13/20 THE SYSTEM WITH LOCAL STABILITIES 



scattered elements will tend to have more immunity to localised 

 injury than one whose elements are few and compact. 



13/20. Lashley had noticed this possibility in 1929, remarking 

 that the memory-traces might be localised individually without 

 conflicting with the main facts, provided there were many traces 

 and that they were scattered widely over the cerebral cortex, 

 unified functionally but not anatomically. He did not, how- 

 ever, develop the possibility further; and the reason is not far 

 to seek when one considers its implications. 



Such a localisation would, of course, be untidy; but mere un- 

 tidiness as such matters little. Thus, in a car factory the spare 

 parts might be kept so that rear lamps were stored next to radia- 

 tors, and ash-trays next to grease guns; but the lack of obvious 

 order would not matter if in some way every item could be 

 produced when wanted. More serious in the cortex are the effects 

 of adding a second reaction; for merely random dispersion provides 

 no means for relating their locations. It not only allows related 

 reactions to activate widely separated variables, but it has no 

 means of keeping unrelated reactions apart; it even allows them 

 to use common variables. We cannot assume that unrelated 

 reactions will always differ sufficiently in their sensory forms to 

 ensure that the resulting activations stay always apart, for two 

 stimuli may be unrelated yet closely similar. Nor is the differen- 

 tiation trivial, for it includes the problem of deciding whether a 

 few vertical stripes in a jungle belong to some reeds or to a tiger. 



Not only does random dispersion lead to the intermingling of sub- 

 systems, with abundant chances of random interaction and con- 

 fusion, but even more confusion is added with every fresh act of 

 learning. Even if some order has been established among the 

 previous reactions, each addition of a new reaction is preceded 

 by a period of random trial and error which will necessarily cause 

 the changing of step-mechanisms which were already adjusted to 

 previous reactions, which will be thereby upset. At first sight, 

 then, such a system might well seem doomed to fall into chaos. 

 Nevertheless, I hope to show, from S. 15/4 on, that there are 

 good reasons for believing that its tendency will actually be 

 towards ever-increasing adaptation. 



183 



