17/3 DESIGN FOR A BRAIN 



subdivision into parts, where there are many subsystems, some 

 simple, some complex, acting sometimes independently and some- 

 times in conjunction, where an adaptation produced for one part 

 of the environment may conflict with an adaptation produced for 

 another part, and where the adaptations themselves have to be 

 woven into more complex patterns if they are to match the com- 

 plex demands of the environment. It will be considered in 

 Chapter 18. Beyond this, for completeness, are the environments 

 of extreme complexity ; but they hardly need discussion, for at 

 the limit they go beyond any possibility of being adapted to — 

 at least, in the present state of our knowledge. 



The environment of the first type, that composed of indepen- 

 dent parts, would, if joined to a multistable system, form an 

 ultrastable whole (S. 16/2). Adaptation will, therefore, tend to 

 occur. But as the whole is also multistable the process will show 

 modifications. Dispersion will occur, so that at each moment 

 only some of the whole system's variables will be active. This 

 allows the possibility that though the whole may contain a great 

 number of variables yet little subsystems may occur containing 

 only a few. A subsystem may become stable before all the rest 

 are stable. By the usual rule such stable subsystems will tend 

 to be self-preserving. There is therefore the possibility that the 

 multistable system will adapt piecemeal, its final adaptation 

 resembling that of a collection of iterated ultrastable systems, 

 like that of Figure 12/1/2. The present system will, however, 

 differ in that the constancies that divide subsystem from sub- 

 system are not unalterable but conditional. 



Such a multistable system, having arranged itself as a set 

 of iterated systems, will show the features previously noticed 

 (S. 12/2) : its adaptation will be graduated ; it can conserve its 

 old adaptations while developing new ; and, most important, the 

 time taken before all its variables become stabilised will 

 be reduced from the impossibly long to the reasonably short 

 (S. 12/4). 



This is what may happen ; but will it actually occur ? The 

 tendency to adaptation may be persistent, but why should the 

 process take the favourable course ? First we notice that as 

 adaptation in some form or other is inevitable the only question 

 is what form it will take. For simplicity, consider an eight- 

 variable environment that can be stabilised either in two inde- 



184 



