<p>Previous research has obtained better memory performance for items that were studied in two modalities than in one modality. We investigated the role of multisensory integration and of study-test overlap. Items were presented once as unimodal (picture or sound), once as bimodal (picture and sound), or twice as unimodal (once as picture, once as sound) stimuli in a continuous recognition task. In Experiment 1 we found a benefit for items that were studied in both modalities. The benefit did not depend on temporal alignment of picture and sound, which poses problems for multisensory integration as an explanation. In Experiment 2 we found that repetition of items in the same modality resulted in better memory performance than repetition in different modalities, and we found that memory performance was better when study and test format were the same than when they were different, supporting a role for encoding specificity. We conclude that multimodal presentation during study benefits memory only when the test item is also multimodal. Moreover, this benefit is more likely explained by study-test overlap than by multisensory integration.</p>