Temporal correlation in cat striate-cortex neural spike trains

The authors draw on point-process theory and wavelet analysis to examine the variability and correlation properties of spike trains from single neurons in the cat striate cortex, under conditions of spontaneous and stimulated (driven) firing. It is not possible to infer the long-term correlation properties of a spike train from measures that reset at short times; thus, often-used spike-train measures, such as the interevent-interval histogram (IIH) and post-stimulus time (PST) histogram, cannot serve this purpose. Rather, the authors make use of the event-number histogram (ENH), also called the spike-number or spike-count distribution. This measure affords the experimenter the opportunity of externally controlling the counting time, T, and therefore the duration over which spike correlations can be viewed. A useful and relatively simple gauge of the correlation properties is obtained from the first two moments of the ENH. In particular, the Fano factor (FF), defined as the ratio of the spike-count variance to the spike-count mean: F(T)=var(N)|N, plotted as a function of the counting time, T, serves this purpose quite well. The FF is a special case of the wavelet Fano factor (WFF) implemented using the Haar wavelet basis