17/4 ANCILLARY REGULATIONS 



main variables change. Too rapid a change at the step- 

 mechanisms means that the appropriateness (or not) of a set of 

 values does not have time to be communicated round, through 

 the brain and environment as they carry out the trial, to the 

 essential variables, which would thus be acting before the arrival 

 of their necessary information. If it takes ten seconds for the 

 goodness of a trial to be tested, then alterations should obviously 

 not be made more frequently than at about eleven-second intervals. 

 And if it takes ten years to observe adequately the effect of a 

 profound re-organisation of a Civil Service, then such re-organisa- 

 tions ought not to occur more frequently than at eleven-year 

 intervals. The amount of communication from essential variables 

 to step-functions can thus become harmful if excessive. 



(2) In Chapter 10 we considered how the organism could take 

 advantage of the recurrent situation, so that if, having adapted 

 first to A and then to B, A were presented again, it could produce 

 the behaviour appropriate to A at once. It was shown in S. 10/8 

 that during the adaptation to B, the step-mechanisms concerned 

 with the adaptation to A must not be affected by what happens 

 at the essential variables. The allowing of such communication 

 would thus be harmful. 



(3) In S. 16/11 it was shown that a multistable system's 

 chance of getting adapted in a reasonably short time is closely 

 related to its approximation to the iterated form. Thus every 

 addition of channels of communication takes the system further 

 from the iterated form and, whatever else it may do, increases 

 the time taken to arrive at adaptation. 



Thus, in adapting systems, there are occasions when an increase 

 in the amount of communication can be harmful. 



17/4. It may still be objected that Figure 16/6/1 should show 

 connexions directly between the reacting parts, because such 

 connexions are necessary for co-ordination to be achieved between 

 part and part. The objection in fact is mistaken; connexions are 

 not necessary. Let me explain. 



First we can dismiss at once the case in which the parts of the 

 environment (as in Figure 15/7/1) are not joined; for then the 

 threats to the various essential variables come independently and 

 can be responded to independently. In this case the necessity 

 for co-ordination between parts does not arise. 



219 



