To determine the usefulness of the Center for Epidemiological Studies Depression Scale (CES-D) as a measure of depressive symptoms among frail and disabled elderly, we conducted a first- and second-order confirmatory factor analysis testing the four-factor structure of the CES-D and the existence of a single underlying second-order factor. We also examined the relative contribution of the somatic factor of the total symptom score, and relationships of age, race, functional status, and health with the four factors. We found that the factor structure is replicated when used with frail elderly, and scores are not distorted by the disproportionate influence of somatic items. Age and health do not affect the somatic subscale more than the other three subscales. Functional limitations are associated with higher somatic scores. We conclude that the CES-D is an appropriate tool to measure depressive symptoms in frail elderly.