17/2 DESIGN FOR A BRAIN 



to stall it ; and he has to learn to change gear, neither burning 

 the clutch nor stripping the cogs. On an open, level, empty road 

 he can ignore accelerator and gear and can study steering as if 

 the other two systems did not exist ; and at the bench he can learn 

 to change gear as if steering did not exist. But on an ordinary 

 journey the relations vary. For much of the time the three 

 systems 



driver + steering wheel -f . . . 

 driver -f- accelerator 4- . . . 

 driver -f gear lever -f . . . 



could be regarded as independent, each complete in itself. But 

 from time to time they interact. Not only may any two use 

 common variables in the driver (in arms, legs, brain) but some 

 linkage is provided by the machine and the world around. Thus, 

 any attempt to change gear must involve the position of the 

 accelerator and the speed of the engine ; and turning sharply 

 round a corner should be preceded by a slowing down and by a 

 change of gear. The whole system thus shows that temporary and 

 conditional division into subsystems that is typical of the whole 

 that is composed largely of part-functions. 



17/2. Before supposing that the nervous system, in its con- 

 struction and function, resembles the multistable, we may ask 

 to what extent the supposition is necessary. S. 9/4 showed the 

 necessity for ultrastability ; is the hypothesis of multistability 

 equally necessary ? 



Our basic facts and assumptions are now as follows : 



(1) the nervous system adapts by the process of ultrastability 



(S. 9/4), 



(2) it can retain one adaptation during the acquisition of 



another (S. 11/3), 



(3) this independence is not achieved by a division of the 



nervous system into permanently separate parts (S. 11/8), 



(4) no special mechanism is to be postulated for special en- 



vironmental conditions (S. 1/9) : if possible, the variables 

 are to be statistically homogeneous. 

 Given these, what can be deduced ? 



In the system, label the main variables M and the step-functions 

 S. Call those variables immediately affected by the first environ- 



182 



