Models for the cross-correlation between retinal ganglion cells

Neighboring ganglion cells in the vertebrate retina not only respond to the same stimuli but also display cross-correlated activity on a millisecond time scale. Recent studies of this cross-correlation have indicated that simple linear addition of common variability to each ganglion cell signal does not account for the observations (Levine 1997). In this report, Monte Carlo simulations of various linear and nonlinear models are presented that confirm the earlier speculations. Models in which common variability alters the leakages of a pair of leaky integrate-and-fire neurons account for the data and predict the cross-correlogram lag without invoking temporal delay lines.