Categorization and memory retrieval by three-month-olds

84 3-month-olds were tested in 3 studies of the acquisition and long-term retention of category-specific information. Infants who were trained with a perceptibly different member of an alphanumeric category on each of 3 days generalized responding to a novel instance of the original training category but not a novel member of a novel category during a 24-hour novelty test. 2 weeks later, when infants displayed no evidence of remembering their prior training experience, categorization was reinstated if a novel exemplar from the original training category was used as the retrieval cue in a memory reactivation procedure. A novel exemplar from a novel category was not an effective retrieval cue. The effectiveness of the category-specific retrieval cue was a function of its physical similarity to the individual exemplars encountered during training, not testing. The background against which the alphanumeric exemplars were displayed during training was not an effective retrieval cue in either the 24-hour novelty test or the memory reactivation procedure, indicating that all invariant stimulus attributes do not contribute equally as category cues. These data are the first to document retention of category-specific information after extended intervals. A popular account of categorization holds that infants abstract invariant features from individual exemplars and form a schema or distinctive memory representation of these shared features against which subsequent exemplars are compared. The present data provide support for a more parsimonious account of categorization, based on the retrieval of information about individual exemplars, that does not require an assumption of prototype formation.