Two hypotheses were raised and empirically tested to account for the failure of previous controlled validation studies to find evidence of literacy in nonspeaking persons with autism using facilitated communication: (a) The naming tasks used in other studies have triggered specific "word retrieval" problems, or anomia, and (b) a perceptual problem, visual agnosia, prevents subjects from recognizing objects without touching them. Three nonspeaking autistic children who had used facilitation for at least 2 years were evaluated with four experimentally controlled tasks, over a period of 5 months. In descriptive and object handling tasks, and in a traditional picture identification task, subjects failed to type correct answers when facilitators were blind; one subject, however, occasionally engaged in signing and vocalizations that were context-appropriate. Results reflected a generalized language deficit, rather than isolated word-finding or perceptual difficulties, and were consistent with many previous studies revealing facilitator cuing. Questions are raised about inconsistencies in pseudo-correct scores, a measure of facilitator influence, reported here and in previous research.