Not All Forms of Morphological Mismatch are Acceptable in Verb-Phrase Ellipsis

Abstract

<p>The Recycling Hypothesis of verb-phrase ellipsis states that elided verb phrases with non-parallel antecedents are interpreted by reconstructing the appropriate verb phrase structure using the information available in the antecedent (Arregui, Frazier, Clifton, & Moulton, 2006). The hypothesis predicts that structurally more complex antecedents will involve more complicated reconstruction operations, which will lower the acceptability of the sentences. The experiments reported in this thesis tested two underlying assumptions of the Recycling Hypothesis as well as one prediction that follows from the proposal. First, the hypothesis assumes that elided verb phrases with parallel antecedents are interpreted by copying the structure of the antecedent into the ellipsis site (Frazier & Clifton, 2001). Second, Arregui et al. (2006) argued that changes in verbal morphology were “really easy (p. 242)” to recover from, suggesting that verbal morphology is not a factor in determining parallelism between the antecedent and elided verb phrases. Results from three written survey experiments in which participants were asked to judge the acceptability of verb-phrase ellipsis with matching or non-matching verbal morphology contradicted these assumptions. Morphologically more complex antecedents were rated less acceptable than simpler antecedents, regardless of whether the antecedent morphology matched the morphology on the elided verb phrase. The fact that verbal morphology affected acceptability ratings suggests that this factor plays a critical role in determining parallelism in ellipsis. Furthermore, the fact that parallel antecedents patterned with non-parallel antecedents suggests that the two must be processed in a similar fashion. Finally, if more complex antecedents require more complicated reconstruction operations, it might be predicted that word-by-word reading times at the ellipsis site should be correlated with the level of difficulty (Gibson, 1998). One self-paced reading experiment using the same materials showed no such correlation. These results are discussed with reference to two other psycholinguistic theories of verb-phrase ellipsis comprehension.</p>