Stochastic dependence of the consecutive interspike intervals of the spontaneous discharges of the central single neurons has been studied in cats. Out of 35 neurons examined with the stationarity test, 16 passed it: 3 from the ventrobasal complex where the modality was hair (VB), 6 from the lateral geniculate nucleus (LGN), 4 from the mesencephalic reticular formation (MRF), and 3 from the red nucleus (RN).
The interspike interval histogram in the VB and the LGN nuerons resembled the exponential and the bimodal distribution, while that in the MRF and the RN neurons did the gamma distribution.
The quantitative measures of correlation between interspike intervals were furnished by the serial correlation coefficient, the variance function and the joint interval histogram. They showed that there were no correlation in most of the VB and the LGN neurons, while there existed correlation in most of the MRF and the RN nuerons. Approximately similar results were obtained by the X
2 test for Markov process.
A new quantitative measure which reflects Markov properties of the consecutive impulse sequences has been developed. It estimates the statistical dependency,
d(τ), on the basis of Shannon's entropy. It has been confirmed theoretically that this measure is applicable in practice.
The value of
d(τ) of adjacent interspike intervals obtained from the VB and the LGN neurons was approximately equal to that of the shuffled ones, while that from the MRF and the RN neurons was larger than that of the shuffled ones. This is to be regarded as the evidence of the fact that the spontaneous impulse activity of the VB and the LGN neurons has no Markov property, while that of the MRF and the RN neurons has at least first order Markov property.
The differentiation in two groups of the neurons is suggested to correlate with other functional characteristics of these neuron populations. However, the number of neurons sampled would have to be increased several-fold before the statistical differences in patterns of discharge could be made the basis for widely applicable generalizations about function.
View full abstract