A single timescale is insufficient to describe both intermittency and entrainment. A, Model schema. Input elicits a spike with a probability determined by the excitability x, causing depletion by a factor U, followed by recovery with a timescale τ. White noise of amplitude σ is added to x. B, Fano factors of response probability as a function of time window. Data (black) are compared with model (σ = 0, blue; σ = 18U, red). C, Autocorrelation of response probability to constant stimulus (calculated in 1 s resolution) as a function of time. Data (black) are compared with models chosen to fit either the autocorrelation (red) or the input–output covariance under white noise stimuli (green). D, Input–output covariance for white noise stimulus for the same three neurons. Colors are the same as in C.