The expressed emotion (EE) measure is a robust predictor of relapse in schizophrenia. At least twelve studies from a range of cultures have now shown this, although the authors of one maintain their findings are the result of confounding factors. A thirteenth study which failed to demonstrate a predictive effect was seriously weakened by methodological defects. There is now also evidence for the construct validity of EE, which has been shown to reflect ongoing family interactions. High EE relatives display less effective coping responses, and may adversely affect the patient by creating an unpredictable environment. EE may change with time, and probably reflects the interaction of external stresses, the real problems with the patient, and inappropriate coping responses. The association of high EE with arousal in the patient, as indicated by skin conductance abnormalities, has now been replicated. Although the EE phenomenon can be subsumed under a vulnerability-stress model, the way in which it might mediate relapse has not as yet been specified. Interventions have been successful in reducing relapse, and do so by increasing the tolerance and coping skills of the relatives, and by establishing more realistic expectations. As most have also involved the patient in the process of treatment, they have not fully resolved the question of causal direction. Future studies will clarify these issues, but will not detract from the value that EE has already had in energizing research. It has also provided a framework within which clinicians have begun to recognize and deal with the disabling problems that beset schizophrenic patients and their families.