[Epistemological remarks on the question of cerebral organization]

Two complementary models of brain organization are nowadays offered: one anatomical and functional, the other computational. The first model implies a subdivision of brain functions in order to find the corresponding brain areas. It is easy to show that this cutting up can only be artificial and is influenced by some kind of "folk psychology", thereby raising suspicion on the "top down" method. On the other hand, it seems impossible to do without any psychological and even any subjective reference and work only according to the bottom up way. It must therefore be concluded that our subdivisions of brain functions must always be regarded as provisional as long as the two methods do not exactly overlap. The computational model uses either the conventional method or the parallel distributed processing. It can be shown that the conventional language utilized impoverishes the normal language and even more thought and is therefore unable to yield more than restricted results. It is suggested that the main advantage of parallel distributed processing is precisely to erase the formalization of language and it is then possible to expect from it much more interesting models. Although it seems possible to improve continuously our understanding of the functional conditions required for consciousness to appear, there seems to be theoretical limits to a complete scientific explanation of self awareness.