A NEURONAL LEARNING RULE FOR SUBMILLISECOND TEMPORAL CODING

A PARADOX that exists in auditory and electrosensory neural systems(1, 2) is that they encode behaviourally relevant signals in the range of a few microseconds with neurons that are at least one order of magnitu de slower, The importance of temporal coding in neural information pro cessing is not clear yet(3-8), A central question is whether neuronal firing can be more precise than the time constants of the neuronal pro cesses involved(9), Here we address this problem using the auditory sy stem of the barn owl as an example, We present a modelling study based on computer simulations of a neuron in the laminar nucleus. Three obs ervations explain the paradox. First, spiking of an 'integrate-and-fir e' neuron driven by excitatory postsynaptic potentials with a width at half-maximum height of 250 mu s, has an accuracy of 25 mu s if the pr esynaptic signals arrive coherently. Second, the necessary degree of c oherence in the signal arrival times can be attained during ontogeneti c development by virtue of an unsupervised hebbian learning rule, Lear ning selects connections with matching delays from a broad distributio n of axons with random delays, Third, the learning rule also selects t he correct delays from two independent groups of inputs, for example, from the left and right ear.