Assessing the occurrence of eleborative inference with recognition: Compatibility checking vs compound cue theory

Abstract

In previous research, we have used delayed item recognition to investigate the information about elaborative inferences that is generated during reading. Recently, Potts, Keenan, and Golding suggested that results from recognition experiments are not determined by information encoded during reading, but rather by information calculated at the time of the recognition test. They proposed that test items that represent potential inferences from studied sentences are compared to the sentences at the time of the recognition test, and performance on recognition is a function of the compatibility of the test items with the sentences in memory. In this article, we present both theoretical and experimental results that argue against this compatibility-checking hypothesis. In Experiment 1, on-line lexical decision is used to demonstrate that relations between sentences and test items that affect on-line performance do not affect performance in delayed recognition, as they should according to the compatibility hypothesis. In Experiment 2, it is shown that subjects' ratings of compatibility do not predict recognition performance. It is concluded on the basis of the experimental results that the compatibility hypothesis does not account for delayed recognition performance. In addition, a review of theories of recognition memory shows that the compatibility hypothesis is not consistent with the temporal dynamics of processes in these theories.