38 BRAIN MECHANISMS AND LEARNING 



mammalian behaviour involves an enormous number of synapses, the 

 changes that occur simultaneously may be in opposite directions, and 

 there may be little detectable relation between the course of events at any 

 one synapse and changes of overt behaviour —just as there may be little 

 relation between the activity of an individual neurone and the gross record 

 of the EEC 



If we assume that many of the neurones in the brain of a waking 

 mammal arc tning at any given time, and especially if we further assume 

 that some of these are inhibitory, there are certain consequences which 

 have sometimes been overlooked in physiological discussions of learning. 

 The possession of a large brain capable of learning a great many different 

 things inevitably nieans that there are far more neurones present than is 

 necessary for learning some one specific task. Any random activity in 

 these excess neurones (the ones not needed for the task being learned) is 

 'noise', which must tend to interfere with the learning. (If the activity is 

 organized instead of random it is not noise, technically speaking, but the 

 effect may still tend to be adverse.) It seems obvious that the number of 

 excess neurones must be very much greater — perhaps thousands of times 

 greater in the brain of the higher animal — than those needed for the 

 learning going on at the moment. It therefore seems that the rate of 

 learning, as observed in the behaviour of the whole animal, may be not 

 an index of capacity tor adding new synaptic connections so much as an 

 index of the noise level, and that learning will be fast or slow according as 

 one is successful in establishing an environmental control of the excess 

 neural activity, in order to prevent or niinimize interference. 



Practically, the significance of this point of view is clear in the term 'lack 

 of concentration', in the case of the student whose learning is inefficient 

 even in a quiet environment, because other thoughts obtrude besides the 

 ones he should be concerned with. Experimentally, the point is made by 

 Ricci, Doane andjasper (1957), in reporting that a significant part of the 

 conditioning process — perhaps the main part — is in the dropping out of 

 irrelevant connections, rather than the acquisition of new ones. It is 

 clearest of all in a classic experiment of Yerkes. 



Yerkes (19 12) trained an earthworm to choose one arm of a T-maze, 

 using electric shock as punishment for error and the moist burrow as 

 reward for correct choice. The habit was acquired in twenty trials, 2 days 

 at ten trials per day, about what might be necessary for the laboratory rat. 

 No errors were made on the third day, though behaviour was somewhat 

 inconstant in the following week as between good days and bad days 

 (even worms have them). Yerkes then removed the brain, or principal 



