OBJECTIVE: To explore the issue of diagnostic specificity for major depression in the primary care setting by examining the relative accuracy of 3 methods to detect major depression in primary care. DESIGN: Comparison of performance characteristics of 3 case-finding methods for depression (ie, the Center for Epidemiologic Studies-Depression scale [CES-D], unaided physician detection, and "augmented" physician detection by use of a case-finding instrument), with the standard criterion being the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-III-R). SETTING: The offices of 50 family physicians from private and academic practice in southeastern Michigan. PATIENTS: Adult patients (N = 1580) who presented for routine care, from which a weighted random sample of 425 patients completed the Structured Clinical Interview for DSM-III-R. MAIN OUTCOME MEASURES: Sensitivity, specificity, positive predictive value, and positive likelihood ratio for each case-finding method. RESULTS: Major depression was present in 13.4% of the sample. Both the CES-D and unaided physician detection methods performed poorly in identifying patients who met DSM-III-R criteria for major depressive disorder. The CES-D had high sensitivity but low specificity at standard and high cut points, resulting, respectively, in low positive predictive values (0.307 and 0.385) and low positive likelihood ratios (2.9 and 4.0). Unaided physician detection showed lower sensitivity, higher specificity, and a slightly higher positive predictive value (0.45) and positive likelihood ratio (4.9). Raising of the CES-D threshold for a positive test did not enhance the detection of depression. Augmented physician detection with CES-D scores resulted in minimal improvement. Although the positive predictive value and positive likelihood ratio increased to 0.50 and 6.1, respectively, using the most stringent case-finding definition (ie, physician identification plus the CES-D score [score > or = 22]), the proportion of depressed patients who were correctly identified decreased to 26.9%. CONCLUSIONS: Neither high scores on the CES-D nor unaided physician detection accurately identified patients with major depression who were seen in primary care settings, while the supplementation of physician detection with CES-D scores had a minimal net effect on the accuracy of detection. The data do not support the routine use of the CES-D as a primary care screening instrument for depression, either as a stand-alone measure or as a supplement to physician detection.