Late closure in context

The debate surrounding the use of extra-sentential context to inform early parsing decisions has focused primarily on the class of syntactic ambiguity to which Minimal Attachment (Frazier, 1979) applies. The present paper extends the debate to Late Closure (Frazier, 1979). We argue that Crain and Steedman's (1985) Principle of Parsimony predicts a specific circumstance in which referential context should override the tendency for late closure or right association (Kimball, 1973). The first of four eye-movement reading time studies failed to confirm the prediction. However, when similar materials were embedded in a context which explicitly directed attention toward the appropriate predicate (i.e., the predicate associated with the high attachment), we found evidence of contextual override. We argue that the data fit well with the constraint-satisfaction view of sentence processing (e.g., MacDonald, Pearlmutter, & Seidenberg, 1994; Trueswell & Tanenhaus, 1994), although we offer an augmentation of the account based on thepredictiveactivation of forthcoming structure (cf. Elman, 1990).