Stages of spike time variability during neuronal responses to transient inputs

In cerebral cortex, cells tend to fire in response to strong transient fluctuations in input, produced by synchronous population activity, which reset the precision of firing and erase correlations between prior and future spike times. Here, using experiments and modeling, we study the accumulation of spike time variance in response to single decaying transient stimuli. All such responses go through distinct stages in time. When the stimulus is high, variance is held low, while at low stimulus levels near threshold, variance rises dramatically, approaching a Poisson level. This behavior was reproduced in a stochastically simulated Hodgkin-Huxley model, and in two simpler models, class 1 (Morris-Lecar) and class 2 (FitzHugh-Nagumo), incorporating Ornstein-Uhlenbeck noise. Early stage variance represents perturbation of uniform limit-cycle motion of the dynamical variables. Late stage variance reflects random motion of the dynamical variables captured within the basin of the resting fixed point. We show that the two stages have different sensitivities to the amplitude and time scale of noise, and relate this to coherence resonance. This rapid breakdown in reliability during responses to transient stimuli may restrict precise signalling by spike times to brief time windows, and limit the duration of coherent synchronous responses in the cortex.