"All models are wrong; but some are useful." This well-known adage by the statistician, George Box, contains several important insights (Box 1979) -first, that a model is an abstraction of reality, and second, that the parts of a model which are important for its intended purpose should be emphasized, while those which are nonessential can be deemphasized or even left out. Consider, for example, "model" airplanes. Display models, such as those designed to be hung from ceilings, are concerned with proportions vis a vis size, material composition, or flying capability. If they look right, they make for satisfied users. Models intended for flying, on the other hand, sacrifice proportions and composition in order to produce a device that "performs." Since neither model reproduces reality, each is inherently "wrong"; however, each may meet its intended purpose quite well. The situation is similar with software models. A skilled modeler should focus on those aspects of reality that are germane to the issue of interest. If he or she succeeds, and the abstraction effectively addresses the factors relevant to the model's purpose, then it is possible that the model will be "useful."The questions of adequate abstraction and of the model's usefulness are particularly challenging -and poorly explored at this time -for PMESII models (societal models). Like the challenge of visualization (Chap. 10), the challenge of validation handicaps the current generation of societal models. This is the topic of our chapter: having constructed a societal model, how do we know that the model is useful?Verification, validation, and accreditation (VV&A) are processes that address the correctness of models. Just as models can have many forms, they can refer to different parts of the real world (e.g., hurricanes, flight characteristics of an airliner, price sensitivity of a commodity to scarcity, or the impact of international interventions). Further, models can have different purposes (e.g