Subjects played strategically similar 4x4 and 6x6 constant sum games under varying payoff scales. Substantial divergences from equilibrium predictions were exhibited. The dynamic pattern of play is best expalined by a stimulus learning model whereby players allocate weight to different actions...