SERIAL ADAPTATION 17/3 



ment, M 1 ; those immediately affected by the second, M 2 ; those 

 step-functions which changed during the second adaptation, S 2 ; 

 and those main variables that S 2 directly affects, M 3 . It is not 

 assumed that the M-classes are exclusive. 



After the step-functions S 2 have changed value, the behaviours 

 of the variables M x are unchanged, by postulate 2 ; so M x is in- 

 dependent of *S 2 (S. 14/3). But S 2 affects M 3 ; so M x must be 

 independent of M 3 (S. 14/5). There must therefore be a wall of 

 constancies between them (S. 14/8), which must be only temporary, 

 by postulate 3. We can deduce therefore that some of the main 

 variables must be part-functions. 



Since M 1 is independent of S 2 , it follows that the step-functions 

 S 2 can have no immediate effect on the main variables M v In 

 other words, some of the step-functions' immediate effects are 

 restricted to a few of the main variables. 



If we now use the fourth postulate, that these particular main 

 variables and step-functions are typical, it follows that part- 

 functions must be common, and step-functions must usually be 

 restricted in the variables they immediately affect. We conclude, 

 therefore, that if the nervous system is to show the listed proper- 

 ties, the main features of the multistable system are necessary. 



17/3. We can now start to examine the thesis that the nervous 

 system is approximately multistable. We assume it to be joined 

 to an environment that contains many part-functions, and we 

 ask to what extent the thesis can explain not only elementary 

 adaptation of the type considered earlier but also the more complex 

 adaptations of the higher animals, found earlier to be beyond the 

 power of a simple system like the homeostat. 



We may conveniently divide the discussion into stages accord- 

 ing to the complexity of the environment. First there is the 

 environment that, though perhaps extensive, is really simple, for 

 it consists of many parts that are independent, so that they can 

 be adapted to separately. Such an environment was sketched 

 in Figure 12/1/2. It will be considered in this section. Then 

 there is the environment that has some connection between its 

 parts but where the adaptation can proceed from one part to 

 another, perhaps in some order. It will be considered in the 

 remainder of this chapter. Then there is the environment that 

 is richly interconnected but in which there is still some transient 



183 



