A principal method of studying physician practice patterns has been to examine physicians' responses to brief written cases. We have compared this method with practice patterns of the same physicians derived from chart audit. Subjects were 98 family practice residents for whom data were available in actual patient encounters for the workup of asymptomatic hypertension. Short, carefully structured case reports using four cues were designed and a checklist similar to the one used for test ordering in practice was employed. Chart reviews and billing encounter forms were used for comparison. Results indicated residents ordered fewer tests in clinical practice, due, in part, to practice constraints not represented in the written cases. Physicians tend to make the diagnosis of hypertension incrementally in practice, with no one visit adequately representing the point of diagnosis. Studies based on data bases using a patient encounter as the unit of analysis in chronic disease such as hypertension may spuriously underestimate the actual number of tests ordered for the workup. Judgment cases may better reflect the patterns of use of information in a well-defined problem. Prediction of number of tests ordered in the clinical setting has not been established in this case.