Access to higher education is one important prerequisite for later employment possibilities. Often access is regulated inter alia by multiple-choice entrance exams. The application of this testing format is problematic if it favors the answering strategies of certain groups. We present the results of a field experiment in answering multiple-choice questions. Our sample consists of 2113 pupils from different school types. We find that girls skip more answers than boys only if the questions are difficult. This gender gap vanishes when extrinsic rewards are provided. This suggests that our findings are compatible with a stereotype threat explanation. Moreover, the gender gap is found only for pupils of school types preparing for the academic track. It is therefore important to consider all social strata in the design of testing formats.