DESIGN FOR A BRAIN 14/8 



Minor disturbances 



14/8. Exactly the same type of argument — of looking for what 

 can be terminal — can be used when S is not an accurately repeated 

 stimulus but is, on each application, a sample from a set of dis- 

 turbances having some definite distribution. In this case, Figure 

 14/3/1, for example, would remain unchanged in its confluents 

 and states of equilibrium, but the arrow going from each state of 

 equilibrium would lose its uniqueness and become a cluster or 

 distribution of arrows from which, at each disturbance, some one 

 would be selected by some process of sampling. 



The outcome is similar. The equilibrium whose arrows all go 

 far away to other confluents is soon left by the representative 

 point; while the equilibrium whose arrows end wholly within its 

 own confluent acts as a trap for it. Thus the polystable system 

 (if free from cycles) goes selectively to such equilibria as are 

 immune to the action of small irregular disturbances. 



14/9. The fields (of the main variables) selected by the ultrastable 

 system are subject to this fact. Thus, consider the three fields 

 of Figure 14/9/1 as they might have occurred as terminal fields 

 in Figure 7/23/1. In fields A and C the undisturbed representa- 



Figure 14/9/1 : Three fields of an ultrastable system, differing in their 

 liability to change when the system is subjected to small random dis- 

 turbances. (The critical states are shown by the dots.) 



tive points will go to, and remain at, the states of equilibrium. 

 When they are there, a leftwards displacement sufficient to 

 cause the representative point of A to encounter the critical states 

 may be insufficient if applied to C; so C's field may survive 

 a displacement that destroyed A's. Similarly a displacement ap- 

 plied to the representative point on the cycle in B is more likely 



190 



