii) why the subjects of ASCs must stay in situ. Why do Spanish subjects raise? One possibility could be to assume that [Spec,AgrSP] optionally projects strong features in Spanish.10 An altemative is suggested in Contreras (1976) and Suñer (1982). They show how the SV order makes the subject theme, whereas the VS makes it rheme. For the purposes of this paper, I adopt the notion of Suñer (1982:4): rheme is the part of a sentence that is asserted, theme is the part that falls out of the scope of assertion. l0 Chomsky (1993) proposes this for Standard Arabic. However, the case of Arabic is different from Spanish: when the subject is in situ in hbic, the verb does show poorer morphology. Let us see it with an example. In (20) there are two possible conversations. In (20a), the subject is preverbal, hence theme, whereas the verb is rheme. This means that the issue that is talked about is what Juan is doing, and we are asserting that he is coming. In (20b), the subject is rheme, so the issue is whether Juan or someone else is coming: (20) a. A: Creo que Juan viene. 'I think that Juan is coming.' B: No, se queda. 'No, he's staying.' a. B': *No, Pedro. 'No, Pedro.' b. A: Creo que viene Juan 'I think that Juan is coming.' B: *No, se queda. 'No, he's staying.' B': No, Pedro. 'No, Pedro.' I suggest that overt NP raising in Spanish ordinary clauses is triggered by this semantic reason: NP subjects raise in order to make the distinction between theme and rheme: theme being the position(s) external to the VP, rheme the position(s) internal to the VP.ll Tentatively, I suggest including the themelrheme distinction in the Discourse Representation Theory tradition stemming from Heim (1982) and the Mapping Hypothesis in Diesing (1992). Given a logical representation of the forn Op, [P(x)] Q(x), the restrictive clause is the non- - - In the most recent version of the Principies and Parameters Theory, that of Chomsky (1993). movement is only triggered by morphological feame checking. Semantic or discourse factors are not supposed to play a role. This entails that my proposal in this section does not fit into the Minimalist Program as is. However, it seems to me that Chomksy's strong position vvill have to be revised in order to reintroduce at least Quantifier Raising back into the thwry. Attempting to do this is out of the scope of this paper. asserted and the nuclear scope is the asserted part.12 Syntactically, thematic subjects will show up in the Spec of a high functional category, rhematic ones inside the VP projection. My claim concerning ASCs is that, being too small for a tripartite structure, they cannot license any of the operators (as for instance the generic operator) that take scope in a quantification structure. This has the following set of predictions: (i) IL predicates are not licensed in ASCs (see section 2), (ii) generic subjects are not licensed in ASCs, (iii) lacking a motivation to raise, the subjects of ASCs should stay in situ. In the next paragraphs I look at the interpretations of NPs in ordinary clauses in preverbal and postverbal position and in ASCs. I show how all those predictions hold. Let us first consider the behavior of indefinite subjects in ordinary and small clauses.13 In ordinary clauses, preverbal indefinites get a generic reading, and they cannot get an l2 As h4olly Diesing points out, not all discourse phenomena can be represented in a quantification structure. There is, however, some overlap, which I exploit here. l3 The discussion concerning (21). (22). (24) and (25) should be considered a rough sketch where some variables are left out. An important variable, pointed out by the reviewer is perfectivity. indefinite NPs cannot be generic in a perfect tense, even if placed before the verb: (i) IJn espaiiol Ueg6 con retraso. a Spaniard anived with late The reviewer suggests that the ahsence of generic readings in ASCs is tied to perfectivity -it has been noted that ASCs must have the perfective aspect (see De hliguel (1990, 1992). Hemanz (1991)). However, in a full dause, an NP with a definite determiner can have a generic reading even in the perfect: (ii) Desde tiempos inmemoriales 10s espaíioles han, llegado tarde a sus citas. from times immemorial ihe Spaniards have anived late to their appointments However, in the context of an ASC. NPs with a definite determiner cannot have a generic reading (see (26)). So, the fact remains that generic readings are not available in ASCs and that this is nota by product of perfectivity. It seems that a variety of factors infiuence generic and existential readings in main clauses, some of which I have to abstract away from unti1 they are better understood. 