Regeneration in Recall and Verb Phrase Ellipsis

In this thesis, I propose the Ellipsis Regeneration Hypothesis, whereby resolution of verb phrase ellipsis (VPE) involves accessing a conceptual/propositional representation of the antecedent, and regenerating structure using recently activated lexical items. This hypothesis pulls together strands from two independent areas of research: anaphor resolution and recall. Previous research on the processing of VPE constructions has demonstrated that the linguistic structure of the antecedent appears to be accessed and reactivated during resolution. At the same time, in a series of experiments investigating immediate recall, Potter and Lombardi (1992, 1998) and Lombardi and Potter (1992) argue that such linguistic representations do not persist in memory. Under their view, immediate recall (which is normally highly accurate) entails access to a conceptual (rather than linguistic) representation of the target and regeneration of structure using recently activated, but unordered, lexical items. The primary piece of evidence for this hypothesis is lure intrusion: if recall consults a surface representation of the target, then it should not be possible to induce changes to lexical items or structure. And yet, Potter and Lombardi show that when subjects are presented with synonyms (lures) to a word in the target, they will sometimes select the synonym in recall. Moreover, if that synonym is a verb (e.g., donate) that requires a different structure to that of the target (e.g., give), subjects will make the appropriate structural changes in recall. Bringing the two strands together, one comes to an intriguing possibility: that VPE does not access a linguistic representation of the antecedent but rather, like recall, accesses a conceptual representation and regenerates using recently activated lexical items. Taking lure intrusion to be a hallmark of regeneration, three experiments were designed to determine whether the lure intrusion effect would obtain: a recall study, an ellipsis study, and a study using 'do it'. The prediction of the current thesis is that lure intrusion should occur with both recall and ellipsis, but not with 'do it'.. Evidence from the three experiments supports the hypothesis that the structural output of VPE is, like recall, not accessed, but regenerated.Ph.D