Background A key problem in the management of first episode psychosis is that patients are often reluctant to take antipsychotic medication, especially once their presenting symptoms have resolved. Clinicians may be tempted to trial a ‘break in treatment’ in such patients. Aim To assess the impact of interruptions in the antipsychotic treatment of first episode psychosis. Method Treatment adherence and clinical course were assessed during the 18 months following presentation in 136 consecutive patients with a first episode of psychosis in 2003–2005 by a systematic retrospective casenote review. Regression analyses were used to examine the time to remission and the risk of relapse in patients who had stopped antipsychotics for one month or more. Results There were breaks in antipsychotic treatment of ≥ 1 month in more than half of the patients (n = 73; 58%). When these occurred before they had recovered (n = 22; 17%), the time to remission was almost twice as long as in patients in whom treatment was continuous (t = 2.9, P = 0.01). Patients in whom treatment was interrupted were 5 times more likely to have relapsed than those in whom it was continuous (p = 0.0001, 95%CI 2.1–11). The mean time to relapse following an interruption in treatment was 3 months. Conclusion If the treatment of first episode psychosis with antipsychotic medication is stopped for a month or more, remission may be delayed and the risk of relapse following remission may be substantially increased.