In this paper, we examine the performance of a pencil-and-paper screening questionnaire on depressive symptoms (the Center for Epidemiologic Studies Depression Scale, or CES-D) in a sample of 120 adult American Indians belonging to a single Northwest Coast tribe. Results of factor analyses suggest that somatic complaints and emotional distress are not well differentiated from each other in this population. CES-D scores (which have shown good sensitivity and specificity for depressive disorders in this sample) also show weak and apparently nonsignificant trends to be elevated in the presence of other psychiatric diagnoses (including alcoholism) or general impairment. However, because of the use of a convenience sample (rather than a probability sample), analyses of associations between study factors--including comorbidity--are liable to produce spurious results due to selection bias (including Berkson bias). On this basis, we suggest that the use of probability samples should assume a high priority in cross-cultural studies. The study of the entire population of interest is another solution to the sampling problem, particularly in small communities.