A list of eight words was presented serially to S at a 1-sec. rate and followed after 2 sec. by a test word for item recognition. Under a second condition, 5 simply named the test word as rapidly as possible. It was argued that subtracting naming latency (NL) from decision latency (DL) yielded a relatively pure measure of comparison times in memory. The "corrected DL" values showed strong recency effects; this result has implications for models of recognition memory. Recency effects were also present in the NL measures, demonstrating a short-term priming effect in word naming. Manipulating input modality yielded faster NL and corrected DL values when the presentation series and test word were both given auditorily, and, since this faciliation was present over all serial positions, it was concluded that both naming and decision processes were speeded by the presence of relatively long-lasting auditory information in memory.