68 existential one, as seen in (21). On the other hand, postverbal indefinites get an existential and not a generic reading, as seen in (22):14 (21) Un espaiiol llega con retraso. a Spaniard arrives with late *& x a Spaniard A x is late Gen, [x is a Spaniard] x is late (22) Llega con retraso un espaiiol. arrives with late a Spaniard & x is a Spaniard A x is late *Gen, [x is a Spaniard] x is late Indefinite subjects in ASCs must also be existential. In example (23) I do not provide the main clause so that the reader makes sure that I arn not favoring one interpretation over the other: (23) Llegado con retraso un espaiiol, ... arrived with late a Spaniard & x is a Spaniard A x is late *Gen, [x is late] x is late Exarnples (24) through (26) show the same point: a plural definite NP receives a generic reading in preverbal position, but not in postverbal position (where the reading they get is more like a universal quantifier): l4 Regardless of whether there are other constituents between the verb and the subject. I assume that when the VOS order obtains, the subjectis right adjoined to the VP, as in Bonet (1989). Rhematic readings are not related to one position in particular, the only proviso being that they be located low enough to avoid being mapped onto the restrictive clause. It seems that the Case that the predicate of a ASC checks is some kind of default Case. This could ceriainly be argued for Latin, since ablative is a synchretic Case that embraces locatives, oblique agentives and instrumentals. This type of solution fits into the program set out in the introduction of this paper. Though the hierarchy of functional categories does not vary, the features of the functional categories do vary, giving rise to crosslinguistic variation and, as I have shown in this section, also intralinguistic. The problem of at least this particular type of 'sub-grammar' (Harris (1968)) -i.e. the fact that certain constructions seem to have a grammar that intersects instead of being properly contained in the grammar of the language as a wholeis resolved by positing feature variation rather than structure variation. We still have one more wrinkle to iron out. Transitive verbs are assumed to have an accusative Case to check. If verbs are generated with this Case, then ASCs have two Cases available: nominative Case of [absolute] in Asp and the accusative Case of the verb. This problem becomes more apparent than real if we incorporate Collins and Thr&inssonts (1993) Case theory. Drawing on Icelandic data, they conclude that V2 does not have a Case feature, and that accusative Case is actually borne by V1, which itself projects an AgrOP -thus, there are two AgrOPs in the clause, one projected by V1, where Case is checked, and the other is projected by V2, a place where +features are checked, but no case because V2 has no Case to check. ~herefore, the participle of ASCs, which is a V2, never has any accusative Case to l6 AS the reviewer points out, something should be said about what I termed AFCs: (i) Habiendo Juan disuelto el sulfuro, el experimento pudo realizarse en menos de dos horas. having Juan dissolved the sulphur the experiment could accomplish-SE in less of two hours The questions should be asked concerning (i): (a) does the gerundive in (i) have an [absolute] feature?. @) does the subject of the AFC check Case against this feature [absolute]? I do not have any evidence in Spanish to answer any of those questions. However, consideration of crosslingustic &ia could give us a clue. Latin and Classical Greek have AFCs with active participles -at ali effects, equivalent to AFCs like (i) because they admit two arguments. In these constructions the subject NP exhibits ablative Case in Latin, genitive in Greek, and these 7. Mid-Way Conclusions Now we are in a position to answer Q1-5. The remarkable properties of ASCs have been shown to be derived from their 'smallness': their functional structure includes AgrOP and excludes any other higher categories, which has the following consequences: (i) the subjects of IL predicates lack a Spec position where they can be licensed, following Diesing's (1992) theory; (ii) since the upper verbal head is not generated, 'deep' subjects cannot be generated either; (iii) the word order participle+NP is obligatory because the order NP+predicate in Spanish is the result of raising a thematic subject to a restrictive clause in Heim's (1982) sense and this is unavailable in ASCs. Finally, the