Problem-based small-group discussions are the cornerstone of health sciences education at the University of Limburg (The Netherlands). In each of three courses, fifteen discussion groups of about eight students were randomly assigned a staff-tutor (control condition) or student-tutor (experimental condition).In two of the courses no significant differences in cognitive test achievement between the two conditions were found. In one course students tutored by staff-tutors performed significantly better than students tutored by student-tutors. However, no significant differences in test performance were found between students tutored by high-achieving versus average-achieving studenttutors. These findings contradict the "congruence" hypothesis from information processing theory. Several other explanations are explored.Health sciences education at the University of Limburg (Maastricht, The Netherlands) features problem-based small-group discussions alternated with periods of self-directed learning. This problem-based self-directed learning method is designed to teach problem-solving skills, self-learning skills and enhance motivation and knowledge retention (Schmidt and De Volder, 1984). One of the cornerstones of this method is the discussion group consisting of about eight students in which health science phenomena (so-called problems) are analyzed and learning goals are formulated (Schmidt, 1983). These groups are guided by tutors who are faculty members. Their role is to stimulate the discussion, not to give lectures. In 1981 it was decided to explore the possibility of assigning the role of tutor not only to faculty staff but also to (undergraduate) students~ The main reason for this was what Goldschmid and Goldschmid (1976, p. 14) called "economic considerations": "Given today's high student-faculty 0018-1560/85/$03.30 9 1985 Elsevier Science Publishers B.V.