A total of 89 depressed out-patients were randomly allocated to one of three groups: group A received one dose of antidepressant medication at night; group B received three doses of medication during the day; group C were allowed to choose either A or B above. Compliance with medication was assessed at three, six, nine and 12 weeks by interrogation and pill count; at the same time, depression and side-effects were rated. No overall significant difference was found between doctor-prescribed and patient-chosen regime, or between once-a-day and three-times-a-day dosage. However, compliance was significantly better in those patients who were allowed to choose, when they selected the three-times-a-day regime. There was a significant decline in compliance for all regimes over the 12 weeks. There was no evidence that better compliance produced a better therapeutic result, and possible reasons are given for this finding.