Three experiments are reported showing that diagnosis of skin disorders by medical residents and general practitioners was facilitated by similar cases previously seen in the same context. Diagnosis of similar cases was facilitated more than that of dissimilar cases in the same diagnostic category, demonstrating that facilitation was not solely due to activation of the diagnostic category as a whole. Because diagnosis was posed in a multiple-choice format that always included the correct diagnosis, the relative disadvantage of dissimilar items was not due to the unavailability of the category name. The similarity effect also occurred with 2-week delay between the initial case and the test cases. Variations in diagnostic procedure, ranging from giving a quick first impression to arguing for given alternative diagnoses before selection, did not interact with the effect of similarity. This result suggests that the similarity effect is not strongly dependent on a particular diagnostic strategy.