Source: Dissertation Abstracts International, Volume: 78-11(E), Section: A.

Adviser: Maria Mercedes Pinango.

Access restricted by licensing agreement.

This item is not available from ProQuest Dissertations & Theses.

Complement coercion refers to a phenomenon that an entity-denoting complement following verbs such as begin and enjoy, receives an eventive reading. For example. sentence (1) "The princess began/enjoyed the book." is often read as (2) The princess began/enjoyed [ reading/writing] the book. The eventive meaning in (2) is not morpho-syntactically supported in (1) but is understood by comprehenders. In real-time comprehension, sentences involving complement coercion are reported to engender additional processing cost.

Two hypotheses have been proposed to account for this additional processing cost. The Type-Shifting Hypothesis assumes that verbs like begin (aspectual verbs) and enjoy (psychological verbs) select for eventive complements. Combining these verbs with an entity-denoting complement results in a semantic mismatch, which calls for a type-shifting operation to coerce the complement to an event in order to satisfy the selectional restriction of the verbs. On this hypothesis, the additional processing cost results from the interpolation of this operation in comprehension, and should be observed for all cases involving a type mismatch.

On the other hand. the Structured Individual Hypothesis captures the complement coercion effect as the processing; of aspectual verbs (AspVs). These verbs are dissociated from other verbs (e.g., psychological verbs) by their semantic properties (Pinango k Deo. 2012. 2015). On this hypothesis, the processing cost that is previously argued to reflect complement coercion in fact results from two processes of comprehending AspV sentences: (i) exhaustive lexical retrieval, and (ii) ambiguity resolution based on context to obtain an appropriate interpretation. This predicts that only aspectual verbs, but not psychological verbs, require more processing effort.

I conducted four studies (using self-paced reading, fMRI, eye-tracking, and questionnaire) to adjudicate between the Type-Shifting Hypothesis (TSH) and the Structured Individual Hypothesis (SIH). Results are consistent with the SIH and inconsistent with the TSH: (i) sentences with aspectual verbs engender additional processing cost as compared to sentences with psychological verbs, (ii) the ultimate reading and acceptability of AspV sentences are influenced by context. The findings have implications for the real-time comprehension of meaning underspecification, as the two hypotheses imply different views of sentence processing.

Based on the results, I argue that the complement coercion phenomenon should be treated as the processing of aspectual verbs, which involves underspecification among multiple semantic representations, rather than a mismatch followed by a repairing operation. I discuss the processing architecture implicated by the SIH and identify the properties that need to be taken into consideration by any processing model of meaning underspecification. During comprehension, lexical representations are exhaustively retrieved and incrementally composed. The composition of lexical representations gives rise to multiple semantic representations in parallel, causing an ambiguity to be resolved. Comprehenders determine an appropriate representation among multiple alternatives based on contextual constraints, and this process of ambiguity resolution engenders greater computational cost. This view of meaning underspecification is consistent with the constraint-satisfaction approach to sentence processing.