It has been known for over a century that the visuomotor transformations underlying reaching are highly adaptive. When optical prisms are placed over the eyes, for example, subjects quickly adapt to the displacement of visual feedback. This is shown by the fact that if the prisms are removed after even only a few movements, subjects exhibit reaching errors in the direction opposite the visual shift, despite the fact that their vision is now correct. Furthermore, the adaptive capabilities of the sensorimotor system are complex, e.g. displaying context dependent effects. Finally, sensorimotor learning is a continuous process: even in natural environments, sensory feedback from each behavior influences performance in subsequent movements. How sensorimotor learning takes place is a crucial question both for understanding the sensorimotor system and for developing more general theories of neural plasticity at the functional level. The mechanisms of sensorimotor adaptation are the primary focus of my lab.
Human psychophysical studies will be used to gain an understanding of the computational principles underlying sensorimotor adaptation. With our virtual reality setup, we can artificially manipulate the visual feedback available to the subject while they perform arbitrary reaching tasks. Upcoming experiments will focus on adaptation in the face of alterations of the visuomotor map. While a simple prism-like shift of the visual world is easily learned, earlier experiments of mine have shown that human subjects are unable to adapt to more complex perturbations. By probing the limits of learnable maps and observing how the sensorimotor system generalizes to movements outside the training domain, it will be possible to identify the adaptive degrees of freedom.
We are also preparing to investigate the development of visually guided behavior by studying children of various ages. One question of particular interest is whether there is a varying tradeoff during development between flexibility and robustness of behavior. This could be manifested, for example, by a progressive restriction in the degrees of freedom of the sensorimotor map with a concurrent increase in the ability of the system to generalize across the workspace.
The sensorimotor pathways responsible for visually guided reaching are an ideal system for studying the neural basis of adaptation. As discussed above, the transformation from vision to action is rapidly adapting, allowing us to monitor the fine temporal dynamics of cortical changes in relation to behavioral changes. In addition, the primary cortical pathways underlying reaching have already been identified. The areas of greatest interest here are in the Superior Parietal Cortex (Areas 5, V6a, and MIP which combine visual and somatosensory information into the precursors of movement commands) and the Frontal motor areas (PMd and M1). Furthermore, rapid progress is being made in understanding the neural representations recorded in these areas. We are in the process of building a state-of-the-art physiology laboratory which will be able to address the role of these areas in sensorimotor adaptation.
The central nervous system (CNS) has access to an overabundance of perceptual information and has control over a body with a vast number of independent degrees of freedom. For a given set of behaviors, what information does the CNS utilize from this perceptual flow, what aspects of the movement does it attempt to regulate, and what computational principles and physiological mechanisms underly the sensorimotor transformation?
Psychophysics : Starting with my thesis work at MIT, I have used human psychophysics to investigate the planning of visually guided reaching. This work is continuing, with an emphasis on the way in which information from the various sensory modalities and from internal models are combined. Extensive use is being made of perturbation techniques, in which the visual input-output mapping is artificially altered.
Physiology : My postdoctoral research focused on the the planning of eye movements in the posterior parietal cortex (PPC). In the laboratory of Richard Andersen, Boris Breznen and I investigated the behavior of PPC neurons during an object-based saccade task.
I have worked on a variety of statistical methods for the analysis of neural data. This work has been driven by my own need for such methods, as well as a belief that they can help shed new light on the manner in which information is coded in the brain. Topics include using log linear/graphical models to detect interactions between simultaneously recorded units, optimal filters for smoothing spike trains and Joint Peristimulus Time Histograms, and measures for relating neural firing to concurrent stimuli or behaviors. I am currently finishing a paper on the use of likelihood methods to get a fine temporal decoding of a population of spatially tuned cells.
