The brain has evolved a division of labour amongst component systems which
link different sorts of processing with precise actions. Debate is over cen
tralized versus decentralized control at different processing levels, from
cognitive systems to motor-control systems. With simultaneous activation of
alternative expert systems which link (a) picture-processing with drawing
and (b) reading with writing, decentralized modelling predicted both the av
eraging of action-production times and additive effects of neural noise. Su
ch modelling has the advantage of being able to measure the cost of regulat
ion both within and between systems, in a common metric of performance vari
ability. That commonality strengthens the trend to regard the brain as a di
stributed super-system with a great deal of regulation done towards the out
put end. (C) 2000 Elsevier Science B.V. All rights reserved.