Preventive behavior in a sense of self-protection is defined as the action taken by an agent, at a cost, to mitigate the probability of an adverse consequence. Yet, even though these behaviors are widespread in economic life, it has never been studied in a controlled environment. We propose to do so by first clarifying the theoretical predictions in various theoretical frameworks (expected utility, rank-dependent utility, prospect theory), and second to test experimentally these predictions. More specifically, we study the cost that makes the decision-maker indifferent with a fixed change of the probability of the adverse consequence. At the theoretical level, we show that under expected utility, the intensity of prevention action decreases with the probability of the adverse consequence, while for rank-dependent expected utility and prospect theory, the main determinant of prevention is the derivative of the probability weighting function, leading to a U shape form of prevention. Experimentally, our results suggest that prevention is mostly driven by the shape of the probability weighting function, a result that seemed to have been overlooked in the literature, that has mostly focused on the curvature of the utility function