This paper presents a small theoretical model to compare school systems that segregate students by ability ("tracking") with comprehensive ones, which allow for mixing of differently skilled students into same classes. The outcomes of interest are the achievement levels of weaker and better students, and the average achievement of all students. In the model, the instructional pace is tailored to the skill distribution of a class, and higher-Achieving peers are an additional source of learning. The results show that differences in both the share of high-Achievers and degree of interaction between student types can explain the mixed (quasi-)experimental evidence on the effect of de-Tracking on student achievement. As changes in peer quality affect good and weak students' achievement in very different ways, the term "peer effect" should be used with caution. © 2017 Walter de Gruyter GmbH, Berlin/Boston.