This study is a field verification of one of the most important hypotheses concerning decision making under risk and uncertainty: the loss aversion. Towards this goal, the framing of grading rules was differentiated in two exams at the University ofWarsaw, with only half the students facing explicit penalty points in the case of giving an incorrect answer. In fact, the numbers of points per answer and grade thresholds were adjusted such that the two treatments were fully equivalent. Loss aversion predicts that less risk will be taken (less questions will be answered) in the Loss Treatment. In fact, no treatment effect was observed. This null result was not due to ceiling effect as nearly two-thirds did leave at least one question unanswered. Interestingly, there was a large gender effect in both treatments, with males trying to respond to more questions (despite their inability to actually pick the right answer more often). This is fully in line with earlier results showing males as more overconfident and less risk averse.