We run a randomized control trial in an Italian university to study the effect of test anxiety on a high stakes exam. We separate students in two groups and we expose them to two random treatments, silence and music, that influence their level of pre-test anxiety. We measure the variation of test anxiety by observing the difference in individual biomarkers collected before and after the treatments. We find that a reduction in the mean arterial pressure and systolic pressure improve females test scores, and the effect is much stronger if the treatment is silence. For males we do not find any significant effect. Hence, we conclude that test anxiety may help to explain gender differentials in performance.