Where before we had a product of two scalars on each trial, now we have the cross-correlation of two spike trains. Operationally, this is defined as
follows: take each spike of cell 1 (shown in orange on the left); find the relative times from this spike to all the spikes of cell 2 (green). At each of those
relative times, put a spike down in the cross-correlation graph (shown in red at right). Do this for all the spikes of cell 1. The result is the cross correlation of
the two spike trains. Notice that if cell 1 and cell 2 had fired almost synchronously, this operation would have resulted in a high density of spikes near t=0 in
the cross-correlation graph on the right.
.
.
An important variant of the experiment that Sillito et al. carried out was this: When the LGN cells were stimulated with flashing squares instead of drifting
gratings, the peak in the covariogram disappeared (compare the two covariograms in the slide). The same two cells were used for both covariograms, but
the stimulation conditions differed.