The clinical picture, risk factors and natural history of tardive dystonia resulting from dopamine-receptor antagonist (DRA) treatment in 107 patients (57 male and 50 female), seen between 1972 and 1995, are described. The mean age at onset (+/- SD) was 38.3 +/- 13.7 years (range 13-68 years), and the age at last follow-up was 46.3 +/- 15.7 years (range 15-80 years). These patients had received DRAs for schizophrenia (39%), for other psychiatric conditions (51.5%) and for non-psychiatric disorders (9.5%). All classes of neuroleptics used were implicated in producing tardive dystonia, which was found to develop at any time, ranging from 4 days to 23 years after their introduction (median 5, mean 6.2 +/- 5.1 years); there was no 'safe' period. Men were significantly younger than women at onset of dystonia, which developed after shorter exposure in men. At onset, the dystonia was focal in 83% of cases, but progressed over months or years and remained focal in only 17% at the time of maximum severity. The craniocervical region was involved in 87% of cases, and was the most commonly affected site both at onset and at maximum severity. There was a correlation between the site and age of onset; the site of onset ascended from the lower limbs to the face as the mean age of onset increased. Overall, the phenomenology of tardive dystonia was indistinguishable from that of primary (idiopathic) dystonia, although retrocollis and anterocollis, as well as torticollis to the right, were significantly more common in tardive dystonia. It is a very persistent disorder; only 14% of our patients had a remission over a mean follow-up period of 8.5 years. Remission occurred after a mean of 5.2 years from onset (range 1-12 years) and 2.6 years after discontinuation of neuroleptics (range 1 month to 9 years). Discontinuation of neuroleptics increased the chances of remission fourfold. Patients with < or = 10 years on neuroleptics had a five times greater chance of remission than those with > 10 years exposure, suggesting that the pathogenetic changes in tardive dystonia may become irreversible after long-term use of these drugs. None of the numerous treatments tried in these patients, including clozapine and botulinum toxin injections, seemed to relate to overall outcome, but there was a significant negative association between the occurrence of remission and the use of benzodiazepines. Although there were hints of a possible genetic predisposition, the question as to whether patients with tardive dystonia have an underlying vulnerability remains unanswered.