5/14 



AN INTRODUCTION TO CYBERNETICS 



not, and four stepping-switches, each of which can be in any one 

 of 25 positions (not shown accurately in the Figure). Each position 

 carries a resistor of some value. So B has 2 x 25 x25 x 25 x 25, 

 i.e. 781250, states. To this system yi is input, fi has been built so 

 that, with the relay energised, none of 5's states is equilibrial (i.e. the 

 switches keep moving), while, with the relay not energised, all are 

 equilibrial (i.e. all switches stay where they are). 



Finally, B has been coupled to A so that the relay is non-energised 

 when and only when A is stable at the central positions. 



When a problem is set (by a change of value at some input to A 

 not shown formally in the Figure), A has a variety of possible states 

 of equilibrium, some with the needles at the central positions, some 

 with the needles fully diverged. The whole will go to some state of 

 equilibrium. An equilibrium of the whole implies that B must be 



A 



B 



ti^ 



U 



• • • a' -♦ • 



Fig. 5/14/1 



in equilibrium, by the principle of the previous section. But B has 

 been made so that this occurs only when the relay is non-energised. 

 And B has been coupled to A so that the relay is non-energised only 

 when yi's needles are at or near the centres. Thus the attachment 

 of B vetoes all of ^'s equilibria except such as have the needles at 

 the centre. 



It will now be seen that every graph shown in Design . . . could 

 have been summed up by one description: "trajectory of a system 

 running to a state of equilibrium". The homeostat, in a sense, does 

 nothing more than run to a state of equilibrium. What Design . . . 

 showed was that this simple phrase may cover many intricate and 

 interesting ways of behaving, many of them of high interest in 

 physiology and psychology. 



(The subject of "stability" recurs frequently, especially in S.9/6, 

 10/4, 12/1 1 ; that of the homeostat is taken up again in S. 12/15.) 



84 



