Three interactional analysis (IA) systems (Bales', Roters modified Bales and Stiles' "Verbal response modes") were used to characterize behavioral elements of provider-patient dialogues of 101 new-patient visits in a general medical clinic. In a previous article, the explanatory power of these IA systems was compared. In this article, specific provider and patient behaviors within segments of the encounter (introduction-history, physical examination and conclusion), which were shown to be related to encounter outcomes of knowledge, compliance and satisfaction, were examined. Review of interactional behaviors entering regression analysis with a significant F-to-enter (p less than or equal to 0.05) and supplementary contextual analyses suggested the importance of several categories of physician and patient behavior. Behaviors manifesting tension bear important and complex relationships to encounter outcomes. For example, patient and physician expressions of tension generally bear strong negative relationships to patient satisfaction, while patient expressions interpreted as tension release are positively related to both satisfaction and compliance. The timing of other behaviors appears to be critical to subsequent outcomes. If patient requests for medication occur early in the encounter, this behavior is positively related to subsequent patient satisfaction. However, if they occur in the concluding segment, a negative relationship results. Finally, several relationships taken together indicate that physician teaching in the concluding segment may be important. While useful observations may emerge from application of currently available IA techniques, the resulting information is best characterized as hypothesis-generating. These IA systems have many limitations, and research is needed to derived more clinically oriented systems that may permit more consistent demonstrations of critical process-outcome relationships.