We put forward a new experimental economics design with monetary incentives to estimate students' perceptions of grading discrimination. We use this design in a large field experiment which involved 1,200 British students in grade 8 classrooms across 29 schools. In this design, students are given an endowment they can invest on a task where payoff depends on performance. The task is a written verbal test which is graded non anonymously by their teacher, in a random half of the classrooms, and graded anonymously by an external examiner in the other random half of the classrooms. We find significant evidence that students' choices reflect perceptions of biases in teachers' grading practices. Our results show systematic gender interaction effects: male students invest less with female teachers than with male teachers while female students invest more with male teachers than with female teachers. Interestingly, female students' perceptions are not in line with actual discrimination: Teachers tend to give better grades to students of their own gender. Results do not suggest that ethnicity and socioeconomic status play a role. Appended are: (1) Survey Questionnaire; and (2) Experimental Procedure. (Contains 13 tables and 17 footnotes.)