FULLY CONNECTED SYSTEMS 11/4 



has shown in endless ways that the easy problem is the one 

 whose components are few and independent, while the difficult 

 problem is the one with many components that form a complex 

 whole. So when confronted with environments of various ' diffi- 

 culties ', the ultras table system and the living organism are likely 

 to fail together. 



It seems, then, that the ultrastable system's modes of failure 

 support, rather than discredit, its claim to resemble the living 

 brain. 



11/2. Now we can turn to those features in which the simple ultra - 

 stable system, as represented by the homeostat, differs markedly 

 from the brain of the living organism. One obvious difference 

 is shown by the record of Figure 8/8/4, in which the homeostat 

 made four attempts at finding a terminal field. After its first 

 three trials its success was zero ; then, after its next trial, its 

 success was complete. The homeostat can show no gradation in 

 success, though this is almost universally observable in the living 

 organism : day by day a puppy becomes steadier on its legs ; 

 year by year a child improves its education. 



11/3. A second difference is seen in their powers of conservation. 

 If the homeostat adapts to an environment A and then to an 

 environment B, and is then returned to A again, it has no adapta- 

 tion immediately ready, for its old adaptation was destroyed in 

 the readjustments to B ; it does not even start with a tendency 

 to adapt more quickly than before : its second adaptation to 

 A takes place as though its first adaptation had never occurred. 

 This, of course, is not the case in living organisms, except perhaps 

 in the extremely primitive : a child, by learning what two times 

 three is, does not thereby destroy its acquired knowledge of what 

 is two times two. 



11/4. Although the homeostat, in adapting to B, usually de- 

 stroys its adaptation to A, this is not the case necessarily, and 

 we should notice a property, inherent in the ultrastable system, 

 that might enable it to adapt to more than one environment. 

 It will be described partly for its intrinsic interest, as it will be 

 referred to later, and partly to show that it is insufficient to 

 remove the main difficulty. 



133 K 



