Ljiljana Progovac Wayne State University

7 downloads 0 Views 112KB Size Report
Wayne State University. The Issue. It is typically assumed in various linguistic frameworks that sentences without a finite. (auxiliary) verb such as (1) below in ...
ROOT SMALL CLAUSES WITH UNACCUSATIVE VERBS Ljiljana Progovac Wayne State University The Issue. It is typically assumed in various linguistic frameworks that sentences without a finite (auxiliary) verb such as (1) below in Serbian are full finite clauses, whose auxiliary verb first gets merged and then deleted/elided. This implies that they are syntactically/semantically equivalent to the clauses in (2), which feature an overt auxiliary. An obvious alternative analysis would be to treat examples in (1) as auxiliary-less at all levels or representation, that is, to analyze them as “root small clauses (RSCs)”, a la Progovac’s (2006) account of English examples such as Me first!, Him worry?!). In the analysis of Stowell (1983) and Chomsky (1995), embedded small clauses are lexical projections of the predicate, and contain no TP projection. I present three arguments in favor of such TP-less analysis of RSCs in (1), and then address the apparent problem for this analysis, the fact that their subjects surface in the nominative case. (1)

(2) (3) (4)

Stigla pošta. /Pala vlada. / Došla zima. / Pao sneg. arrived.FSG mail fallen-FSG government come-FSG winter fallen-MSG snow /Umro Petar. died-MSG Peter Pošta je stigla. /Vlada je pala./ Zima je došla. / Sneg je pao. /Petar je umro. Aux.3SG Stigla je pošta. /Pala je vlada. /Došla je zima. /Pao je sneg. /Umro je Petar. ??? Pošta stigla. / ???Vlada pala. / ??Zima došla. / ???Sneg pao. / ???Petar umro.

Background. Crosslinguistically, unaccusative verbs such as those in (1) are typically analyzed as taking an internal (theme) argument, which first merges as a complement of the verb (for general discussion, see e.g. Perlmutter 1978, Burzio 1986, Harves 2002). This internal argument would become the subject of the clause (TP) only upon the merger of Tense, and upon its Move to the specifier of TP, deriving (2). The (derived) subject agrees in person/number features with the auxiliary verb/Tense, gets its nominative case by Agree, and checks the EPP feature of Tense by Move (see e.g. Chomsky 2000; 2001). (On the other hand, the subject in (3) can be argued to check its nominative case feature in situ by Agree, but that Move does not take place here, for whatever reason: perhaps the finite T in pro-drop languages is optionally selected with an EPP feature – the issue is beyond the scope of this paper, however). First Argument: Word Order. But now, if (1) is a root small clause, there is no syntactic reason (such as an EPP feature of Tense) for its only argument to move to the subject position (SpecTP), since there is no T there in the first place (in other words, this is not just a defective T in the sense of Lavine & Freidin (2002) – it is a clause without any T). This would account for the strongly preferred word order in (1), contrasting with the highly marked order in (4). (The default nature of VS word order of unaccusatives, in contrast to unergatives, is also noted in Babyonyshev (1996) for Russian.) On the other hand, if both (2) and (4) are full TPs, there is no (obvious) reason why (2) is perfectly natural/unmarked, while (4) is highly marked. Second Argument: Time Adverbials. If there is a syntactic difference between (1) and (2/3) having to do with the absence/presence of TP, then one would expect a semantic difference as well, and moreover a difference having to do with tense/time representation. Indeed, small clause examples such as (1) imply that the event being reported pertains to the present situation/context, which is confirmed by the fact that such clauses cannot be modified by e.g. adverbials denoting remote past, such as ‘pre tri meseca’ (‘three months ago’) (5-6). On the other hand, such modification is perfectly grammatical with fully sentential

counterparts (7). (5) ???Stigla pošta pre tri meseca. / ???Pala vlada pre tri meseca. ???Došla zima pre tri meseca. ???Pao sneg pre tri meseca. / ???Umro Petar pre tri meseca. (6)

??? Pošta stigla pre tri meseca. / ???Vlada pala pre tri meseca….

(7) Pošta je stigla pre tri meseca. / Vlada je pala pre tri meseca. / Zima je došla pre tri meseca. / Sneg je pao pre tri meseca. / Petar je umro pre tri meseca. Third argument: Embedding. Functional categories typically select a unique complement: e.g. C selects a TP (e.g. Abney 1987). If so, then the RSC analysis of (1) correctly predicts that these small clauses cannot be embedded under C, and that they thus cannot serve as subordinate clauses such as (8). This is in sharp contrast with full finite clauses (9), which can, of course, embed. This contrast is unexpected under an ellipsis account of the examples in (1). (8) *Ja znam da stigla pošta. / *Ja znam da pala vlada. / *Ja znam da došla zima. / *Ja znam sneg. / *Ja znam da umro Petar. (9)

a) b)

da

pao

Ja znam da je stigla pošta/pala vlada/došla zima/pao sneg/umro Petar. I know that AUX Ja znam da je pošta stigla/vlada pala/zima došla/sneg pao/Petar umro.

Default Nominative Case: But if there is no tense/TP in (1), how is nominative checked? The general assumption is that nominative case must be checked by the process of Agree with finite T (see above). Even though the participles in (1) agree in gender/number with the subject, they do not agree in person, and are thus not φ–complete. Such gender/number agreement is also operative in e.g. noun phrases, between nouns and adjectives, and can clearly not be taken to imply finiteness/T. These participles are thus not finite (if they were, there would be two finite verbs in (2-3)), and should not be able to check nominative case by Agree. I propose that nominative in (1) is not structural, but rather default case, that is, the form an NP takes when not case-marked at all. Schütze (2001) argues that default case is found in non-argument positions (e.g. predicate positions), and in isolation (e.g. labels, titles, etc.). In this sense, accusative is default in English, while nominative is default in Serbian, for which ample independent evidence will be provided. If correct, this analysis has far reaching consequences for the structure of clauses/sentences in general: it challenges the widespread belief that every meaningful utterance has to involve a full TP, especially if it expresses a proposition. References Abney, S. 1987. The English Noun Phrase in Its Sentential Aspect. MIT PhD. Diss. Babyonyshev, M. 1996. M. Structural Connections in Syntax and Processing: Studies in Russian and Japanese. MIT Ph.D. Dissertation. Harves, S. 2002. Unaccusative Syntax in Russian. Ph.D. Dissertation, Princeton. Lavine, J. & R. Freidin. 2002. The subject of defective T(ense) in Slavic. Journal of Slavic Linguistics 10: 253-89. Progovac, L. 2006. Phrases and Small Clauses at the Root. In Progovac et al, eds. The Syntax of Nonsententilas: Multidisciplinary Perspectives. Benjamins. Schütze, T. C. 2001. On the nature of default case. Syntax 4.3: 205-238. Stowell, T. 1983. Subjects across categories. The Linguistic Review 2/3:285-312.