Three experiments used an implicit memory test and a surprise recognition test to assess the retention of words that subjects had earlier read on a CRT screen. In Experiment 1, subjects were required to name words presented on the screen for either 36 ms or 54 ms. In Experiments 2 and 3, the words were presented at a point where roughly half of them could be identified correctly, by manipulating the brightness of the CRT screen. Following the naming phase, all subjects were given a word-fragment completion or a perceptual identification test, and a recognition test. Recognition memory for words which subjects could name correctly on the naming task was better than that for words they could not read in the study phase. Similarly, priming effects in the implicit memory tests were found for only the words subjects successfully read. There were no priming effects in near-zero identification condition. We conclude that repetition priming is dependent on conscious identification of stimuli.