The grades that students receive in school have been used to determine important social outcomes like educational tracks, scholarships, and even wages. However, grades do not always accurately reflect student achievement. Studies show, for example, that teachers discriminate in grading based on students’ ascribed characteristics (such as race or caste). The literature is largely silent, however, on whether teachers discriminate based on students’ achieved characteristics. In particular, we hypothesize that grades are not only the outcome of grading discrimination. Instead, they may also be the cause of grading discrimination or amplify discrimination on the basis of other characteristics. To test this hypothesis, we conducted a field experiment in which teachers graded a series of essays with randomly assigned student characteristics of (either low or high) prior grades crosscut with gender (girl or boy). We find that essays ostensibly written by students with low as opposed to high prior grades are graded 0.55 standard deviations (SDs) lower. Moreover, we find a negative interaction effect between prior grades and gender: teachers impose an additional penalty when essays are ostensibly written by girls with low prior grades.