Stochastic resonance in neuron models

Abstract

Periodically stimulated sensory neurons typically exhibit a kind of “statistical phase locking” to the stimulus: they tend to fire at a preferred phase of the stimulus cycle, but not at every cycle. Hence, the histogram of interspike intervals (ISIH), i.e., of times between successive firings, is multimodal for these neurons, with peaks centered at integer multiples of the driving period. A particular kind of residence time histogram for a large class of noisy bistable systems has recently been shown to exhibit the major features of the neural data. In the present paper, we show that an excitable cell model, the Fitzhugh-Nagumo equations, also reproduces these features when driven by additive periodic and stochastic forces. This model exhibits its own brand of stochastic resonance as the peaks of the ISIH successively go through a maximum when the noise intensity is increased. Further, the presence of a noise-induced limit cycle introduces a third time scale in the problem. This limit cycle is found to modify qualitatively the phase-locking picture, e.g., by suppressing certain peaks in the ISIH. Finally, the role of noise and possibly of stochastic resonance (SR) in the neural encoding of sensory information is discussed.