A multiple cue hypothesis for the testing effect

Abstract

Three experiments investigated the role of variable processing in the testing effect. All three experiments utilized a paired-associate paradigm in which the same target items were presented, followed by an intervening cued-recall test or additional study opportunity, followed by a final free-recall test over all target items. In Experiment 1, related cue-target pairs (dough: bread) were presented, followed by either an intervening test in which the cue was either the same or different from the one encountered during initial presentation (dough:______ , toast:_______ , respectively), or by an intervening study opportunity in which the cue was either the same or different from the one encountered during initial presentation (dough: bread, toast: bread. respectively). In Experiment 2, cue-target pairs were the same from initial presentation to intervening test or study, and were either highly related (dough: bread) or weakly related (basket: bread). In Experiment 3, unrelated cue-target pairs (basket: bread) were presented, followed by either a yes/no recognition test or additional study opportunity in which the correct target was presented after either 0 incorrect targets (basket: bread). 1 incorrect target (basket: wicker, basket: bread), 2 incorrect targets (basket: wicker, basket: eggs, basket: bread), or 3 incorrect targets (basket: wicker, basket: eggs, basket: carry: basket: bread). A significant testing effect was obtained in all three experiments. In addition, a significant advantage in retention was obtained for target items presented with different as opposed to same cues (Experiment 1), a significant Trial x Cue Interaction was obtained in which retention for tested items benefited from unrelated as opposed to related cues whereas retention for studied items was unaffected by cue-target relatedness (Experiment 2), and finally, retention for studied items but not tested items benefited significantly from a greater number of incorrect targets such that the relative advantage for tested over studied items decreased with each additional incorrect target (Experiment 3). Results of these three experiments suggest that retention for both tested and studied items can benefit from variable processing, and that the advantage for tested over studied items is greatest under conditions that promote the sampling of extra information during retrieval.