Shawn R. Lockery
Electrophysiological and Genetic Studies of Behavior in the Nematode C. elegans
Information in the vertebrate brain is distributed widely across many brain regions that operate in parallel to control behavior. Parallel distributed processing seems to be essential for the versatility and robustness of human and animal behavior, but the relationship between cellular properties of single neurons and information processing in large neural networks is poorly understood. We study this problem by analyzing simple, well-defined parallel distributed processing networks controlling bending in the leech, Hirudo medicinalis, and chemotaxis in the nematode, Caenorhabditis elegans.
The function of these networks is studied using a combination of physiological and theoretical techniques. Intracellular and patch clamp recordings are made to determine the biophysical properties of identifiable neurons and their responses during behavior. These data are combined with neural network training algorithms such as dynamic backpropagation to produce realistic computer simulations of the biological networks controlling bending and chemotaxis. This approach has enabled us to construct large-scale working models of both behaviors. Theoretical predictions from the simulations are then tested physiologically to confirm and improve the models. These results provide new insights into parallel distributed processing in biological networks and new modeling strategies for studying larger networks in the vertebrate brain.
Current and future research will seek to determine (1) how single networks control qualitatively different behaviors, (2) how distributed representations of information are changed by learning, and (3) how distributed processing networks are constructed in development and maintained in adulthood.