In a semi-replication study, 103 videotaped real-life general practice consultations of patients with hypertension were observed with Roter's interaction Analysis System (RIAS). RIAS consists of a detailed category system meant to measure each verbal utterance of physician and patient (distinguished in taskrelated behavior and socio-emotional behavior) and a set of global affect-ratings. In this article, only general practitioner (GP) behavior is studied. GP's behavior is related to panel-assessed quality of care on three separate dimensions (technical-medical, psychosocial, and the management of the physicianpatient relationship). A remarkably high percentage of the variance in the quality assessments (ranging from 59% to 70%) was explained by RIAS. The global affect-ratings proved to have the strongest influence in all quality assessments. In addition, task-related behavior seems to be more important in medical technical behavior, whereas socio-emotional behavior, and especially the psychotherapeutic categories like reflecting, paraphrasing, showing agreement, and others, seem to be more important in the other quality measures. The results are compared with Roter's study; similarities and differences are discussed in light of adjustments in the methodology. A plea is made for cross-cultural comparisons in physician behavior.