DESIGN FOR A BRAIN 9/14 



Dancoff s principle — that systems made efficient by natural selec- 

 tion will not use variety or channel capacity much in excess of 

 the minimum — then we can deduce that when organisms regularly 

 use the method of trials there is strong presumptive (though not 

 conclusive) evidence that their trials will be controlled by material 

 entities having (relative to the rest of the system) not much more 

 than the minimal variety. There is therefore strong presumptive 

 evidence that the significant variables in S (of Figure 7/5/1) are 

 step-functions, and that the material entities embodying them are 

 of such a nature as will easily show such functional forms. 



Levels of feedback 



9/14. We can next consider how the formulation of Chapter 7, 

 and Figure 7/5/1, compares with the real organism's organisation 

 in respect of the division of the feedbacks into two clearly dis- 

 tinguishable orders: between organism and environment by the 

 usual sensory and motor channels, and that passing through the 

 essential variables and step-mechanisms to the reacting part R. 



Chapter 7 followed the strategy described in S. 2/17: we 

 attempted to get a type-system perfectly clear, so that it would 

 act as a suitable reference for many real systems that do not 

 correspond to it exactly. To get a clear case we assumed that 

 the system (organism and environment joined) was subject to 

 just two types of disturbance from outside. Of one type is the 

 impulsive disturbance to the system's main variables ; by this their 

 state is displaced to some non-equilibrial position; this happens if 

 a Homeostat's needle is pushed away from the centre, as at D x 

 in Figure 8/4/1 ; or if the fire by the kitten suddenly blazes up. 

 Then the organism, if adapted, demonstrates its adaptation by 

 taking the action appropriate to the new state. A number of 

 such impulsive disturbances, each with an interval for reaction 

 to occur, are necessary if the organism's adaptation is to be tested 

 and demonstrated. 



Of the other type is the disturbance in which some parameter 

 to the whole system is changed (from the value it had over the 

 many impulsive disturbances, to some new value). This change 

 stands in quite a different relation to the system from the change 

 implied by the impulse. Whereas the impulse made the system 

 demonstrate its stability, the change at the parameter made the 



130 



