Abstract
This article highlights several issues from simulating agent-based financial markets. These all center around the issue of learning in a multiagent setting, and specifically the question of whether the trading behavior of short-memory agents could interfere with the learning process of the market as whole. It is shown in a simple example that short-memory traders persist in generating excess volatility and other features common to actual markets. Problems related to short-memory trader behavior can be eliminated by using several different methods. These are discussed along with their relevance to agent-based models in general.