On bound variable interpretations: the LF-only hypothesis.

Under what conditions do perceivers prefer to assign a bound variable interpretation to a pronominal that is ambiguous between a bound variable and a coreferential interpretation? Several experiments were designed to test the hypothesis that language perceivers prefer a bound variable over a coreferential interpretation of a pronoun because the former only requires consultation of a logical form (LF) representation, while the latter requires access to a discourse representation. The hypothesis was disconfirmed in two respects. First, although bound variable interpretations show a processing advantage over coreferential interpretations in VP ellipsis constructions, the preference for bound variable interpretations is not general--it does not extend to other quantificational contexts. Second, the preference for bound variable interpretations in VP ellipsis constructions is not limited to examples in which the antecedent and the ellipsis site occur in the same sentence. If the bound variable advantage were due to the ready availability of the LF for the current sentence, the advantage should disappear across sentence boundaries. An alternative hypothesis is then considered which could explain the source of the bound variable advantage in VP ellipsis contexts.