Although behavioral/psychological factors are believed to contribute to today's health problems, such as physical inactivity and obesity, it is generally agreed that rapid change in the environment is the major contributor (Jeffery & Utter, 2003). The complex relationship between health behaviors/outcomes and the environment, however, remains unclear. Lack of appropriate analytical methods is one of the reasons. Most conventional, group-centered linear statistical methods, including some sophisticated statistical measures (e.g., hierarchical linear model), cannot handle the complex relationships among the variables examined. This paper introduces agent-based modeling (ABM), a newly developed simulation method for analyzing and modeling clustered, multi-level complex data. The heart of ABM is the “agent,” the subject/individual with a set of characteristics and behaviors. An agent's behavior is determined by a set of rules governing its decision-making and communication protocols. The purpose of ABM is to understand how sufficient numbers of agents interact with each other and their environment and how changes in the environment will affect their measurable behavior. An agent-based simulation consists of a set of agents, a set of agent relationships, and a framework for simulating decisions and interaction. In contrast to traditional equation-based, top-down modeling, ABM grows a simulated complex adaptive system from the bottom up; individual agents make up the system, and they interact among themselves and with the environment according to rules governing behavior and environmental interactions. As a simulation technique, ABM differs from earlier approaches in the number of objects simulated. Simulating 100,000 individuals in a small city for “24 hours” in one hour means that each agent gets 400 nanoseconds for each minute of simulated time. Clearly, most decision making must be simple and fast. For example, ABM applied to urban transportation activities would begin by defining a street grid (the environment), and driver agents with scenario guiding roles for these agents – commuters, students with flexible schedules, deliverymen, etc. Each agent would seek to accomplish its goal, such as arriving at work on time, by adjusting its driving patterns to accommodate the environment and other agents. A given agent will learn over time which route is the fastest and alter its behavior accordingly. After an overview of ABM, its application to modeling physical activity behaviors will be illustrated using data/examples from a walking space project. Finally, commonly used ABM software will be briefly introduced.Keyword(s): exercise/fitness/physical activity, measurement/evaluation, research