This study focuses on the analysis of temporal patterns in the spike train of cells in the lateral geniculate nucleus (LGN) of cat. Two-hundred eighty-three units have been recorded extracellularly in anesthetized animals during visual stimulation with flashing spot stimuli of different size. We used a novel method of temporally local computed interval distributions (intervalogram; Funke & Wörgötter, 1995) to visualize the statistical distribution of interspike intervals during different phases of the visual response. Multimodal interval distributions were observed mainly in X- and Y-ON cells, reflecting the tendency of these cells to fire with preferred intervals during the sustained light response. The shortest preferred interval is called the fundamental interval and the longer ones (higher-order intervals) are, in general, multiples thereof. During increasing surround inhibition a redistribution of the intervals towards the higher orders was observed. We regarded the different peaks in the interval distributions as different components of possible temporal spike sequences and performed a pattern search up to the level of five subsequent intervals. While it is obvious, that the dominant peak is most strongly represented in any interval sequence, we also show that a significant overrepresentation of short sequences of similar intervals exists. The repetition rate is rather small (4–5 intervals) and, therefore, no long-lasting oscillatory pattern was observed in the autocorrelograms. Power spectral analysis of the peristimulus-time histograms, however, revealed that the sequential firing pattern is strongly stimulus locked at least for the majority of sweeps in the records.
The mean firing rate of an LGN cell decreases with increasing stimulus size as well as with decreasing contrast. Therefore, the mean rate cannot be used to distinguish between these situations. While in the whole network this tradeoff can be resolved by the combined activity of multiple cells, our findings additionally suggest that contrast and size can be distinguished already at the single-cell level using different temporal patterns.