Between-method comparisons of patient data from standardized self-report instruments and corresponding patient interview data frequently have poor agreement. We investigated the concordance and predictive validity of self-report measures and expert-rated interview data of women's subjective theories of illness in a psychosomatic-gynecological consultation liaison service. 31 patients completed two questionnaires (KKG, PATEF) on subjective theories of illness and were interviewed with a semistructured instrument on the same topic at the time of the initial appointment and after 1.5 years. External criteria for assessing the predictive validity of questionnaire versus interview data were: follow-up participation and psychotherapy utilization during the follow-up interval. Questionnaire and interview rating data were only modestly associated. Rater concordance on the interview data was low. Neither the initial-assessment interview nor the questionnaire data had predictive validity for patients' participation in the follow-up. None of the interview rating data but a portion of the questionnaire data, had predictive validity for patients' utilization of psychotherapy during the follow-up interval. The cognitive conceptualization of subjective theories of illness may fail in patients with psychosomatic disorders, due to their very mode of experience. A wider approach that includes assessment of nonverbal communication and countertransference might be useful.