Sodium valproate, a GABAergic agent (800 mg), and placebo were administered orally, as a single dose, to nine chronic schizophrenics with tardive dyskinesia (TD), seven chronic schizophrenics without TD and ten healthy controls, according to a double blind design. Blood samples were collected before and after drug administration, to determine plasma prolactin concentrations. Sodium valproate decreased plasma prolactin levels in healthy subjects (P less than 0.001) and in schizophrenic patients with TD (P less than 0.001), but not in chronic schizophrenics without TD. Moreover, in dyskinetic subjects, the maximum per cent decrease of plasma prolactin from basal value was positively correlated to the score of the abnormal involuntary movement scale (r = 0.724, P less than 0.02). Although the neural or biochemical substrate underlying the different responses of plasma prolactin to sodium valproate in schizophrenics with and without TD remains unclear, these results provide the first neuroendocrine evidence able to differentiate dyskinetic subjects from those without TD within a schizophrenic population.