Experimentation, imitation, and stochastic stability

Do boundedly rational agents repeatedly playing a symmetric game with a unique symmetric equilibrium learn over time to play it? In this paper we model the dynamic interaction of two types of such agents, experimenters and imitators, whose behavior is characterized by simple rules of thumb. We iind that the stochastic process describing their play is stable in the large: it converges globally and with probability one to st compact neighborhood of the equilibrium. However, its local behavior near the equilibrium depends in interesting ways on the details of the model. Journal of Economic Literature Classification Number: D83.(C) 1999 Academic Press.