NP can check Case with an extra feature in Asp, absent in ordinary clauses. In the next section I test the hypothesis that ASCs should have the same structure crosslinguistically against the Italian data presented in Belletti (1990). Though the results obtained in this part of my research are tentative, I believe that this hypothesis gives us better empirical predictions than assuming a different structure -as argued for, in a Relational Grammar framework, by Addis (1993). are the Cases that the subject of an ASC gets. On the other hand, the object gets accusative Case. (u) is a Latin example (from Allen and Greenough (1991)) and (ui) is Greek (from Smyth (1984)): (u) E0 imperi-um tenente. eventum timeo. he-Abl-sg power-Acc-sg holding-Abl-sg outcome fem-lsg With him in power, I fear the outcome.' (iii) Tosaut-a eipont-o:n t-o:n platai-o:n, Arkhidam-os so-many-things-Acc-p1 having-said-Gen-pl the-Gen-pl plataean-Gen-pl Archidamus-Nom-sg hupolab-o:n eipen. answering-Nom-sg said 'Having the Plataeans said so many things, Archidamus replied.' Since these subjects do not get the expected nominative of the subject of a main clause but the Case of the NP of an ASC, it seems plausible to assume that the same feature that checks ablative or genitive in ASCs checks ablative or genitive in AFCs. Mutatis mutandis, it could be argued that in (i), and in general in AFCs in any language, the subjectis licensed by the feature [absolute]. 8. Italian Transitive ASCs Apparently, Italian unaccusative ASCs work the same way the Spanish do. (34) is an exarnple: (34) Amvato io, la reunione 6 cominciaia. amved I the meeting is siarted However, transitive ASCs differ significantly. In the first place, they do not have the 'passive look' that transitive ASCs have in Spanish: the NP receives accusative Case, as we can see in (35) and by-phrases are disallowed, as can be seen in (36) (from Belletti (1990)): (35) Salutata mel*io, si t accorto che c'era molta altra gente. greeted me I SI is realized that there-were many other people (36) . Salutata me (*da Gianni), tutti uscirono dalla saia. greeted me by Gianni everybody left of-the room As I mentioned above, it is the case that [absolute] checks different cases crosslinguistically, the variation being more or less unprediciable. Belletti argues that transitive ASCs are control shuctures, i.e., apart from the overt NP there is also another argument PRO. According to the theory of clause structure laid out here, this extra argument would have to be generated in the Spec of VP1. Though it is theoretically possible that Italian transitive ASCs had more structure than the Spanish ones (Addis (I=)), I argue here that they do not. Let us review and critique Belletti's arguments for PRO. First, according to Belletti, this PRO has to be controlled by the subject of the main clause. The ungrammaticality of (37a) contrasts with (37b) because in (37a) the PRO is not controlled: (37) a. *Chiamato i1 taxi, smise de piovere. called the taxi stopped3sg of rain b. PRQ chiamato i1 taxi, Mariai usci. called the taxi Maria went-out However, Rosen (1988) shows that control in Italian ASCs can be purely pragmatic. In (38a), it is probably the judge who absolved the defendant, so a case could be made for subject control, but in (38b) the judge or jury is absent from the linguistic context: (38) a. Assolto I'imputato, i1 giudice se n'andb a pranzo. absolved the-defendant the judge SE NE-went to lunch b. Assolto l'imputato, scopiarono gli applausi. absolved the-defendant burst-out-3pl the applauses (38) weakens Belletti's argument because pragmatic control does not require a PRO subject. (39) is the Spanish equivalent of (37). In Spanish, however, the person who calls the taxi is not necessarily Maria -it could be the hotel porter, for instance: (39) a. *Llamado el taxi, par6 de Ilover. called the taxi stopped3sg of rain b. [e]i/j llamado el taxi, Mm'ai sali6. called the taxi Maria left In Italian, like in Spanish, therefore, transitive ASCs are not controlled: the construction allows for an arbitrary subject to perform the action -the mirror image of a verb like eat which, in its unergative usage, allows for any arbitrary thing to be eaten. As for the unacceptability of (37a). I believe it is due to a lack of obvious semantic connection between the main and the absolute clause and not to any grammatical reason. Belletti's second argument is that the overt NP can be a reflexive: (40) a. Elogiata solo se stessa, Maria restb del tutto isolata. praised only herself Maria remained completely isolated