A theoretical framework within neuroscience distinguishes humans' implicit and explicit systems for category learning. We used a perceptual-categorization paradigm to ask whether nonhumans share elements of these systems. Participants learned categories that foster implicit or explicit categorization in humans, because they had a multidimensional, information-integration (II) solution or a unidimensional, rule-based (RB) solution. Then humans and macaques generalized their category knowledge to new, untested regions of the stimulus space. II generalization was impaired, suggesting that II category learning is conditioned and constrained by stimulus generalization to its original, trained stimulus contexts. RB generalization was nearly seamless, suggesting that RB category knowledge in humans and monkeys has properties that grant it some independence from the original, trained stimulus contexts. These findings raise the questions of (a) how closely macaques' dimensional categorization verges on humans' explicit/declarative categorization, and (b) how far macaques' dimensional categorization has advanced beyond that in other vertebrate species.