Computational neurobiology is an approach to studying the nervous system that attempts to describe complex behavior, often observed in systems of neurons, in terms of information, and algorithms for encoding such information. My research has focused on a combination of these approaches to advance our understanding of brain/behavior relationships. Although each model system has its unique attributes and technical strengths and weaknesses, the core evolutionary principles that guide neuroethological research guarantee that principles learned in one model system apply generally. A wealth of research in ethology and neurobiology substantiates and elaborates this claim. In my own research, the major focus has been the study of birdsong learning and the attendant avian song system. Song learning is an example of a very general class of sensorimotor learning phenomena that are modeled as reinforcement learning. Thus, the lessons we have learned from this emerging premier model system have broad implications for explaining the rules whereby systems of neurons organize in response to environmental cues. Because song learning is a form of vocal learning, our studies also impact upon research in speech and language processing.

Two principal computational problems need to be solved in neural implementations of vocal learning. The first is to form the high dimensional mapping between sensory (auditory) input and vocal (motor) output. The second is to establish a mechanism that can learn to adaptively modify motor output in the presence of feedback delay. Recent observations in zebra finches suggest that "offline" mechanisms during sleep as well as "online" mechanisms during singing are important for birdsong learning. We are currently testing the central prediction of this hypothesis, that sleep acts to reinforce or suppress various motor patterns practiced during the day. We have identified the cholinergic system as a central player in behavioral state modulation in this system. This motivates studies at the systems and cellular level (the latter in a brain slice preparation) to understand how sensory feedback is differentially regulated in different neural pathways by neuromodulators. These studies are pursued in the context of observations regarding temporal coding and hierarchical organization of the vocal motor system. Our recent neurophysiological studies of song development suggest that the traditional instructional model of birdsong learning is incomplete. Rather, selectional processes may act to establish motor programs during the subsong and early plastic song phases of song development. We also study the neuroethology of auditory perceptions, currently by pursuing an analysis of individual vocal recognition in starlings by combining behavioral and neurophysiological approaches. Finally, we are investigating certain predictions of human speech perception and production that have resulted from the birdsong work.

To address these and related questions, we use a variety of approaches, including extracellular single-cell and multielectrode array recordings in the context of auditory neurophysiology, intracellular brain "slice" recordings, chronic motor recordings from singing birds, neuroanatomical techniques, computer based behavioral observation and shaping, and (connectionist) neural nets modeling and other quantitative approaches. I will sponsor students interested in these or other aspects of neurobiology and behavior.

I teach a course in Neuroethology. The design of this course considers the needs of advanced students who plan to pursue graduate work, particularly in neurobiology or experimental psychology. It covers topics in systems, computational, and behavioral neuroscience. There is a heavy emphasis on original literature, and oral and written scientific presentations. Labs include exposure to instrumentation and electronics, and involve work with live animals. The labs are taught in one of our state-of-the-art laboratory rooms in the Biological Sciences Learning Center building.