In five experiments, we examined the role of global and local cues in memory retrieval in infancy. Six-month-old infants were trained at home in a distinctive context (playpen liner) to kick to move a mobile. The liners were yellow and displayed either green stripes, green squares aligned vertically in stripe-like columns, or green squares in a grid pattern. The stripes and columns liners had a similar global configuration but different local components; the columns and grid liners had identical local components but different global configurations. When infants were tested 24 h after training in the presence of context liners that differed from the training context in either global configurations or local features, their memory retrieval was disrupted (Experiments 1 and 2). However, a change from stripes to columns failed to disrupt memory retrieval, even though the reverse change, from columns to stripes, did. Experiments 3, 4, and 5 revealed that this asymmetry was due to the fact that, when discriminative local information is not directly associated with training, a postperceptual strategy enables infants to disregard a mismatch in local information between training and test contexts and to generalize on the basis of a match in global information during the 24-h retention test. Thus, infants encode and remember for substantial periods of time both global configuration information and local component information in the incidental context in which an event occurs and flexibly utilize this information when responding to new events.