16/8 ADAPTATION IN THE MULTISTABLE SYSTEM 



u j'k ' l » m»- T7 



-\j — \r 



v 



Tim<z 



Figure 16/8/1 : Three units of the Homeostat interacting. Bars in the 

 central positions prevent 2 and 3 from moving in the direction corre- 

 sponding here to upwards. Vertical strokes on U record changes of 

 uniselector position in unit 1. Disturbance D, made by the operator, 

 demonstrates the whole's stability. 



1 and 3 interacted, and 2 was independent. 1 was set to act on 



2 negatively and on 3 positively, while the effects 2 — > 1 and 



3 — > 1 were uniselector-controlled. 



When switched on, at J, 1 and 2 formed an unstable system 

 and the critical state was transgressed. The next uniselector 

 connexions (K) made 1 and 2 stable, but 1 and 3 were unstable. 

 This led to the next position (L) where 1 and 3 were stable but 

 1 and 2 became again unstable. The next position (M) did 

 not remedy this; but the following position (N) happened to 

 provide connexions which made both systems stable. The values 

 of the step-mechanisms are now permanent; 1 can interact re- 

 peatedly with both 2 and 3 without loss of stability. 



It has already been noticed that if A, B and C should form 

 from time to time a triple combination, then the step-mechanisms 

 of all three parts will stop changing when, and only when, the 

 triple combination has a stable field. But we can go further 

 than that. If A, B and C should join intermittently in various 

 ways, sometimes joining as pairs, sometimes as a triple, and 

 sometimes remaining independent, then their step-mechanisms 

 will stop changing when, and only when, they arrive at a set of 

 values which gives stability to all the arrangements. 



Clearly the same line of reasoning will apply no matter how 

 many subsystems interact or in what groups or patterns they 



211 