b. Lavatasi le mani, Maria 2 uscita. washed-herself the hands Maria is left She claims there has to be a PRO argument in the ASC that binds the reflexive. However, this is not the only analysis available. If there is no 'deep' subject, ASCs do not contain an accessible SUBJECT, therefore the binding domain of the anaphor is the main clause and not the ASC. My Italian informants certainly agree that se stessa ('herself) has to be coreferent with Maria, which confirms my hypothesis.17. Thus my claim is that we do not need to posit a PRO in ASCs any more than we need it in the adjunct phrase near each other in the sentence Near each other they saw snakes (thanks to Wayne Harbert (p.c.) for suggesting this exarnple to me). On the other hand, the claim that there is a PRO in 'deep' subject position raises an empirical problem: why can't there be unergative ASCs? The subject of the unergative could be generated as PRO and a sentence like (41) should be grammatical. But (41) is ungrammatical in Italian, as it is in Spanish: (41) *PRO, telefonato, Giannii andb alltappuntamento. telephoned Gianni went to-the-appointment Therefore, the reasons adduced by Belletti to have a PRO in transitive ASCs do not seem so convincing. On the other hand, it seems that there are good reasons not to have a PRO, and l7 Italian ASCs are therefore transparent to externa1 binders, whereas the Spanish ones are not. I consider this a consequence of the fact that in Spanish absolutes the NP bears nominative Case and nominative anaphors are disaiiowed. 80 stick to our analysis of ASCs without a VP1. It seems, therefore, that my hypothesis that the structure of absolute clauses does not vary crosslinguistically can be maintained. 9. The Problem of a CP Node in ASCs 9.1 Belletti's Arguments for a CP in ASCs In the remainder of this paper, I will discuss Hernanz's (for Spanish) and Belletti's (for Italian) claims that the AgrP is dominated by a CP. If Hemanz's and Belletti's claim were to hold, it would seriously threaten my hypothesis that the structure of ASCs is properly contained in the grammar of ordinary clauses. The surface structure that they offer for ASCs is the following: I NP Agr' n AspP n i "i' A There are three points that need to be discussed with respect to (42): 1 1. Whether there is direct evidence for a CP in ASCs. This will occupy the rest of this subsection. 2. Whether predicate raising can provide indirect evidence for a CP in ASCs. This will occupy subsections 9.2 and 9.3. 3. Whether NP movement provides indirect evidence for a CP in ASCs. This will occupy subsection 9.4. Belletti provides two types of evidence for a CP in Italian ASCs. First, she claims that the free relative quanto ('what') can introduce an ASC, as in (43): (43) Quanto comunicatoci, ... what communicated-us Under current assumptions on the licensing of wh-words, the most likely analysis for (43) should include a [Spec,CP] that would house quanto. However, quanto communicatoci is an infinitival relative and cannot be used as an adjunct.18 (44) is an example of its use: (44) Quanto comunicatoci ci ha disturbato molto. what communicated-us us has disturbed a-lot Belletti's second argument is that Italian ASCs can be introduced by anche se ('even if') and benchk ('although'), which she takes to be complementizers. This entails that there must be a C where they can be generated. In (45) are two examples: (45) a. BenchC partita da sola, Maria si diverti. although left by alone Maria SI had-fun b. Anche se amvata in ritardo, Maria non si scusb. even if amved in late Maria not SI apologize However, both benché and anche se may introduce other types of phrases in which it is not assumed there to be a CP:l9 (46) a. Questo t spiegato in una maniera detagliata bench6 [Ap non chiara]. This is explained in a detailed though not clear manner.' b. Questo ragazzo, anche se stupido, reuscid a conseguire la laurea. This young man, even if stupid, will manage to obtain the BA.' The examples in (46) show that there is not a one-to-one correspondence between these words and the presence of a C position, therefore they do not provide a strong reason to assume that there is a CP in Italian ASCs. A careful analysis of anche se ('even if) and benché ('although') is out of the scope of this paper. Pending further research, I suggest they could be analyzed as adjoined elements. 