44 



BRAIN MECHANISMS AND LEARNING 



repeat. Some change occurs which means that when the experimenter 

 stops speaking, the corresponding trace systems fire, and in the proper 

 order. Undoubtedly there are complexities which this does not take 

 account of; there is certainly not a mere chaining of the systems involved, 

 because — for example — the subject may be able to tell you what the 

 first and last of a group of digits were, though he has lost the intervening 

 ones. Memory for the individual item in the scries is not entirely depen- 

 dent on the preceding item. But the order of repetition seems to require a 

 changed synaptic relation between the specific cell-assembly groups, or 

 trace systems, so that in the example of Fig. 3, j tires 2, and not 4 or the 



Fig, 3 

 Diagrammatic representation of the synaptic 

 relations involved in the repetition of the 

 digits 3-2-1. The heavier arrows 1-2 and 2-3 

 represent more strongly established connec- 

 tions ; the encircled .v indicates the temporary 

 strengthening of less strongly established 

 ones. 



central representative of some other number; and 2 fires 1 and not j. The 

 synaptic 'strengthening' implied is shown by the encircled .v in the figure. 

 What can this be? 



The difficulty here is that all the synapses concerned must be already 

 asymptotic, in the development of their structural connections. Synapses 

 connecting system 2 with other systems are highly developed; if now 

 when 2 fires it is to fire 1 reliably, and not 4 or 6 or some other system, it 

 must be sii^^iifiavitly more closely related to 1 for the moment, and it seems 

 most unlikely that this closer relation can consist of a sudden further 

 development of the size of the knobs of the systeiii-2 — system- 1 synapses. 

 Some other mechanism besides growth of the knob, or its closer juxta- 

 position with the cell body, must be involved. 



The mechanism could well be that recently proposed by Milner (i959)- 

 He has pointed out that failure of conduction along some axon fibrils may 



