UC San Diego

This dissertation presents three studies using the event- related potential methodology (ERPs) that offer augmentative evidence for investigations of lexical properties during on-line auditory sentence comprehension. Because ERPs are sensitive to specific linguistic properties and have relatively fine-grained temporal resolution, the methodology can be brought to bear on significant issues in psycholinguistic inquiry that have yielded conflicting or inconclusive results. In particular, ERPs can differentiate between alternatives indistinguishable using behavioral methods. Experiment 1 examines whether the property of antecedent concreteness in available at the pronoun during co-reference resolution. Results suggest that antecedent concreteness is available relatively immediately at the pronoun in auditory sentences, supporting hypotheses that antecedents are reactivated at coreferent pronouns at a deep, conceptual level. Experiment 2 investigates the operation of aspectual coercion, in the case where an event that is typically interpreted as a point action event (e.g., "The light flashed at dawn") is "coerced" into a durative, iterative interpretation by an aspectually marked modifier (e.g., "The light flashed until dawn"). Consistent with linguistically motivated hypotheses, results indicate that aspectual coercion indeed may be a (primarily) semantic operation, manifesting as a lasting negative shift in the ERP in coercion conditions starting immediately when, or shortly after, coercion is licensed. Furthermore, a different effect is found when aspect is "coerced" in process verbs with a continuous but bounded sense (e.g., "The light glowed until dawn"). Experiment 3 examines the effects of phrase structure violations on the ERP. In this study, participants listened to grammatical and ungrammatical English and Jabberwocky sentences (comprised of semantically null nouns but English verbs and closed class words) to determine whether lexico-semantic influences of verbs properties (e.g., verb subcategorization rules) would be preserved even in language that contains meaningless (semantically null) arguments. Results indicate that phrase structure violations elicit a P600 effect in both English and Jabberwocky sentences, suggesting increased processing costs in the violation conditions. No LAN to violations was evident in either language for either violation type, contrary to what has been found with identical stimuli presented visually