9.2. Does the Participle Raise into C? Hernanz and Belletti agree that the participle raises to C, though their rationale for the movement differs. If their claim were demonstrated, then we woul-d have indirect evidence that there is a CP in ASCs. However, I will argue against their analyses. Hernanz shows that there cannot be an ASC as complement of a subordinating preposition like por ('for'), para ('in order to, for'), sin ('without'), in contrast to the very similar infinitivals: (47) a. Sin comprar comida, no te atrevas a venir. b. *Sin comprada comida, no te atrevas a venir. 'Without buying food, do not dare come.' l9 Thanks to Carol Rosen for providing these examples. (47) c. Para comprar comida, ve al supermercado. d. *Para comprada comida, ve al supermercado. 'In order to buy food, go to the supermarket.' e. Por hacer travesuras, te quedas castigado. f. *Por hechas travesuras, te quedas castigado. 'For doing mischief, you will be punished.' Her argument is that, assuming that these prepositions occupy the CO position, the reason why they are ungrammatical in ASCs is because the participle raises into CO, therefore two elements are competing for the same position. However, under the same asumptions, it could alternatively be suggested that if there is no C in ASCs, there cannot be subordinating prepositions either, which would neatly account for the ungrammaticality of (47b,d,f). On the other hand, it is well known that in Romance prepositions appear to select for CPs rather than occupy the Coposition: (48) a. No s& c6mo rechazarlo sin que parezca grosero. 'I don't know how to reject him, without that it seems rude.' b. No s& c6mo disimularlo para que no parezca grosero. 'I don't know how to hide it for that it doesn't seem rude.' Moreover, there may be other reasons why (47b,d,f) are ungrammatical. Let us consider the above prepositions a bit more carefully. Sin ('without') can take NPs as complements, but not adjectives or adverbs: (49) a. Sin la comida, no te atrevas a venir. 'Without the food, do not dare come.' b. *Sin cansado, no te atrevas a venir. without tired not dare to come (49) c. *Sin rápidamente, no te atrevas a venir. without quickly not dare to come It seems that sin requires an NP complement. I claim that the infinitival in (47a) contains nominal features that make it, in some intuitive sense, like a noun. Since ASCs are not nominal but adjectival or adverbial in nature, the source for the ungrammaticality of (47b) is a violation of the selectional properties of the preposition sin. Let us take a look at por and para (both meaning 'for'). Selectional properties are again crucial to account for the ungrammaticality of (47d) and (47f). Por and para may take adjectives as complements, but only if they form IL predicates, never if they form SL predicates: (9) a. Para tonto, mi hermano. for silly my brother 'My brother is an example of a silly person.' b. Por tonto perdiste el tren. for silly missed-2sg the train 'Because you were silly, you missed the train.' c. *Para furioso, mi hermano. for angry my brother 'My brother is an example of an angry person.' d. *Por furioso, te castigaremos. for angry youpunish-FUT-lpl 'Because you were angry, we will punish you.' It seems that the generalization that por and para cannot select for SL predicates is clear. Since ASCs must be SL predicates (see (3) and discussion in section 2.4.), it follows that they Pesetsky, D. (1989) 'Language Particular Processes and the Earliness Principle', ms. MIT. Pollock, J.-Y. (1989) 'Verb Movement, Universal Grammar and the Structure of IP', Linguistic Inquiry 20,365-424. Postal, P. (1974) On Razking, The MIT Press, Cambridge. Rigau, G. (1991) 'La legitimaci6 de les construccions temporals d'infinitiu', ms. Universitat Autbnoma de Barcelona. Rizzi, L. (1986) 'Nul1 Objects in Italian and the Theory of pro', Linguistic Inquiry 17, 501557. Rosen, C. (1988) The Relational Structure of Reflexive Clauses: Evidence fiom Italian, Garland, New York. Smyth, H. (1984) Greek Grammar, Harvard University Press, Cambridge. Sportiche, D. (1988) 'A Theory of Floating Quantifiers and Its Corollaries for Constituent Structure', Linguistic Inquiry 19,425-450. Stowell, T. (1981) Origins of Phrase Structure, PhD dissertation, MIT. Stump, T. (1985) The Semantic Variability of Absolute Constructions, Reidel, Dordrecht. Suñer, M. (1982) Syntax and Semantics of Spanish Presentational Sentence-Types, Georgetown University Press, Georgetown. Travis, L. (1991) 'Derived Objects, Inner Aspect and the Structure of VP', paper presented at NELS 22. Zagona, K. (1994) 'Perfectivity and Temporal Arguments' in M. Mazzola (ed) Romance Linguistics: Theory and Issues. 203 Morrill Ha11 Cornell University Zthaca NY 14853 USA E-mail: [email protected]