In two recent cases involving the University of Michigan (Gratz v. Bollinger and Gruttinger v. Bollinger), the Supreme Court examined whether race should be allowed to play an explicit role in the admission decisions of schools. The arguments made in support of affirmative action admission policies in these cases and others raise two fundamental questions. First, do students actually have incorrect beliefs about individuals from different races at the time of college entrance? Second, if students do have incorrect beliefs at the time of college entrance, can diversity on a college campus change these beliefs? While a small literature has recently shed some light on the second question, no previous work has been able to provide direct evidence about the first one. In this paper we examine the first question by taking advantage of unique data collected specifically for this purpose.