In this paper, we present a new framework for analyzing and designing virtual reality (VR) techniques. This framework is based on two concepts -system fidelity (i.e., the degree with which real-world experiences are reproduced by a system) and memory (i.e., the formation and activation of perceptual, cognitive, and motor networks of neurons). The premise of the framework is to manipulate an aspect of system fidelity in order to assist a stage of memory. We call it the Altered-Fidelity Framework for Enhancing Cognition and Training (AFFECT). AFFECT provides nine categories of approaches to altering system fidelity to positively affect learning or training. These categories are based on the intersections of three aspects of system fidelity (interaction fidelity, scenario fidelity, and display fidelity) and three stages of memory (encoding, implicit retrieval, and explicit retrieval). In addition to discussing the details of our new framework, we show how AFFECT can be used as a tool for analyzing and categorizing VR techniques designed to facilitate learning or training. We also demonstrate how AFFECT can be used as a design space for creating new VR techniques intended for educational and training systems.