Can a person with a damaged medial-temporal lobe learn a category implicitly? To address this question, we compared the performance of participants with mild Alzheimer's disease (AD) to that of age-matched controls in a standard implicit learning task. In this task, participants were first presented a series of objects, then told the objects formed a category, and then had to categorize a long sequence of test items [Knowlton B. J., Squire L. R. (1993). The learning of categories: parallel brain systems for item memory and category knowledge. Science, 262, 1747-1749]. We tested the hypotheses that: (1) both Control and AD participants would show evidence for implicit learning after the unwanted contribution of learning during test is removed; (2) the degree of implicit learning is the same for AD and Control participants; (3) training with exemplars that are highly similar to an unseen prototype will lead to better implicit category learning than training with exemplars that are less similar to a prototype. With respect to the first hypothesis, we found that both AD and Control participants performed better on tests of implicit learning than could be attributed to just learning on test trials. We found no clear means for evaluating our second hypothesis, and argue that comparisons of the degree of implicit learning between patient and control groups in this paradigm are confounded by the contribution of other memory systems. In line with the third hypothesis, only training with similar exemplars resulted in significant implicit category learning for AD participants.