BACKGROUND. This paper investigates individual patient characteristics predicting differential response to each of three physician-delivered smoking interventions after 6 months. METHODS. Participants were 1,286 currently smoking patients seen by 196 medical and family practice residents in five primary care clinics affiliated with the University of Massachusetts Medical School. Of the participants, 57% were female, 92% were white, their average age was 35 years, and they smoked an average of 23 cigarettes per day. Physicians were trained to provide the following interventions: advice only (AO), a brief (< 10 min) patient-centered counseling intervention (CI), and counseling plus prescription of the nicotine-containing gum Nicorette (CI+NCG). The CI+NCG condition included NCG only when appropriate and if acceptable to the patient. Patients were randomized to one of these three physician-delivered intervention conditions. RESULTS. Overall, stratified univariate analyses revealed that AO produced consistently lower cessation rates across most subgroups (generally 9-13%) but was somewhat more effective for certain groups of lighter smokers. Relative to AO, CI was somewhat more effective (about 20-24%) for less addicted smokers, for those with more previous quit attempts, and for those with fewer close associates who smoke, but generally failed to produce higher quit rates for harder core smokers or for women. CI+NCG had an overall pattern of greater effectiveness for both more addicted and less addicted smokers, with the highest absolute levels of cessation (about 27-30%) among less dependent smokers. Women in this group had cessation rates (20.6%) comparable to those of men (23.6%). Condition-stratified logistic regression analyses, controlling for a wide range of covariates, revealed associations similar to those observed in the univariate analyses: An overall logistic model in which intervention conditions were fitted as dummy variables produced the following significant main effects: sex, years smoked, contact with other smokers, symptoms, and CI+NCG condition. Significant interactions were observed for both CI and CI+NCG and smoking when feeling too ill to continue normal activities and CI+NCG and amount smoked. CONCLUSIONS. We observed significant main effects on cessation of variables related to addiction, sex, social factors, and physician counseling interventions. Specific interactions were observed between reported smoking when feeling ill and each of the counseling interventions as well as by amount smoked in the CI+NCG condition.