In humans, neural signals are most easily recorded using EEG recordings. Graduate student Paul Hammon and I have recently begun a project to improve analysis of EEG recordings with the goal of improving non-invasive brain-computer interfaces (BCI). The overall goal of our proposed research is to design and test a BCI system with vastly improved mind-reading abilities. These improvements will include: increasing the types and dimensionality of interpretable desires that can be read, increasing classification rates, and significantly decreasing training time for BCI users. In order to achieve this goal we intend to develop a qualitatively different motor-imagery based BCI system by using natural human movements with rich and intuitive visual feedback to make use of the user's innate motor learning skills. Our hypothesis is that the mapping from natural (imagined) movements to realistic feedback of the intended movement will be easier and more intuitive to learn and use than the somewhat arbitrary pairing of mental task and feedback currently employed in most BCI systems. So far we have created an automated method for feature selection that earned a second place finish in an international BCI competition and demonstrated that viewing realistic animated movements during motor imagery improves EEG classification rates relative to motor imagery alone[#!hammon06a!#]. We have also shown that EEG recorded during reaching movements can accurately predict the reach target and trajectory.