The author describes a protocol for classroom experiments for courses that introduce undergraduates to signaling games. Signaling games are conceptually difficult because, when analyzing the game, students are not naturally inclined to think in probabilistic, Bayesian terms. The experimental design explicitly presents the posterior frequencies of the unobserved events. The protocol's emphasis on the posterior enhances convergence to the equilibrium prediction, relative to a treatment in which posterior frequencies are not explicitly computed. This convergence reinforces the development of the theory in subsequent lecture periods. (Contains 6 figures, 1 table, and 5 notes.)