The US Armed forces, and increasingly its coalition allies, continue to embrace simulation technologies as a solution to many training challenges. As these applications broaden in scope, it will no longer prove effective to use high-end systems to train all tasks. Rather, it will become critical to match training requirements to specific training technologies. Simulation tools such as Tactical Decision-Making Simulations (TDSs) seem to offer substantial benefits as modalities for training complex cognitive skills including planning and decision making, while keeping development and procurement costs significantly lower than those associated with high-end systems. However, the training community lacks a means of ensuring that utilization of these technologies for cognitive training will enhance battlefield performance. To aid TDS developers in designing tools that support the acquisition of expertise, a framework was generated to describe the evolution from beginner to expert in ill-structured environments. The framework reflects the notion that effective training enables learners to move toward a state of expertise, where good decisions are made quickly and automatically despite environmental ambiguity and chaos. Thus, the framework describes 1) five levels of proficiency and characteristics of learners at each stage; 2) the process by which individuals transition from one stage to the next; and 3) indicators of proficiency that enable diagnosis of a learner's current stage. Two main assertions anchor the framework. First, in cognitively complex domains such as tactical thinking, learning consists of developing and refining mental models which enable individuals to accurately size up situations and apply action scripts to accomplish the objectives. Second, mental model development relies differentially on a range of training environments and techniques depending on the learner's current proficiency level. Since the framework has implications for training in ill-structured environments regardless of specific domain content, it should prove adaptable across services and even across alliances.