Objective: Self-administered questionnaires are frequently used in studies of hospitalized physically ill patients to identify and measure psychiatric morbidity. This study examines the validity of some commonly used questionnaires in this context. Method: One hundred and seventy-nine patients in a general hospital completed the GHQ, HAD, BDI and STAI and were interviewed using the SCID-R. The findings were evaluated against DSM-III-R diagnoses using the QROC curve. Results: Thirty-eight percent of patients obtained one or more DSM-III-R diagnoses; 25 percent mood disorder, 12 percent anxiety disorder, 11 percent drug abuse or dependence, 2 percent a somatic syndrome. Eleven percent had more than one diagnosis. As screening instruments for general psychiatric morbidity there were no statistically significant differences between versions of the GHQ and the HAD. With respect to identifying depression, the GHQ tended to perform better than the other instruments. Conclusions: The questionnaires identified general morbidity and depression satisfactorily but anxiety and drug abuse and dependence syndromes poorly. The results also support the notion that the DSM-III-R classification has a number of deficiencies when used in this population.