THE MULTISTABLE SYSTEM 16/7 



and apparent confusion. Travel along the other line, from Q to 

 R, will also activate various combinations of subsystems ; and 

 the set made active in the second line may be very different from 

 that made active by the first. 



In such conditions it is no longer profitable to observe par- 

 ticular subsystems when a multi stable system adapts. What 

 will happen is that instability, and consequent step-function 

 change, will cause combination after combination of subsystems 

 to become active. So long as instability persists, so long will 

 new combinations arise. But when a stable field arises not 

 causing step-functions to change, it will, as usual, be retained. 

 If now the multistable system's adaptation be tested by dis- 

 placements of its representative point, the system will be found 

 to respond by various activities of various subsystems, all co- 

 ordinated to the common end. But though co-ordinated in this 

 way, there will, in general, be no simple relation between the 

 actions of subsystem on subsystem : knowing which subsystems 

 were activated on one line of behaviour, and how they interacted, 

 gives no certainty about which will be activated on some other 

 line of behaviour, or how they will interact. 



Later I shall refer again to ' subsystem A adapting to, or 

 interacting with, subsystem B ', but this will be only a form of 

 words, convenient for description : it is to be understood that 

 what is A and what is B may change from moment to moment. 



16/7. In S. 12/4 it was shown that the division of a system 

 into parts reduced markedly the time necessary for adaptation. 

 The multistable system, being able to adapt by parts (S. 16/5), 

 can adapt by this quicker method. But no reason has yet been 

 given why this quicker method should be taken if offered. There 

 is, however, a well-known principle which ensures this. 



When changes can occur by two processes which differ in their 

 speeds of achievement, the faster process, by depriving the 

 slower of material, will convert more material than the slow ; 

 and if we imagine the material marked in some way according 

 to its mode of change, then the major part of the material will 

 bear the mark of the faster process. If the difference between 

 the speeds is great, then for practical purposes the slow process 

 may not be in evidence at all. The important fact here is that 

 we can predict a priori that if the change be examined, it will 



177 



