We propose to combine two methodologies: experimental economics and agent-based simulation for the study of rational behavior of individuals in a market environment. The related market represents a competitive continuous double auction (CDA). Much of what we know about CDA is based on experimental research. We still do not understand sufficiently the choice processes observed during behavior experiments. Therefore, in this paper, we will build an artificial society, which allows us to analyze how perceptions of the market price impact on individual strategies and collective behavior. More precisely, using an analytical science approach, we studied learning (for example, number of transactions memorized) and use of global vs. local information. The result of our work mainly shows that the memory of past transactions is of little importance, and that the revision of prices models the memory of agents adequately to demonstrate how rapidly price converges on an equilibrium state of the market.