Nagel, Harold Nicholas

This project investigates how subjects' preferences for a specific argument structure of a verb affect the on-line processing of sentences. Three experiments are conducted. The first experiment uses a forced choice task to gather preference ratings for several verbs. The second and third experiments use these preference ratings to examine on-line performance for a subset of the subjects participating in Experiment I. We find that a subject's preference for a specific argument structure influences the first pass analysis of sentence but that this effect may be overridden by a structural effect in sentences containing syntactic ambiguities. These results are discussed in terms of lexical guidance vs. phrase structure-driven models of sentence processing. It is concluded that both lexical and phrase structure information are utilized during first pass analysis of a sentence, but that the structure of a sentence determines which of these effects is measurably observable.

This project comprises a series of experiments investigating the role of prosody--the timing and intonation of an utterance--in syntactic disambiguation. Acoustic analyses isolated two parameters--main-clause verb segment and pause durations, and the pitch contour over the verb and the following phrase--that reliably predicted syntactic structure in two sets of temporarily ambiguous sentences. The manipulation of one of these parameters--verb and pause duration--resulted in increased processing load over the disambiguating region of sentences temporarily ambiguous between a direct object and an embedded clause syntactic structure (e.g., "John knew the answer by heart" vs. "John knew the answer was correct"). Also, differences in the prosodic contours associated with temporarily ambiguous "filler-gap" sentences determined whether or not a gap was posited during on-line sentence processing. These findings suggest that prosodic information is used early, perhaps immediately, to make informed on-line parsing decisions and support a model of sentence processing in which both lexical and prosodic information interact on-line to generate the syntactic representation of an utterance.