A software behavior implements an action of a simulated entity. For example, a behavior can change the stance of an individual combatant (IC) or move an IC to a new position. Currently, behaviors are used to control ICs in many Computer Generated Forces (CGF) simulations such as Closed Combat Tactical Trainer (CCTT) and Modular Semi-Automated Forces (ModSAF). A behavior approach, on the other hand, is a software technique used to implement a behavior. For example, in CCTT and ModSAF the software technique utilizes finite state machines (FSMs). In the past few years, computer hardware technology has provided massive improvements. These improvements combined with the need for more realistic and autonomous behaviors as well as decision-making that handles multitudes of different inputs resulted in the Non-Traditional Human Behavior Models project. The goal was to research several behavior approaches and implement these approaches within a CGF simulation. The project examined in detail traditional FSM, Q-Learning reinforcement, evolutionary, and fuzzy rule-based approaches as each of these approaches provided different mechanisms with different strengths and weaknesses to control ICs in specific use cases. A previous paper was published describing the overall design of the behavior approaches and their relationship to the CGF Test-bed (Gugel, Pratt, & Smith, 2001). This paper, the second in a series, details the scenario (and its variants) selected to evaluate the four behavior approaches. The paper describes the specific scenario design for each approach. The next section describes the results of the experimentation for each approach and scenario variant combination. The final section outlines the overall results across all of the experimentation. This section also outlines overall benefits and weakness of these approaches with respect to implementation. We believe that understanding different behavior approaches and allowing different approaches to exist within the same CGF simulation will allow a diversity of new behaviors to be developed that provide more realism as well as more automation. We believe that these approaches can provide an accurate portrayal of CGFs in training simulations and provide a more versatile simulation for the analysis of new doctrine and tactics.