662 BRAIN MECHANISMS AND LEARNING 



three times as many as would be needed in other circmnstanccs — because, prior to 

 that experience, he had arranged things so as to direct the animal's attention to the 

 size of the diagrams only; the necessary sensory stimulation occurred, but the 

 central factor of attending was absent. Attention is fundamental; where does it fit 

 in such theories as the ones we are discussing now; 



There are the phenomena of visual transfer and patterning. These theories talk 

 as though the stimulus input from one object went to one part of the brain, from 

 another to another part: a spatial separation is assumed for different stimulus 

 objects. But inevitably there is overlap of excitations and the differentiation be- 

 tween objects is a matter of patterning, not the gross locus of stimulation or 

 central conduction. 



When theories can deal with that kind ot problem they will be getting closer to 

 the realities of behaviour. There are the problems ot time relations in serial order, 

 and the fact that most learning is not a one-shot output with a single input; charac- 

 teristically it is serialized, each reaction producing further stimulation, with further 

 reaction, and so on. As Lashley showed in his Hixon symposium paper, the time 

 properties of conduction, and reaction time, are such as to prevent us from treating 

 serialized behaviour as simply a series of conditioned reffexes; there is an intra- 

 cerebral serialization as well. This aspect of the problem is directly related to the 

 question of ideation, as demonstrated experimentally by Broadbent's holding 

 phenomenon. 



These conditioned-reflex theories assume a one-way, through transmission m 

 CNS, with the exception that as I understand Dr Gerard, Beurle's theory postulates 

 a reverberatory process going on until a 'satisfactory outlet' is achieved, at which 

 point it would stop. What stops it, in physiological terms? The whole question ot 

 reinforcement is raised here, and this is a difficult question indeed. There is the fact, 

 for example, that frequent repetition of a response may decrease, instead of increase, 

 the probability that that response will be made again (and this without overt punish- 

 ment of any kind). The phenomena oi memory and forgetting and the facts to 

 which I referred briefly in earlier discussion (i.e. the conditions in which visual 

 learning is or is not retained), all these and other tactual data have to be taken 

 account of When there is so much selection, in the phenomena on which a theory 

 of learning is based, it cannot be taken seriously. 



EccLES. I am very grateful for these criticisms, because they give me a chance to 

 show that the hypothesis is much more adaptable than would appear at first sight. 

 The full potentialities of the hypothesis are not revealed by the simplified account 

 I gave. For example, the neuronal networks CC and NN would be very widely 

 dispersed over the cortex and involve millions of neurones, as would necessarily be 

 the case if convergence is to occur between different sensory modalities and to be 

 so highly specific, as for example occurs with conditioning for an ellipse and not for 

 a circle. Multiple representation of the neuronal networks is also not excluded, 

 hence conditioned reflexes could survive extensive cerebral lesions and ablations. 

 In fact the hypothesis will encompass any sensory experience that can be analysed 



