The contribution of specific contextual attributes to recognition of a well-learned cue was examined in four experiments with 6-month-olds. 24 h after learning to move a given mobile in a distinctive visual surround by kicking, recognition of the training cue was tested in either the original context or in one in which only a single contextual attribute was altered. Retrieval was completely disrupted by all form changes involving the deletion of angles and by a chromatic figure/ground reversal, but a discriminable change in form color had no effect. Although infants displayed partial retention in a degraded context after 1 day, they displayed no retention when a reminder was administered in the same degraded context after 20 days. These data reveal that infants do not encode contextual information holistically; moreover, they imply a privileged status for highly specific information about the incidental setting in which an event occurs. Unless this same highly specific information is perceptually identified at the time of testing, the memory of that event will not be retrieved. These data are consistent with our hypothesis that the visual context serves as an initial attention gate for memory retrieval.A common finding in studies of animals, children, and adults is that retention is better the greater the similarity between the encoding and retrieval contexts (

No abstract