Bull. Amer. Phys. Soc. 45:943, 2000.
Richard Kempter
Computational maps are of central importance to the brain's representation of the outside world. The question of how maps are formed during ontogenetic development is a subject of intense research (Hubel & Wiesel, Proc R Soc B 198:1, 1977; Buonomano & Merzenich, Annu Rev Neurosci 21:149, 1998). The development in the primary visual cortex is in principle well explained compared to that in the auditory system, partly because the mechanisms underlying the formation of temporal-feature maps are hardly understood (Carr, Annu Rev Neurosci 16:223, 1993). Through a modelling study based on computer simulations in a system of spiking neurons a solution is offered to the problem of how a map of interaural time differences is set up in the nucleus laminaris of the barn owl, as a typical example. An array of neurons is able to represent interaural time differences in an orderly manner, viz., a map, if homosynaptic spike-based Hebbian learning (Gerstner et al, Nature 383:76, 1996; Kempter et al, Phys Rev E 59:4498, 1999) is combined with a presynaptic propagation of synaptic modifications (Fitzsimonds & Poo, Physiol Rev 78:143, 1998). The latter may be orders of magnitude weaker than the former. The algorithm is a key mechanism to the formation of temporal-feature maps on a submillisecond time scale.