It has been suggested that low frequency transcranial magnetic stimulation (TMS) over left temporo-parietal cortex may reduce the frequency and intensity of auditory hallucinations in schizophrenia. Sixteen patients with hallucinations, treatment-resistant for at least 2 months, were randomised into a placebo-controlled crossover study of TMS at 1 Hz and 80% of motor threshold over left temporo-parietal cortex. Treatment periods lasted for 4 days, with daily duration escalating from 4 to 8, 12 and 16 min on subsequent days. Each minute of stimulation was followed by 15 s of rest to check coil position and allow the patient to move, if necessary. Both patients and symptom raters were unaware of the treatment condition. Patients' hallucination scores improved from baseline with both real and sham TMS, and there was no significant difference between real and sham treatments. There was a trend for second treatments, whether sham or real, to be more effective than first treatments. Other psychopathology scales (apart from positive symptoms) and verbal memory were not affected by real or sham TMS. Previous positive studies could not be replicated with these parameters. TMS is safe if applied within the protocol used.