The objectives of this study were to examine patterns of resident-patient communication and the relationship between resident patterns of speech with patient satisfaction. Forty consultations, ten in each of the four gender combinations (male resident/male patient, male resident/female patient, female resident/female patient, female resident/male patient) were audiotaped and microanalyzed using the Roter Interaction Analysis System. Several findings depart significantly from previous studies with physician-only or physician-resident-mixed samples. First, the average length of the 40 consultations was 19.5 minutes, 11.3 minutes longer than consultations in a physician-only sample drawn in the same clinic previously. Second, male residents engaged in twice as much psychosocial talk as female residents and conducted longer consultations. Third, residents asked 80% of the total questions while patients asked 20% of the questions. Previous studies with physician-only or physician-resident-mixed samples reported that physicians ask 89-99% of the total questions. Finally, patients' overall satisfaction and communication satisfaction were negatively correlated with residents' positive talk, which constitutes 31% of a given resident's total utterances. In the study conducted in the same clinic with a physician-only sample, physician positive talk was 26% and physician positive talk was not correlated with patient satisfaction. Is this a signal that residents should reduce the amount of positive talk? Apparently more studies with resident-only samples are needed to answer this and other unanswered questions in the field to offer directives to resident training.