Semantic Relatedness and the Efficacy of Retrieval Practice

There is growing interest in the role that semantic memory plays in the retrieval practice (or testing) effect. We investigated how semantic relatedness impacts the efficacy of testing, relative to a restudy control, using a cued-recall paradigm with immediate correct answer feedback. High and low relatedness word pairs were created by selecting two different response words for each cue word based on associative norms. The parameter-free predictions of the dual-memory model of test-enhanced learning for low relatedness pairs were used as an empirically well-supported scaling benchmark for evaluating the effectiveness of testing for high relatedness pairs, allowing us to overcome the issue of removable interactions. Results showed a robust testing effect at both relatedness levels. For the high relatedness group, however, the testing effect was about 26% smaller than expected by the model. Nevertheless, the predicted quadratic relation between test and restudy proportion correct in the cumulative distribution held for both groups. One candidate account of those results within the dual-memory framework is that, for the high relatedness case, study and test memory strengths for tested items are positively correlated rather than independent. A simulation implementing that version of the model provided a good distribution fit for the high relatedness group. Other theoretical accounts of the semantic relatedness effect are considered.