To test and further develop a causal model of the influence of tutor behaviors on student achievement and interest in the context of problem-based learning. Data from 524 tutorial groups involving students participating in the four-year undergraduate health sciences curriculum at the University of Limburg in 1992-93 were analyzed. The tutorial groups were guided by 261 tutors. Overall, 3,792 data records were studied, with each student participating in an average of 2.3 groups. Correlations among tutors' social-congruence, expertise-use, and cognitive-congruence behaviors, small-group functioning, and student' self-study time, intrinsic interest in subject matter, and level of achievement were analyzed using structural-equations modeling. This statistical technique allows the investigator to test causal hypotheses on correlational data by comparing the structure of data with a theoretical model. After minor adaptations, the hypothesized causal model of the effective tutor fitted the data extremely will. Each tutor's level of expertise use and social congruence not only directly affected his or her level of cognitive congruence but also affected other elements of the model. Level of social congruence influenced group functioning in a direct fashion, while expertise use had a slightly negative effect on the students' level of self-study time and a slightly positive effect on level of achievement. As hypothesized, the level of cognitive congruence influenced tutorial-group functioning. Level of group functioning affected self-study time and intrinsic interest. Finally, time spent on self-study influenced level of achievement. The results suggest that subject-matter expertise; a commitment to students' learning and their lives in a personal, authentic way; and the ability to express oneself in the language used by the student are all determinants of learning in problem-based curricula. The theory of the effective tutor, presented in this article, merges two different perspectives prevalent in the literature. One perspective emphasizes the personal qualities of the tutor: his or her ability to communicate with students in an informal way, coupled with an empathic attitude that enables the tutor to encourage student learning by creating an atmosphere in which open exchange of ideas is facilitated. The other stresses the tutor's subject-matter knowledge as a determinant of learning. The data presented in this article suggest that what is needed, really, is much of both.