In 3 experiments, we manipulated 3-month-olds' attention to different components of a training display and assessed the effect on retention of those components. Attention was manipulated via a pop-out display (one target amidst 6 distractors) that enhances selective attention to the target relative to the distractors, and retention was assessed with displays composed entirely of targets or distractors. In Experiment 1, infants recognized both target and distractors after 1 day, confirming that both are initially encoded at some level. In Experiment 2, infants recognized a target L and distractor Ls after delays longer and shorter, respectively, than infants trained with Ls in a homogeneous display. Experiment 3 replicated the preceding pattern of results with a + stimulus. Thus, increasing or decreasing attention to an item during encoding produces a corresponding increase or decrease in its memorability. This finding is consistent with a levels of processing account and is inconsistent with accounts that deny a capacity for explicit memory to prelinguistic infants.