We examined the state-dependent contribution of neuroleptic withdrawal and psychotic relapse in influencing sleep measures. Eighteen clinically stable male schizophrenic patients taking haloperidol were studied with 3 nights of polysomnography for baseline measures and again after neuroleptic withdrawal. Sleep measures were also obtained at the point of relapse (n = 9) or after a 6-week drug-free period if the patient remained clinically stable (n = 9). Neuroleptic withdrawal led to a global deterioration of rapid eye movement and non-rapid eye movement sleep and a reduction of rapid eye movement latency in both groups. Relapsers differed from nonrelapsers in that they had a larger decrease in total sleep time, sleep efficiency, total non-rapid eye movement sleep, and stage 2 sleep. The level of psychosis was inversely correlated with sleep efficiency, total sleep time, and stage 4 sleep in the drug-free patients. Our data suggest that clinical state needs to be identified in sleep studies of drug-free patients.