A variety of interfaces have been used to enable students to interact with learning technologies. In addition to point and click, text and various media are in common use. However, these technologies seldom attempt to provide an interface that is similar to a conversation between two people. Building on work at the Carnegie Mellon Human-Computer Interaction Institute, we have applied such a conversational interface to the enhancement of physics teaching and incorporated it into the Physics Teaching Web Advisory (Pathway), an environment which has proven effective in physics teacher education. This project is extending this unique learning environment; formally studying student learning within the environment; and establishing, through this research, design principles which are specifically focused on this new mode of computer-student interaction. CMU's Synthetic Interview (SI) provides the user with the feeling of a conversational interaction through a combination of a video of a person and a highly reliable natural language interface. During this project the SI system will be tightly integrated with other media such as video, text, animations, illustrations, and simulations. The result will give the students the impression of a person who is talking directly to them and explaining complex concepts with highly interactive illustrative media. Our hypothesis is that the combination of the SI with other interactive multimedia will create a system that will improve student understanding of topics in physics. This hypothesis will be rigorously tested in the research component of this project. One can imagine a variety of ways in which learning can proceed when the SI software is coupled to other presentations on the screen. The on-screen person can refer to other objects, direct a student to interact with these objects before continuing with any other effort, and help the student directly with his/her learning. Thus, the learning environment is tightly controlled by the system. The control could also be rather loose. The on-screen person can recommend but not require actions with other learning presentations or simply point out that they are available. At present, software for this type of interaction, particularly including the conversational mode, is not available. Thus, no research is available to help developers of learning technologies know what type of coupling is best for improvement of learning. This research will investigate what types of integrations are most effective in providing scaffolding for student learning of physics concepts, for encouraging students to reflect on their learning, and for enhancing transfer of knowledge to the learning of physics.