This paper presents a model of price adjustment in replica economies with two consumer types and two goods. The model provides a trading rule that allows out-of-equilibrium trading and a decentralized price-adjustment rule that features "learning through noisy imitation." It is shown that for all suffciently large economies, the process of experimentation and imitation favors adjustment of prices in the direction of excess demand. When the experimentation probability is small, the price-adjustment process mostly follows a tatonnement-like dynamics, and the limiting distribution is concentrated around the Walrasian equilibrium.