22/5 DESIGN FOR A BRAIN 



second point only in the value of x p : as the subsystem is absolute, 

 an interval t 2 will bring its variables again to x\, • • • , ®»» i-e. these 

 variables' behaviours are the same on the two lines. Now x p 

 either is, or is not, equal to x° r If unequal, then by definition 

 (S. 14/3) x 19 . . . , x n is independent of x v . So the behaviour 

 of x 19 . . . , x n over t 2 will show either that x' v = x® (i.e. that 



x p did not change over t ± ) or that x x x n is independent of 



x p . Similar tests with the other variables of the set x v , . . . , x s 

 will enable them to be divided into two classes : (1) those that 

 remained constant over t v and (2) those of which the subsystem 

 X l9 . . . , X n is independent. By hypothesis, class (2) may not 

 include all of x p , . . . , x g ; so class (1) is not void. 



When a field of x v . . . , x n changes, some parameter to this 

 system must have changed value. As x lt . . . , x n , x p , . . . , x 8 

 is isolated, the ' parameter ' can be none other than one or more of 

 x Pi . . . , x s . As the field has changed, the parameter cannot be 

 in class (2). At the change of field, therefore, at least one of 

 those in class (1) changed value. So class (1), and therefore the 

 set x p , . . . , Xs, contains at least one step-function. 



Reference 



Ashby, W. Ross. Principles of the self-organising dynamic system. Journal 

 of general Psychology, 37, 125 ; 1947. 



236 



