Implementation of Pigovian taxation relies on the presumption that individuals follow self-interested Nash equilibrium predictions of behavior when making decisions. Experimental evidence indicates that, while Nash predictions perform quite well in impersonal exchange, in other environments, subjects behave in ways inconsistent with these equilibria. The predictive power of game-theoretic results with respect to an optimal subsidy in a common-pool resource game (CPR) remains an open question. This paper presents an experiment with training and a simplified decision task, allowing more tractable computerized CPR experiments. In this experiment, subject behavior converges to the Nash prediction, but it takes a number of periods to reach convergence. In addition, this paper provides the first experimental test of theoretical predictions of behavior under an optimal subsidy in CPR games. I find that a Pigovian subsidy effectively moves subject behavior to the pre-subsidy social optimum. Finally, this paper provides evidence of a small and non-persistent effect of information provision in moving subjects toward the social optimum.