Previous experiments with picture sorting and matching tasks have shown aphasics to give more deviant responses than controls when decisions require the identification of single features of concepts, whereas their responses are close to normal whenever decisions have to be based on the relative overlap of broad associative fields. The present experiment was designed to compare picture matching based on single features (property verification) with picture matching based on category membership (category verification). Fifty-five aphasics (14 amnesics, 18 Brocas, 13 Wernickes, 10 global aphasics) and 29 right-brain-damaged control patients served as subjects. Aphasics were poorer than right hemisphere controls on property as well as on category sortings, especially when the sorting criterion was not a dominant property of the object or when the object in question was not a typical member of the criterion category. Contrary to other studies, the “semantic distance” variable did not differentially affect Brocas as compared to Wernickes aphasics. Verbal labels denoting the sorting criterion and added to the picture presentation did not affect the performance of the right hemisphere controls but significantly improved that of the aphasics.