Although many educational institutions allow students to resit exams, a recently proposed mathematical model suggests that this could lead to a dramatic reduction in study-time investment, especially in rational students. In the current study, we present a modification of this model in which we included some well-justified assumptions about learning and performance on multiple-choice tests, and we tested its predictions in two experiments in which participants were asked to invest fictional study time for a fictional exam. Consistent with our model, the prospect of a resit exam was found to promote lower investments of study time for a first exam and this effect was stronger for participants scoring higher on the cognitive reflection test. We also found that the negative effect of resit exams on study-time investment was attenuated when access to the resit was made uncertain by making it probabilistic or dependent on obtaining a minimal, non-passing grade for the first attempt. Taken together, these results suggest that offering students resit exams may compromise the achievement of learning goals, and they raise the more general implication that second chances promote risky behavior.