OBJECTIVE: To compare tobacco control practices of physicians and their staff in Intervention communities with those in Comparison communities of the Community Intervention Trial for Smoking Cessation (COMMIT). DESIGN: COMMIT was a randomised trial testing community-based intervention for smoking cessation carried out over four years. SETTING: Eleven matched pairs of communities assigned randomly to Intervention and Comparison conditions. PARTICIPANTS AND INTERVENTIONS: Physicians in the Intervention communities participated in continuing medical education (CME). Training for office staff focused on tobacco control and office intervention "systems". OUTCOME MEASURES: Smoking control attitudes and practices reported by primary-care physicians in the 22 communities, smoking policies, and practices of 30 randomly selected medical offices in each community, and patient reports of physician intervention activities. RESULTS: Response rates to the physicians' mail survey were 45% and 42% in Intervention and Comparison communities, respectively. Telephone interviews of office staff had response rates of 84% in both conditions. Physicians in Intervention communities were more likely to attend training than those in Comparison communities (53% and 26%, respectively (P<0.0005)). In both conditions, training attendees perceived themselves as being better prepared to counsel smokers than non-attendees (P < or = 0.01) and reported more activity in smoking intervention. Intervention communities carried out more office-based tobacco control activities (P = 0.002). Smokers in Intervention communities were more likely to report receiving reading material about smoking from their physicians (P = 0.026). No other differences in physician intervention activities were reported by smokers between the Intervention and Comparison communities. CONCLUSIONS: The COMMIT intervention had a significant effect on some reported physician behaviours, office practices, and policies. However, most physicians still did not use state-of-the-art smoking intervention practices with their patients and there was little, or no, difference between patient reports of intervention activities of physicians in the Intervention and Comparison communities. Better systems and incentives are needed to attract physicians and their staff to CME and to encourage them to follow through on what they learn. The recently released Agency for Health Care Policy and Research clinical practice guideline for smoking cessation and other standards and policies outline these systems and offer suggestions for incentives to facilitate adoption of these practices by physicians.