11/7 DESIGN FOR A BRAIN 



But if N is twenty the time becomes 3,200,000,000 centuries, which 

 for our purpose, is equivalent to ' never '. Other examples, 

 though quantitatively different, would lead to the same general 

 conclusion : when the number of environments is more than a few, 

 the time taken by this method to find a field stable to all exceeds 

 the allowable. Evidently our brains do not use this method : 

 success by it is too improbable. 



11/7. In the previous section we regarded the animal as having 

 to adapt to a variety of environments, but we can also regard 

 them as constituting a single 4 total ' environment. This makes 

 the number of variables in the system increase. What will be 

 the effect of this increase on the time taken to find a terminal 

 field ? For instance, could the homeostat adapt if it consisted 

 of a hundred units instead of four ? The question cannot be 

 ignored, for the human brain contains about 10,000,000,000 

 nerve-cells, and to this we must add the number of variables in its 

 environment. What is the chance that a field should be terminal 

 when it occurs in a system with this number of variables ? 



If the system worked as a magnified homeostat then, although 

 exact calculation is impossible, the evidence, reviewed in S. 20/12, 

 is sufficient to show that, for practical purposes, there is no chance 

 at all. If we were like homeostats, waiting till one field gave us, 

 at a stroke, all our adult adaptation, we would wait for ever. 

 But the infant does not wait for ever ; on the contrary, the 

 probability that he will develop a full adult adaptation within 

 twenty years is near to unity. Some extra factor must therefore 

 be added if the large ultrastable system is to get adapted within 

 a reasonable time. 



11/8. It may seem that we have now proved that the whole 

 solution must be wrong. But if we re-trace the argument, we find 

 that to some extent the difficulty has been unnecessarily magnified. 

 From S. 8/6 onwards we assumed for convenience of discussion 

 that every main variable was in full dynamic interaction with 

 every other main variable, so that every change in every variable 

 at once affected every other variable. This gives a system that is 

 extremely active and that unquestionably acts as a whole, not as 

 a collection of small parts acting independently. As an intro- 

 duction it has distinct advantages, but it raises its own difficulties. 



136 



