This paper demonstrates that short-term peer exposure can generate achievement effects which persist for several months and years. I study a mandatory freshmen week for first-year undergraduates and exploit the random assignment of students to freshmen teams. I find that the freshmen week contributes to the formation of persistent social ties. Furthermore, peers' observable characteristics impact college achievement for up to three years. Ability peer effects are non-linear, i.e. very high or low levels of average peer ability in a group harm students' grades. These effects are most pronounced for low-ability students.