In the context of a prospective, controlled treatment study, contrasting family interventions with individual treatment, the role of expressed emotion (EE) as a predictor of relapse was examined in patients with recent-onset schizophrenia and related disorders (n = 97). EE was compared with 13 predictor variables. The variables, taken from EE and family intervention studies, related to demography, premorbid functioning, present and past illness history, and comorbid substance abuse. Psychotic relapse was operationalized with a conservatively measured relapse criterion, composed of monthly ratings based on the Brief Psychiatric Rating Scale and on clinical judgment during the 12 months of outpatient treatment. Of the 14 predictor variables entered in stepwise survival analyses, 6 variables had probable predictive power on the conservative relapse criterion. These variables were entered in a Cox regression model. EE turned out to be the major predictor of relapse in the overall sample (hazard ratio [HR] 4.90; confidence interval [CI] 1.05-22.92). This finding remained when only patients with a first psychotic episode (p = 0.02) and patients in the individual treatment condition (p = 0.001) were examined. Cannabis abuse was the major predictor of relapse in patients with high-EE families (HR 4.27; CI 1.12-16.29).