Abnormalities in the integration of auditory and visual language inputs could underlie many core psychotic features. Perceptual confusion may arise because of the normal propensity of visual speech perception to evoke auditory percepts. Recent functional neuroimaging studies of normal subjects have demonstrated activation in auditory-linguistic brain areas in response to silent lip-reading. Three functional magnetic resonance imaging experiments were carried out on seven normal volunteers, and 14 schizophrenia patients, half of whom were actively psychotic. The tasks involved listening to auditory speech, silent Lip-reading (visual speech), and perception of meaningless lip movements (visual non-speech). Subjects also undertook a behavioural study of audio-visual word identification designed to evoke perceptual fusions. Patients and controls both showed susceptibility to audio-visual fusions on the behavioural task. The patient group as a whole showed less activation relative to controls in superior and inferior posterior temporal areas while performing the silent lip-reading task. Attending to visual non-speech, the patients activated less posterior (occipito-temporal) and more anterior (frontal, insular and striatal) brain areas than controls. This difference was accounted for Largely by the psychotic subgroup. Insular and striatal areas were also activated in both subject groups in the auditory speech perception condition, thus demonstrating the bimodal sensitivity of these regions. The results suggest that schizophrenia patients with psychotic symptoms respond to visually ambiguous stimuli (non-speech) by activation of polysensory structures. This could reflect particular processing strategies and may increase susceptibility to certain paranoid and hallucinatory symptoms.