9/4 DESIGN FOR A BRAIN 



The tracing, however, deserves closer study. The action 2 — ► 3 

 was reversed at R, and the responses of 2 and 3 at S 2 demon- 

 strate this reversal ; for while at S ± they moved similarly, at S 2 

 they moved oppositely. Again, a comparison of the uniselector- 

 controlled action 1 — > 2 before and after R shows that whereas 

 beforehand 2 moved similarly to 1, afterwards it moved oppo- 

 sitely. The reversal in 2 — > 3, caused by the operator, thus 

 evoked a reversal in 1 — > 2 controlled by the uniselector. The 

 second reversal is compensatory to the first. 



The nervous system provides many illustrations of such a 

 series of events : first the established reaction, then an altera- 

 tion made in the environment by the experimenter, and finally 

 a reorganisation within the nervous system, compensating for 

 the experimental alteration. The homeostat can thus show, in 

 elementary form, this power of self-reorganisation. 



The necessity of ultrastability 



9/4. In the previous sections a few simple examples have sug- 

 gested that the adaptation of the living organism may be due 

 to ultrastability. But the argument has not excluded the possi- 

 bility that other theories might fit the facts equally well. I shall 

 now give, therefore, evidence to show that ultrastability is not 

 merely plausible but necessary : the organism must be ultra- 

 stable. 



First the primary assumptions : they are such as few scientists 

 would doubt. It is assumed that the organism and its environ- 

 ment form an absolute system, and that the organism sometimes 

 changes from one regular way of behaving to another. The 

 crucial question is whether we can prove that the organism's 

 mechanism must contain step-functions. In S. 22/5 is given 

 such a proof, stated in mathematical form ; but its theme is 

 simple and can be stated in plain words. 



Suppose a ' machine ' or experiment behaves regularly in one 

 way, and then suddenly changes to behaving in another way, 

 again regularly. Suppose, for instance, a pharmacologist, test- 

 ing the effect of a new drug on the frog's heart, finds at every 

 test all through one day that it causes the pulse-rate to lessen. 

 Next morning, taking records of the effect, he finds at every 



110 



