Semantic underspecification and "optimal" linking
Valia Kordoni
ABSTRACT This paper focuses on the semantic properties and the syntactic behaviour of constructions like the ones in (1)-(3). These constructions belong to the Modern Greek (hence MG) Experiencer-Subject Psych Verb Constructions (hence ESPVCs), which feature an experiencer-subject in agreement with the verb, and either an accusative theme (see examples (1) and (3)), or a theme as the object of a prepositional phrase (see example (2)). Interesting about these constructions is the fact that the experienced (hence EXPD) semantic role is syntactically realised as the object of the sentence in constructions like (1) and (3), while in constructions like (2) it is syntactically realised as the object of a prepositional phrase. We should underline here that examples (1) and (2) convey the same meaning. That is, the two variants do NOT differ semantically, although the first one (example (1)) is used more often in the language. In other words, in order to express the meaning Mary is afraid of the storms native speakers do prefer the construction in (1), rather than the one in (2). Our aim, thus, here, is twofold: (i) to account for the semantic and syntactic properties of the predicates arguments in (1) and (2), and (ii) to show that the linking of the semantic roles in (1) to their respective syntactic arguments results to the construction in (1) being ranked higher than the construction in (2) in the native speakers preference. As far as the first of our aims is concerned, we follow Wechsler (1995), and we use his notion of notion in order to account for the semantic properties of the experiencer role in (1)-(3). In other words, the MG ESPVCs in (1)-(3) convey the meaning that in order for Maria to fear some entity x, she must have a notion of x, since that notion is the content of her fear. We also suggest that in the case of the MG ESPVCs the semantic argument denoted by the object NP (or PP) is entailed to be semantically underspecified. The idea of the semantic underspecification in relation to the EXPD semantic argument of the ESPVCs is not new in the literature. Dowtys (1991) linking theory, which argues for a direct mapping from events in the world and their participants to surface grammatical relations via proto-role entailments, predicts that the EXPD argument of the ESPVCs does not bear any Proto-Agent or Proto-Patient Properties. Zaenens (1993) account of the Dutch ESPVCs makes a similar prediction in a slightly different way, i.e., by determining the intrinsic classification of the EXPD argument rather than its surface grammatical role only. In Zaenens account the EXPD argument of the Dutch ESPVCs also lacks Proto-Agent and Proto-Patient Properties and bears the intrinsic classification (IC) feature [-r]. According to her, this argument is a semantically underspecified argument. With the semantic properties of the ESPVCs participants in place, we turn to the linking of the semantic arguments of the ESPVCs in MG to the syntax. For this purpose we adopt the linking theory proposed by Butt, Dalrymple, and Frank (1997). The reason we adopt this approach instead of the most classical LMT is that the linking architecture proposed by Butt, Dalrymple, and Frank (1997) incorporates naturally a set of preference constraints which impose an ordering on the available linking possibilities of given predicators. And this factor is what helps us support the observation that the construction in (1) ranks higher than the construction in (2) in the native speakers preference (see the numeric weights output in (1) and (2)). A clarification is due here: the intrinsic classification of the EXPD argument in (2) as [+r] is compatible with the analyses of this argument proposed by Anagnostopoulou (1995) and Markantonatou (1995). According to these analyses, the EXPD argument of constructions like the one in (2) is syntactically restricted. In conclusion: we show that a well-refined semantic analysis in combination with a robust linking theory like the one proposed in Butt, Dalrymple, and Frank (1997) cannot only account for the syntactic behaviour of constructions like the MG ESPVCs (cf., (1) and (2)), but can also reflect preferences in the use of such constructions by native speakers.
(1) I Maria fovate tis
kategides.
fovame <EXPR EXPD>
(2) I Maria fovate me tis kategides. fovame
<EXPR
EXPD> Subject Preference Total Optimal Numeric Weights +3 +1 +1 5
(3) I Maria fovate ton Giani.
References Anagnostopoulou, E. (1995). Experiencer-Object Predicates in Greek. Studies in Greek Linguistics 16, 254-265. Butt, M., M. Dalrymple, and A. Frank. (1997). An architecture for linking theory in LFG. In M. Butt and T. H. King (Eds.), Proceedings of the LFG97 Conference, University of California, San Diego. CSLI Publications. Dowty, D. (1991). Thematic Proto-Roles and Argument Selection. Language 67, 547-619.Markantonatou, S. (1995). Modern Greek deverbal nominals: an LMT approach. Journal Linguistics 31, 267-299. Wechsler, S. (1995). The Semantic Basis of Argument Structure. Stanford: CSLI Publications. Series: Dissertations in Linguistics, Joan Bresnan, Sharon Inkelas, William J.Poser, and Peter Sells (eds.). Zaenen, A. (1993). Unaccusativity in Dutch: Integrating Syntax and Lexical Semantics. In J. Pustejovsky (Ed.), Semantics and the Lexicon, pp. 129-162. Dordrecht: Kluwer Academic Publishers.
Back to Abstracts index |