Alternative measures of expressed emotion: a methodological and cautionary note.

Undergraduate raters listened to Camberwell Family Interviews that had been conducted with the spouses of depressed patients and then rated each relative with a rating-scale assessment of expressed emotion (EE). Students' ratings of relatives' criticism, hostility, emotional overinvolvement, and warmth were significantly correlated with trained raters' EE assessments obtained in the conventional manner. Despite this correspondence, further analyses revealed that undergraduates' assessments of relatives did not predict 9-month relapse rates in patients. These results highlight the importance of establishing both the concurrent and predictive validity of any alternative measure of EE. They also emphasize the dangers of assuming that significant correlates of EE are necessarily significant predictors of relapse.