To evaluate the clinical and ecological validity of affect recognition (AR) measures in a sample of community-dwelling schizophrenic outpatients (N = 40), we analyzed the relation of facial and vocal AR to intellectual, symptomatic, and quality-of-life criteria. Facial and vocal AR showed virtually identical patterns of association with these criteria, suggesting that both modalities of AR draw on the same underlying heteromodal capacity. Specifically, AR was correlated with a subset of intellectual abilities (verbal–semantic, executive–attentional), but was unrelated to age, education, or neuroleptic dose. In terms of clinical and ecological criteria, AR errors correlated with more severe psychotic symptoms (positive and disorganized) and with lower quality of life (relationships, community participation, and richness of intrapsychic experience). Even after controlling for subjects' intellectual abilities and illness severity, inaccurate AR was associated with bizarre behaviors (involving sociosexual interactions, clothing, appearance) and with impoverished interpersonal relations. Thus, while difficulty identifying basic affective cues is related to general cognitive and illness-severity factors, it appears to have specific functional implications that do not depend on generalized impairment. Assessment of AR may identify a subgroup of schizophrenic patients who have a central defect in the heteromodal monitoring of emotional-social displays, associated with dysregulation of social behaviors and disruption of interpersonal relations. (JINS, 2000, 6, 649–658.)