Sensitivity of biological neuron models to fluctuations in synaptic input timing

摘要

There are several reasons why it is likely that biological neurons may encode information in spike timing patterns rather than simply spike rate. These reasons are based on both information-theoretic and experimental grounds. In order for a biological neuron to function usefully as a reliable spatio-temporal pattern decoder and encoder, its input/output function must straddle two opposing properties: (1) exhibit generalization to a low level of temporal "jitter" in incoming synaptic spike timings that is below a certain threshold, and (2) have significant differences in response to incoming synaptic patterns that differ over this threshold. We show that several single-neuron models of varying levels of complexity all have this threshold property, and that its range varies between 5-30 milliseconds for each model and experiment examined in this study. This time period can be considered compatible with a clocking scheme for neural activity, and is in functional agreement with the observation that a majority of mammalian brain circuits exhibit synchronous oscillations in or near this frequency range.