The possible association between demographic and drug history variables and tardive dyskinesia was explored in patients hospitalized at a facility where computerized records giving complete and detailed information on drug use were available. The variables studied included age, sex, diagnosis, time since first neuroleptic intake, and frequency of drug-free episodes. In addition, total intake, number of treatment days, average daily dose, and current intake were assessed for all neuroleptics, high potency and low potency neuroleptics, and antiparkinsonian and antidepressant medication. There was no significant relationship between most of the variables under investigation and tardive dyskinesia. Only three variables appeared to be significant risk factors: increasing age, time since first neuroleptic intake, and frequency of drug-free episodes. The finding of a correlation between the frequency of drug-free episodes and tardive dyskinesia is of particular interest and may have implications for our approach to the use of drug holidays.