An argument is made against the use of school averages as composition measures by documenting the non-random nature of peer associations and by presenting evidence that different students are influenced by different reference groups. The structure of friendship associations among some 20,000 students in the McDill Twenty High School Sample is examined to illustrate the sources of homogeneity among high school acquaintances. These data are then examined to show that different students in schools are influenced by different reference groups that are visible and meaningful to them. The study concludes that school averages are insufficient and substantively weak measures of how individual students experience their school situation. (Author/EVH)