Analysis and modeling of spike train correlations in the lateral geniculate nucleus

Citation

Brody, Carlos D. (1998) Analysis and modeling of spike train correlations in the lateral geniculate nucleus. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/58Z5-W488. https://resolver.caltech.edu/CaltechETD:etd-01182008-092108

Abstract

In this thesis I consider the cross-correlation analysis of spike train data, in two parts. In the first part, I consider the question of the proper interpretation of peaks in covariograms: It is known that peaks in the covariogram of the spike trains of two cells are due to covariations, not time-locked to the stimulus, in the responses of the two cells. Such peaks, even when they have widths on the order of 10s of milliseconds, are often interpreted as evidence of spike timing coordination between the two cells. However, there are other ways to covary which can generate very similar peaks. I describe two of them here: (1) covariations, over different trials, in the overall latency of the response; and (2) covariations, over different trials, in the overall excitability (i.e. average firing rate) of the response. I show how each of these leads to a peak in the covariogram, and how to distinguish such peaks from each other and from peaks due to spike timing coordination. In particular, I describe how to separate the excitability and spike timing components of a covariogram when both types of correlations are present.

The second part of the thesis studies the spike train data obtained in multiple-cell recordings in LGN of cat by Adam Sillito and colleagues (Sillito, Jones, Gerstein and West, 1994, Nature 369(6480):479-482). Analysis of this data, including the use of some of the novel insights and methods described in the first part of the thesis, shows that (1) the observed correlations between pairs of cells can be well described in terms of covariations in the latencies and excitabilities of the two cells; (2) the correlations have a time scale of lOs of seconds (20-100 s); (3) the correlations are not specific to the orientation of the drifting sine wave gratings used to drive the geniculate cells under study. A computational model, based on Huguenard and McCormick's (1992) model of thalamic cells, shows that covariations in the resting membrane potential of the two covarying cells can lead to covariogram peaks similar to those found in the experimental data. Together with Sillito et al.'s observation that correlations between pairs of cells only occurred when both members of the pair had the same receptive field types (On/Off/X/Y), and the observation that in some of the records, cells not only fail to positively covary but even anti-covary, this suggests that Silltio et al.'s data holds evidence for an intriguing and hitherto undocumented phenomenon: the possibly diffuse, but cell class-specific, control of resting membrane potential in LGN neurons.

Thesis Files