It is by now well understood that very clear incentive schemes can induce gaming: the exploitation of a scheme by an agent for his own self-interest, to the detriment of the objective of the scheme. Ederer, Holden, and Meyer (EHM, Rand Journal, 2018) showed theoretically in a two-task moral hazard model with privately known agent preferences that keeping an agent uncertain about the precise form of the reward scheme can mitigate gaming, which arises in that context in the form of socially inefficient focusing of effort on the agent’s privately-preferred task. By committing to randomly determine which of the two tasks will be rewarded, the principal induces a risk-averse agent to self-insure against the randomness, by choosing a more balanced profile of efforts than would otherwise be optimal for him. But stochastic determination of the task to be rewarded also has drawbacks: The total effort induced is lower, and greater risk is imposed on the agent. In this paper, we test the main theoretical predictions of the EHM model in an “employer-employee” setting in the lab. We investigate the following questions: 1) Do stochastic contracts reduce gaming? 2) Is the reduction in gaming greater for more risk-averse employees? 3) Is the reduction in gaming sufficient to generate higher expected profit for employers? 4) Do employers choose to use stochastic contracts? 5) Does the adoption of stochastic contracts by employers generate potentially undesirable behavioral responses from employees?