The generation manipulation has been critical in delineating differences between implicit and explicit memory. In contrast to past research, the present experiments indicate that generating from a rhyme cue produces as much perceptual priming as does reading. This is demonstrated for 3 visual priming tasks: perceptual identification, word-fragment completion (WFC), and word-stem completion (WSC). This result occurred regardless of the mode of study response (written or spoken) or whether the generation condition was compared with reading words in or out of context. Rhyme generation did not produce priming on the letter height task (Masson & MacLeod, 2002), implying that the effect was not mediated by covert visualization. Nor was the effect due to the mere presence of the rhyme cue. Semantic generation (from definitions) produced a different pattern, exhibiting a reverse generation effect on WFC and WSC but full (read-level) priming on perceptual identification. The present results were not consistent with accounts based on the standard transfer-appropriate processing view, covert visualization, explicit contamination, or conceptual contributions to nominally perceptual tasks. (Contains 5 tables and 7 footnotes.)