An analysis of antecedent mismatch effects under ellipsis is proposed to explain why some cases of verb phrase ellipsis exhibit a sizeable penalty when the elided target is not structurally matched to its antecedent, while other cases show little or no penalty at all. The proposal attributes the penalty in the former case to an information-structural constraint governing contrastive topics, and it is argued that previous accounts have misattributed that penalty to a licensing constraint on ellipsis. Results from four experiments (three off-line acceptability, one on-line self-paced reading) confirm that the relative size of the mismatch penalty can be reliably predicted based on the information structure of the clause containing the ellipsis and that acceptability differences associated with information structure are observable even in the absence of ellipsis.
Stored only when both papers are in the corpus — 179 citations exist upstream.