Abstract Subjects were asked to rehearse word trigrams in a particular prefamiliarized male or female voice for 5, 10, or 15 sec. In Experiment 1, recognition performance improved with the amount of primary (maintenance) rehearsal only if the speaker's voice at test matched the rehearsal voice, but recognition performance improved with the amount of secondary (elaborative) rehearsal regardless of the sex of the speaker at test. With a visual testing procedure in Experiments 2 and 3, the amount of primary rehearsal given to a trigram had no effect on recognition performance unless the original voice context was reinstated mentally at test. These results suggest that: (a) Secondary rehearsal builds up semantic associations, whereas primary rehearsal serves to associate items with their physical characteristics at presentation. (b) There is an important memory search component in recognition as well as in recall. (c) Imaginal operations can yield a product in memory that is similar to that left by perceptual operations.