One of the most basic problems we face in systems neuroscience is understanding how information from the outside world is represented in the activity of populations of neurons. This proposal is directed toward this problem, and uses the retina - specifically, the output cells of the retina - as the model system. We focus on two key questions: 1) What code do the cells use to carry the information, and 2) what roles do the different classes of cells in the population play? To address these questions, we will use a combined electrophysiological and behavioral approach, with the rodent as the model species.
Our specific aims are as follows:
In Aim 1, we will test hypotheses about the codes used by the output cells. Specifically, we will decode their spike trains assuming different codes, measure the performance of the codes on visual discrimination tasks, and compare the performance to that of the animal. We will focus on a series of widely proposed codes, starting with a simple spike count code, and then advancing systematically through a set of more complex codes (a spike timing, a temporal correlation, and a cross-correlation code). We have set up an experimental protocol that allows us to obtain an upper bound on the performance of each code. With an upper bound, we can rigorously determine which codes are viable for the animal and which are not - that is, which codes the animal can be using and which codes must be ruled out.
In Aim 2, we will test hypotheses about the contributions of the different classes of retinal output cells to the representation of visual scenes. We will do this by decoding the output cell spike trains with and without specific cell classes included. These studies will provide fundamental information about the strategies the nervous system uses to represent and process information. Since a substantial portion of the work involves mapping the input/output relationship of the retina, this research will also contribute to the development of algorithms for retinal prosthetics.
Showing the most recent 10 out of 17 publications
