The effect of organizational characteristics of classrooms on student friendship patterns is examined. Classrooms which vary in architectural settings, grouping procedures and pedagogical techniques are seen to place differential constraints on student interaction and to produce different patterns of relations. A number of hypotheses are tested in cross-sectional data from 51 classroom groups and longitudinal data from 9 groups. The results show that a more uniform distribution of choices with fewer isolates and sociometric leaders is found in the open classroom. Asymmetric dyads and intransitive triads occur less frequently and are less stable in the open setting. No clear differences in patterns of mutuality or group cohesion are detected in the two kinds of classrooms. (Author)