Lexical semantics and argument structure in sentence processing

Frequency effects have been demonstrated for a variety of linguistic components, including argument structures. Numerous studies have provided evidence that argument structure frequency biases are one source of information which is used by the parser to constrain the developing interpretation and facilitate syntactic ambiguity resolution. Despite the evidence for the importance of argument structure frequencies to sentence processing, the source of the frequency biases has not previously been investigated. Building on a linguistic view that verbs can be classified into semantically coherent categories according to meaning configurations relevant to argument structures, lexical semantics were explored as a possible factor which can account for variability in argument structure frequencies. Argument structure frequency information was collected for a set of sentence complement (SC) taking verbs (e.g., propose ) and their corresponding nouns (e.g., proposal) which share many semantic features, and for a set of Dative and Benefactive verbs. Analyses of the collected frequency data demonstrated that lexical semantics can account for variability in argument structure frequencies: Frequency bias measures were correlated across verb-noun pairs, and verb semantic category accounted for variance in argument structure frequencies. Given the semantic category effects on argument structure frequencies, the role of semantic category was explored in comprehension and production of sentences. In an on-line syntactic ambiguity resolution reading task with SC-taking verbs from three categories, that are distinct semantically and in overall argument structure preferences, semantic category influenced processing difficulty beyond influences of individual verb preferences. In addition, in a newly developed methodology for syntactic priming, production of the alternative structures for SC-taking and Dative/Benefactive verbs was examined as a function of the semantic similarity between the prime and target verbs. The priming effect was influenced by the semantic relation. Specifically, prime verbs from the same semantic category as the target verb produced larger priming effects than prime verbs from a different semantic category. Taken together, the results point to a model of the mental lexicon in which semantic category is encoded in verb representations and is accessed and used during on-line sentence processing.