When students make the transition from elementary school to junior high school, they experience abrupt organizational changes in their school environment. The hypothesis that social comparison processes mediate the relation between ability grouping practices in mathematics and students' achievement expectancies was tested in a longitudinal sample of 862 students making the transition from grade 6 to grade 7. Compared to between-classroom ability grouping, heterogeneous grouping raised the achievement expectancies of high achievers and lowered the achievement expectancies of low achievers. When controls for the direction of students' social comparison choices and for their mathematics grades were introduced, the independent effect of ability grouping on achievement expectancies was consistently and substantially reduced. It is argued that ability grouping practices constrain the choices available to students and teachers for social comparison of abilities and thereby influence the frame of reference students use for self-assessment and teachers use for assigning grades. Eight tables and four figures are appended. (YP)