Computers now allow the construction of detailed and quasi-natural environments built around simulations of complex processes. These environments can provide interfaces that let users control the simulated processes and observe the consequences in real time. When the interface includes virtual reality technology, the user is, in addition, “immersed” in the environment as if it were the natural world. This presentation is concerned with how such artificial environments can be configured and used in order to help people learn about the simulated phenomena. It begins by suggesting a theoretical framework that explains how learning might occur in artificial environments, and from which hypotheses can be constructed about what features of artificial environments can be expected to help or hinder learning. It then describes the “state of the art” of the implementation and study of artificial environments in education. Examples of what works and what doesn’t, of interesting and less interesting research questions, are provided from our ongoing work with “Virtual Puget Sound”. The presentation concludes with a request for advice on what to do next as the research agenda continues.