British Journal of Psychiatry (1992),160, 385—389 Factor Analysis of the Level of Expressed Emotion Scale, A Questionnaire Intended to Measure ‘¿ Perceived Expressed Emotion' COBY GERLSMA, PETRA M. van der LUBBE and CHIJS van NIEUWENHUIZEN When the factor structure and psychometric qualities of the Level of Expressed Emotion scale, an instrument intended to assess patient's perceptions of expressed emotion, were evaluated, three moderately intercorrelated factors emerged, with good internal consistency; these were lack of emotional support, intrusiveness/control, and irritability. This did not quite correspond to the a priori scales described in the original version. As in the original LEE, the three factors add to a total score intended to measure ‘¿ perceived expressed emotion'. The concept of expressed emotion (EE) (Brown & Rutter, 1966) refers to the amount of criticism, hostility, and emotional overinvolvement expressed by a psychiatric patient's relative. EE has become a well established predicting variable for relapse in schizophrenia, and in unipolar and bipolar depression (Leff & Vaughn, 1984; Kuipers, 1987; Falloon, 1988; Hooley & Teasdale, 1989). EE also appears to be relevant to adherence to treatment in cases of bulimia and anorexia nervosa (Falloon, 1988), and to long a term outcome of behaviour therapy with agora phobia (Peter & Hand, 1988). Expressed emotion is measured by means of a semistructured interview with the patient's ‘¿ significant other' (usually partner or parent), the Camberwell Family Interview (CFI; Vaughn & Leff, 1976b). Owing to its length and the complexity of scoring, as well as the need for a key relative, the CFI is an impractical instrument. The time and money involved ‘¿ @ in its administration preclude a routine application in many settings other than research institutes. This has prompted some researchers to design less cumbersome measuresofEE. Magailaeta!(1986), for instance, introduced assessment of EE based on five-minute speech samples (FMSS) of patients' relatives. Cole & Kazarian (1988) designed a questionnaire intended to assess patient's perceptions of EE, the Level of Expressed Emotion Scale (LEE). Obvious advantages of the questionnaire approach are that the assessment of EE becomes much easier, and that the patient's relative is no longer needed. An apparent disadvantage is the introduction of possible response bias by shifting focus from experimenter-observed EE to patients' perceptions of EE. Although it may seem at first sight hazardous to change the source of information from relative to patient, this change does not seem to decrease the relevance of the EE concept with regard to relapse. Hooley & Teasdale (1989) showed that the simple question “¿ How critical is your spouse of you?― accounted for more variance in relapse rates than EE (measured with the CFI) and marital distress (measured with the Dyadic Adjustment Scale) could explain together. Moreover, as Hooley & Teasdale (1989) suggested, perceived EE may provide importantadditionalinformation:adiscrepancy between experimenter-observed EE on the one hand and patient-perceived EE on the other has clear implications for intervention, since a patient-focused approach may be more appropriate than a family focused one. Thus, a reliable and psychometrically sound instrument assessing perceptions of EE could contribute to research and practice in this area. Kazarian et a! (1990) reported results comparing LEE scores with CFI ratings. The relationship between the two measures was encouraging as far as total scores were concerned, but the correlational pattern of a priori LEE subscales with CFI categories was inconsistent. This apriori structure of the LEE has not yet been evaluated by means of factor analysis, and the factorial structure of the LEE is still unclear. As Briggs & Cheek (1986) have shown, working with unvalidated a priori scales may well distort correlational patterns with concurrent measures and obscure conceptual distinctions. The present study examined the factor structure of the LEE scale and the internal consistency of the factors found. In view of the large numbers of subjects required for factor analyses, data were gathered from healthy subjects from the general population. Although EE has not been studied in a healthy sample before, there is no reason to expect that the EE categories of criticism and emotional overinvolvement are exclusive characteristics of psychiatric patients' relatives, even if the patients' symptoms may be powerful triggers to elicit high-EE behaviour (Kerver, 1990; but see also Goldstein eta!, 1989). 385