Dynamical Networks in Physics and Biology, pp. 285-296. D. Beysens and G. Forgacs, eds., Berlin, Springer, 1998.

Hebbian learning of temporal correlations: Sound localization in the barn owl auditory system

J. L. van Hemmen and R. Kempter

There exists an unresolved paradox in auditory and electrosensory neural systems in that they encode behaviorally relevant signals in the range of a few microseconds with neurons whose time constants are at least one order of magnitude bigger. Here we focus on the barn owl's auditory system, analyze its azimuthal sound localization, and present results of a modeling study based on computer simulations of a neuron in the laminar nucleus. We argue that three observations resolve the paradox. First, the neuron is driven by stochastic presynaptic signals that arrive more or less coherently. Second, the necessary degree of coherence in the signal arrival times can be attained during ontogenetic development by virtue of an unsupervised Hebbian learning rule. It singles out the `right' connections with matching delays from a broad distribution of axons with random delays. Third, the learning rule also selects the correct delays from two independent groups of inputs, for example, from the left and right ear.