The time course and durability of predictive inferences

Abstract The latency to pronounce a probe word has been used as an index of whether or not inferences about highly predictable events are drawn during reading. Using this index. Potts, Keenan, and Golding (1988, Journal of Memory and Language , 27, 399-415) failed to find evidence for predictive (forward) inferencing. Using the same index, we (1) replicated the Potts et al. finding, (2) found that the predictive condition produced significant priming in pronouncing the probe when the probe was presented immediately after the predictive sentence, (3) demonstrated an absence of such priming when the probe was delayed by an interval not filled with additional text input, and (4) found significant priming regardless of when the probe was presented when the sentence format increased processing difficulty. These results are taken as support for the view that predictive inferences are temporarily drawn and subsequently deactivated. With more difficult material, deactivation of predictive inferences may either be delayed or prevented. The mechanisms of such deactivation are discussed in light of the results.