Perceptual implicit memory for previously attended and unattended words was measured either in a word-stem completion task or in a perceptual fluency (perceptual identification) task. Subjects (N = 144) first engaged in a focused attention task in which they were to identify one of two words presented for 100, 200, or 300 msec. Words were classified as attended if they were reported during the focused attention task and unattended if they were not. Results for both implicit memory tests indicated reliable perceptual implicit memory for attended words but not for unattended words, regardless of focused attention exposure duration. The results indicate that perceptual implicit memory tasks reflect attentional encoding processes; that is, words must undergo attentional encoding if they are to affect performance on a later perceptual implicit memory test.