Natural time scales for neural encoding

We propose natural time scales for characterization of two important parameters of neural encoding: spike timing precision dT and average word length T-w. These scales emerge through asymptotic limits of the mutual information between stimuli and corresponding spike trains. We used crickets (acheta domestica) as animal models to illustrate our definitions and test their validity. Intracellular multielectrode extracellular recordings were performed to assess neuronal activity in the crickets' terminal ganglion. The stimuli for these experiments were random waveform air currents with various statistical characteristics. Using an information-theoretic analysis based on the abovementioned time scales, we characterize the maximal information content of spike sequences in this sensory system, independent of their temporal discretization. (C) 2000 Elsevier Science B.V. All rights reserved.