The evolution of mammalian cortex, from lamination to arealization

Abstract We analyse some of the most important anatomical and functional features emerging at different stages of mammalian brain evolution in terms of a possible computational advantage. At the transition from reptiles to mammals, a major anatomical change occurs in the originally sensory dorsal cortex. The principal layer of pyramidal cells is split by the insertion of a new layer of granule cells, giving rise to the laminated isocortex. It has been hypothesized that this qualitative change in the evolution of mammalian brains is necessary to support fine topography in their sensory maps. The simulation of neural network models demonstrates that a nonlaminated patch of cortex must compromise between transmitting "where" information, explicitly mapped, topographically, on the cortical sheet, and retrieving "what" information, represented by the distributed firing pattern across neurons. The differentiation of a granular layer is shown in the model to yield a small quantitative advantage, allowing to transmit a slightly better combination of both information types. Along the same theoretical lines, we are investigating the multiplication of successive sensory areas coding for ever more composite stimuli, such as those in the visual and auditory temporal cortices in primates. In particular we analyse the possible computational advantage for a specific neural population devoted to encode the complex structure of whole stimuli, rather than relying on the coactivation of separate populations encoding their basic elements