Reports of perceived spousal criticism better predict relapse following recovery from major depressive disorder than do measurements of 'actual' spousal criticism from Expressed Emotion interviews coded by outside raters (Hooley & Teasdale, 1989). This study examines relations between perceived and 'actual' criticism in depressed and non-depressed spouses. Married couples were evaluated using structured interviews, questionnaires, and an observationally-coded dyadic interaction. Signal detection methods assessed sensitivity and bias in reports of perceived criticism. Results suggested that depressed and martially discordant spouses are biased towards perceiving spousal criticism, suggesting that criticality bias reflects a latent cognitive style underlying depression and marital discord. That bias was related to reports of global perceived criticism suggests that spousal reports of perceived criticism capture both actual criticism and reporters' cognitive bias towards over- or under-perceiving criticism. Reports of perceived criticism may better predict relapse from depression than ratings of actual criticism in part because they capture such cognitive tendencies.