This study examined the effectiveness of smoking cessation counseling by physicians-in-training (residents) with African-American patients. One hundred fifty-eight family and internal medicine residents at a large urban public general hospital participated in the study; two thirds of the residents underwent a 2-hour smoking cessation training program. Ninety-two of the trained physicians counseled from 1 to 18 patients. The majority of physicians were male, with 8% being current smokers. Over a 26-month period, 1086 patients were randomly assigned to intervention and control (usual care) groups. Mean patient age was 44 years, mean years smoking was 25, and mean number of cigarettes smoked per day was 14. There were no differences in biochemically validated smoking cessation rates between the intervention and control groups at 3 or 12 months postenrollment (2% versus 1.8% and 2.2% versus 2.8%, respectively). Losses to follow-up were high at both 3 and 12 months (38% and 40% respectively). Implications for future trials in minority populations are discussed. A brief physician-based smoking cessation message does not appear to be an effective strategy for use with African-American smokers in a large urban public general hospital.