Recognition of depression in primary care is both important and difficult. To study recognition of depression, we monitored care delivered to new adult patients randomly assigned to primary care providers. At study entry, 508 patients completed the Beck Depression Inventory (BDI) and the Medical Outcomes Study Short-form Health Survey-36 (SF-36), a measure of health status. Chart notes were reviewed at the end of 1 year. Only 36 of 130 patients with elevated BDI scores > or = 9 (moderate-to-severe depression) were noted as depressed on the chart. Patient characteristics predicting chart notation of depression included BDI scores, health status, gender, and education. When controlling for these factors, neither age nor race were statistically significant in the prediction of the recognition of depression. Female patients were more likely to be diagnosed as depressed than men with comparable BDI and SF-36 scores. Greater patient education was associated with enhanced likelihood of diagnosis of depression. Both BDI scores and health status were important predictors of diagnosis of depression. All SF-36 subscales correlated highly with BDI scores, suggesting that these measures may lack adequate discriminant validity. Identifying diagnostic tendencies may help primary care providers improve detection of depression, a critical first step toward effective management.