According to a lexical-access hypothesis, priming on verbal-completion tests requires previous access to higher-order nodes such as morphemes or words. In two experiments, subjects studied words presented in printed color, either by reading the words or by naming the printed colors. In Experiment 2, some of the words were to be ignored in both study conditions. The words presented in Experiment 2 were either conceptually related or unrelated to each other. And the cues in the test phase for repeated targets were either printed in the studied color or in a different color. In both experiments, the word stems of nouns studied and of new targets were presented with word-completion instructions (implicit test of memory). Consistent with the hypothesis, the amount of priming obtained was greater in the reading than in the color-naming condition. The least amount of priming was observed for targets that were to be ignored at study. In addition, Experiment 2 showed considerable effects of surface similarity (i.e., stronger priming effects with cues whose color was not changed from the study to the test phase) and some effects of conceptual processing, as was indicated by different amounts of priming for unrelated and for related nouns. Because these two effects were observed with both types of study task, it is concluded that perceptual and conceptual processes may make independent contributions to priming effects in verbal perceptual tests. This conclusion modifies and extends the original hypothesis, according to which previous lexical access is required in order to obtain effects of perceptual or conceptual processes on the priming scores.