A temporal distinctiveness theory of contextually cued retrieval from memory is presented and applied to recency and modality effects. According to this theory, one part of the mnemonic trace of an item is a representation of the item's time of presentation. Time of presentation may be encoded with a coarse grain (so that it is consistent with a wide range of times) or with a fine grain (so that it is consistent with a narrow range of times). Retrieval proceeds by constructing temporally defined search sets that include representations of items consistent with the temporal bounds of the search set. The temporal width of the search set increases as the retention interval increases. Recency effects arise from retrieval of recently presented items from narrow search sets that include representations of few items; within the context of the search set, these items are distinctive and recalled well. Superiority in recall of recently presented auditory information in comparison with recently presented visual information is attributed to differences in the grain of time of presentation representations for aurally (fine grain) and visually (coarse grain) presented information. Four experiments confirm qualitative and quantitative predictions of the theory, including the prediction of auditory superiority at the beginning of the list when the initial items are temporally distinct.