The goal of this project is to construct a theoretical and empirical framework that can account for and make accurate predictions about the effectiveness of different training methods over the full range of militarily relevant tasks. The ability to predict the outcomes of different training methods on particular tasks will, as a natural by-product, point to ways to optimize training. Many of the basic mechanisms of knowledge and skill acquisition are similar across a variety of perceptual, cognitive, and motor tasks. However, some specific skills have unique features that might demand special training techniques. To date, only a few studies have compared learning in different types of tasks. We will focus on an analysis of which findings, mechanisms, and principles broadly generalize across learning types and task requirements. This evaluation will allow us to make specific predictions about the effectiveness of training and general recommendations to improve training that would apply to virtually any DoD training program. We will also identify the unique features of specific knowledge and skills, where they exist, and how best to train them. We will develop taxonomies for both types of training and types of tasks that will span the range of training types, from classroom to simulator, and task types, from simple individual tasks (e.g., data entry, target detection) to complex tasks involving team cognition. We will extend our taxonomic analysis to include two new dimensions: training principles and performance measures. Two types of working predictive models of training effects will be developed and contrasted for their ability to account for and predict training outcomes: One type of model will be based on IMPRINT, and the other type will be based on the ACT-R approach. We will ensure that these models are mathematically sound, computationally feasible, and DoD applicable.