The proposed research will examine the precision and robustness of the neural code of the retina. Accurately localizing a moving target is one of the most basic tasks of vision. To explore the neural basis of this phenomenon, simulated trajectories of a moving object will be presented to the retina while the extracellular activity of approximately 50 ganglion cells is simultaneously recorded. Three different types of neural codes will be used to reconstruct the target trajectories based on combining information contained in the ganglion cell firing patterns: linear summation of the individual signals, nonlinear summation of the individual signals, and the optimal Bayesian estimate based on the firing patterns. The reconstruction precision for each code will be calculated as a function of insect velocity and background noise level. The control of these parameters will allow the performance of each neural code to be pushed to its operational limits, the point at which successful computation is no longer possible. The suitability of each code as a carrier of spatial information will then be determined by comparing the precision and robustness of each code to the behavioral performance of the salamander. One of the most promising applications of systems neuroscience is neural prosthetics. In order to construct a prosthetic retina that is capable of restoring sight, it is necessary to have a deep understanding of the neural code used in the intact retina. The research described in this proposal will bring us one step closer to that goal.
