The Biological Bases of Syntax-Semantics Interface in Natural Languages: Cognitive Modeling and Empirical Evidence

This paper reviews recent research on representation of events in human languages. We consider empirical evidence for event-structural analysis of language comprehension and production in spoken and signed languages, as well as possible biological bases for it. Finally, theoretical linguistic models, models of language processing, and cognitive architectures, which account for such event-structural basis of syntax-semantics (and, possibly, phonology interface in ASL) in human languages are discussed. Representation of events in human languages: linguistic universals meet language processing. The idea that human languages parse and formulate observable events in a logically restricted fashion is fairly old, dating back to Vendler’s (1967) Aktionsart predicate classes (more recently developed by van Lambalgen and Hamm, 2005). Recent work by Van Valin (2007) claims that the most pervasive components of real-world events have made their way into the morphology of most of the world’s languages (albeit in different form), qualifying them for the status of linguistic universals. For example, he notes that durative, dynamic, causative, and inchoative alternations surface as morphological constants in various language families; while the granularity of the system remains to be (empirically) determined, the pervasiveness of it in the linguistic data is striking. Linguistic theory of verb types has long observed universal correspondences between verbal meanings and syntactic behaviors, including adverbial modification (Tenny 2000), aspectual coercion (Smith, 1991), and argument structure alternations (Levin, 1993, Ramchand, 2008, etc.). Vendler (1967) proposed a system of four basic syntactically relevant semantic types of predicates: atelic States and Activities, and telic Achievements and Accomplishments. The telic/atelic distinction is most clearly analyzed in terms of the internal structure of events. ‘Telic’ is understood as the property of events (linguistic predicates) containing a conceptual (semantic) endpoint. In contrast, ‘atelic’ events do not contain such a point and have the potential to continue indefinitely. Atelic events are homogenous, in that they may be divided into identical intervals, each of which is an instance of the event itself, i.e. ‘walking’ as an instance of ‘walking’. Telic events are composed of at least two subevents, one of which is the final state, and are therefore heterogeneous (cannot be divided into identical intervals). The model was further developed by Pustejovsky (2001), with the primary distinction between static subevent type S(tate) and dynamic subevent type P(rocess). Vendlerian transitions were modeled as combinations of non-identical subevents: either S S (Achievements) or P S (Accomplishments). Ramchand (2008) further developed a syntax-semantics interface model of event and argument structure including three possible event phases: initiation, process, and resultant state (with corresponding argument realization), and elaborated a factorial typology of event-argument predicate constructs based on formal minimalist syntax, overtly incorporating semantic notions of durativity, resultant state/telicity, agentivity and causation into the interface. Linguistic theories beyond minimalism also converge on a similar set of semantic distinctions affecting syntax in the world’s languages. For example, Dixon (2000) lists state/action, control, volition, affectedness and intention among semantic primitives involved in contrastive pairs of causal constructions across the world’s languages; and Pinker (1989), in elaboration of the Grammatically Relevant Semantic Subsystem Hypothesis (GRSSH), distinguishes abstract semantic features, which affect grammar (e.g. punctual vs. durative, causation), and idiosyncratic ones (such as manner), which do not. While complete analysis of all known languages of the world with respect to their event structure is not currently feasible, additional evidence for the fundamental character of a predicate’s event structure comes from psycholinguistic and neurolinguistic research; an increasing body of research is supplying empirical evidence that such concepts as telicity affect the way syntactic structure is processed in human languages. For example, O’Bryan (2003) showed independent effects of telicity and transitivity on response times in a word maze experiment with reduced relative clauses. The subjects were asked to complete grammatical sentences containing Object reduced relative clauses, such as “The actress awakened by the writer left in a hurry”. There was a significant reaction time advantage on the preposition “by” for sentences with telic verbs in the relative clause, as compared to those with atelic verbs; there was also an independent advantage for the second argument in sentences with obligatorily transitive verbs (both telic and atelic). The experiment thus demonstrated independent effects of telicity and transitivity on human language processing ability. A similar study by Friedmann, Taranto, Shapiro, and Swinney (in press) compared sentences with unergative (intransitive atelic) and unaccusative (intransitive telic) English verbs using a cross-modal priming technique, and found a priming effect for nonalternating unaccusatives (obligatorily intransitive telic verbs), but not for unergatives (obligatorily intransitive atelics). Additional evidence of telicity affecting online sentence processing in spoken English was provided in an EEG study by Malaia, Wilbur, and Weber-Fox (2008), which investigated the effects of verbal telicity on the ease of syntactic re-analysis of Object reduced relative clauses. The ERP data demonstrated that subjects’ recovery from garden-path effects required fewer processing resources when telic verbs were used in reduced relative clauses. The same frame alternations in atelic verbs were more difficult to process, and elicited ERP waveforms typically associated with increased difficulty of early syntactic processing and thematic role integration. Because the study controlled for other factors which could affect processing, including argument animacy, frequency of verb occurrence, and transitivity, the results were clearly due to the linguistic telicity influencing sentence processing at the syntax-semantics interface. Theoretical models which account for such empirical data range from construction grammar (Kemmerer 2006, Goldberg 1995), claiming regular correspondence between semantics and syntax in verbal constructions, to a fully developed event-structure based syntax-semantics interface (Ramchand 2008), which attempts to account for the ways regular constructions and cross-linguistically observable morphology could have developed, using the minimalist model of syntax. Psycholinguistic models of processing grammatically relevant semantic information, on the other hand, are still under development (see Kemmerer, 2006, for examples). Even less is known about emergence of grammatically relevant semantic features in the lexicon. We suggest however, that signed languages, which are more closely tied to the visual modality simultaneously the one of perception – might provide the missing link to the puzzle. Overt representation of event structure in