Response of an ensemble of noisy neuron models to a single input

Spike timing precision in response to a subthreshold stimulation can be enhanced by noise in ensembles of neurons [X. Pei, L. Wilkens, and F. Moss, Phys. Rev. Lett. 77, 4679 (1996)]. We elucidate the mechanism underlying this phenomenon by computing the membrane potential distributions of ensembles of Hodgkin-Huxley neuron models. For small noise amplitudes, the membrane potential distribution takes on a Gaussian form centered on the resting potential, while for large fluctuations, there is a significant spread to lower potentials. These two regimes are separated by a relatively narrow band where the distributions transit rapidly from the Gaussian-like shapes to the spread ones. We argue that the optimal noise that maximizes the spike timing precision is situated close to this boundary.