A study tested two hypotheses aimed at explaining the "self-fulfilling prophesy" that results when students are divided into ability group systems. The differential instruction hypothesis explains achievement differentials produced by grouping systems in terms of the differential instruction received by students. The peer differentials hypothesis argues that differential outcomes in schools are influenced by patterns of peer allocation and influence. Subjects were fourth grade students in a single urban school. Two types of instructional grouping systems were analyzed: within-classroom grouping for reading instruction and across-classroom ability grouping. Although the self-fulfilling prophecy was at work in both grouping systems, the two forms of grouping affected the processes of peer allocation and differential instruction differently. Instructional processes and peer group formation tended to place students in lower strata of the cross-classroom ability grouping at a disadvantage, but to work in favor of students in the lower strata of the within-classroom reading group system. These results have implications for grouping research strategy and future research on grouping. (JL)