The use of verb-based subcategorization and thematic role information in sentence processing

A central issue in language comprehension is how and when different kinds of information are coordinated to arrive at an interpretation of a sentence. This thesis explores how verb-based syntactic (subcategorization) and semantic (thematic role) information are used in syntactic processing. This is an important issue within psycholinguistics because it bears on the general mechanisms by which local syntactic ambiguity is resolved in immediate or 'on-line' comprehension. Syntactic ambiguity resolution can be viewed as an example of a more general problem, namely how perceptual and cognitive systems deal with local indeterminacy which comes from having to encode correlated multi-dimensional stimuli. Studying how verb-based syntactic and semantic information are used is particularly informative, because these two types of information are tied to the same word, but provide different classes of constraint. One provides constraints within the syntactic domain, whereas the other provides correlated constraints from the semantic domain. How and when these two classes of constraints are used bears directly on the types of solutions cognitive scientists have explored for resolving local indeterminacy.; The results presented here demonstrate that verb-based information is rapidly used to constrain local syntactic indeterminacy. One set of experiments, which used a variety of on-line measures (cross-modal naming, self-paced reading and eye-movement monitoring), showed that subcategorization information is available almost immediately upon encountering a verb and can constrain ambiguities concerning potential arguments. These studies also demonstrated that the processing system is sensitive to even subtle lexical constraints associated with co-occurrence factors. A second set of experiments, which monitored eye movements during reading, found that correlated semantic information associated with a verb's thematic roles is also used to constrain syntactic ambiguity. Thus, indeterminacy within the syntactic domain can be resolved by recruiting relevant information from the semantic domain. These effects depend upon the degree of constraint, which explains why the literature on this topic has been inconclusive. The results support interactive constraint-based models of sentence processing in which local syntactic ambiguity is resolved in a constraint-satisfaction process using correlated information from within and across domains.