When evaluating new vehicle designs, modeling and simulation offer techniques to predict parameters such as maximum speed, fuel efficiency, turning radius, and the like. However, the measure of greatest interest is the likelihood of mission success. One approach to assessing the likelihood of mission success in simulation is to build behavior models, operating at the human decision-making level, that can execute realistic missions in simulation. This approach makes it possible to not only measure changes in mission success rates, but also to analyze the causes of mission failures. Layering behavior modeling and simulation on underlying models of equipment and components enables measurement of more conventional parameters such as time, fuel efficiency under realistic conditions, distance traveled, equipment used, and survivability. To maximize these benefits, the behavior model must be designed in a way that is able to take advantage of a variety of vehicle capabilities, rather than being tailored to a particular vehicle design. The model needs to be driven more by “first principles,” such that its behavior in any particular situation is dynamically determined by an assessment of its own vehicle capabilities compared to the needs of the situation. For example, the range of some sensor may determine whether or not the appropriate behavior when detecting a superior force is to advance, withdraw, or “circle the wagons” by assuming a defensive posture. Most ground vehicles serve a variety of missions, so to generate a projection of mission outcome requires generating a sufficiently broad range of mission conditions such that there can be confidence in the vehicle evaluation. Hand-coding specific mission scenarios in simulation is too labor intensive, while generating every aspect from scratch may result in scenarios that are unrealistic. A middle ground is hand-coding the scenario basics, but then varying specific aspects. For example, the particular kinds, locations and routes of hostile forces, the time constraints, and the weather could be varied while holding the terrain, starting location, and goal location constant. We discuss past and current examples of behavior models and scenario variation in simulation in support of model-based system engineering, training, and vehicle autonomy.