Background: The purpose of this study was to learn more about barriers to managing depression by comparing knowledge and attitudes about depression among physicians, internists, obstetrician-gynecologists, and a reference group of psychiatrists. Among the non-psychiatrists, we hypothesized that generalist physicians would have more favorable attitudes and greater knowledge about depression than non-generalists., Methods: Survey questionnaires were sent to resident and faculty physicians (N = 375) of two university-affiliated medical centers. The physicians were classified as non-generalists (medicine subspecialists, transitional year interns, and obstetrician-gynecologists), generalists (general internists and family physicians), and psychiatrists. A 33-item written questionnaire assessed knowledge and three attitudinal dimensions: attitudes attributed by physicians to patients; physicians' confidence in managing depression; and physicians' psychosocial orientation. A knowledge scale and an attitudes scale were scored by adding the number of knowledge items answered correctly and the more favorable attitudinal responses. Multivariable regression was used to identify physician characteristics among non-generalists and generalists associated with higher knowledge and attitudinal scores., Results: Response rate was 82%. Sixty percent of the respondents were male, 63% were resident physicians, and 14% had advanced psychosocial training. Non-generalists and generalists had similar demographic characteristics, but psychiatrists were significantly more experienced. Psychiatrists had the most favorable attitudes, followed by generalists and non-generalists. Compared with non-generalists, generalists were more confident in prescribing antidepressants (62% vs 25%), more likely to report that treating depression is rewarding (71% vs 39%), and less likely to refer to a psychiatrist (58% vs 79%). Generalist classification, increased experience, and higher levels of psychosocial training were associated with more favorable attitudes. Knowledge scores were significantly higher for psychiatrists than for non-generalists and generalists. Among non-psychiatrists, correct responses for knowledge items were: treatment efficacy (61%), treatment duration (59%), > or = 5 DSM-III-R criteria (52%), and prevalence of depression (30%). Among those with incorrect responses, both non-generalists and generalists overestimated the prevalence (52%) and underestimated the efficacy of drug therapy (30%)., Conclusions: Generalists and non-generalists have similar and relatively good basic knowledge about depression. Misperceptions about treatment efficacy, and attitudinal barriers, particularly among non-generalists, may compromise the physician's ability to diagnose and manage depression.