Imitation tests encompassing intermingled meaningful and meaningless items are normally used to assess ideomotor apraxia, implicitly assuming that they would test the lexical and the non-lexical route, respectively. However, these mixed lists might induce a "list composition" effect similar to that found in word recognition studies where familiar material can be processed via the non-lexical route. This hypothesis was put to test by examining praxis skills of 23 left-hemisphere damaged patients using the same gestures in two formats: pure and mixed lists (i.e., meaningless or meaningful gestures administered separately or intermingled, respectively). Results showed that patients performed better on the imitation task when pure lists were used. Moreover, asymmetries of performance were observed. Patient SL scored better in the imitation of meaningful gestures in the pure list than in the mixed list condition. Patient CA performed poorly in the imitation of meaningless gestures only in the mixed list condition. Dissociations observed in imitation tasks could be biased by the use of mixed lists. Also "pure" lists should be used for the diagnosis of imitation deficits in apraxia.