660 BRAIN MECHANISMS AND LEARNING 



Anokhin. I should like to add something to the remarks made by Dr Konorski. 



It is true that Pavlov believed that the role ot the cortex of the brain is a decisive 

 one in the creation of a conditioned reflex. However, his views were not absolutely 

 categorical on this point. He recognized clearly that under certain circumstances 

 the conditioning may be carried out bv the subcortical system also. 



hi normal conditions, however, when the animal has a normal cortex, condition- 

 ing is always carried out under the decisive influence of the cortex. 



hi fact, even in those cases when one can note elcctrophysiologicallv a condi- 

 tioned reaction in the subcortical system also, this reaction is not initiated there, but 

 has been tormcd in the very preliminary stage ot the orientating-experimental 

 reaction under the direct control of the cortex. 



Segundo. I feel that the essence of theories put forward by Dr Eccles and Dr 

 Clerard is that the passage of a wave leaves the pathway more permeable. 1 would 

 like to ask whether these schemes would be adaptable to habituation, a process 

 which apparently involves the closure ot permeable pathways? And if so, how? 



Fessard. I would like to add some comments in line with what has been said on 

 the use of models for the understanding of learning mechanisms. I think that 

 recourse to models — either mathematical or physical — is useful, given the fact 

 that the multitude of elements and extreme complexity ot their interconnections 

 leave no hope that we can ever know every detail of a neuronal field and its 

 functional properties. Models help to schematize the main dynamic features of a 

 field but may also be misleading. This is the case, I believe, of models which 

 exhibit a complicated and diversified circuitry as in a radio set: such 'machines' 

 may imitate the different modalities of conditioning, they may 'learn' but they are 

 of little help to give us a better understanding of how the CNS actuallv works 

 during learning operations. This is why probabilistic theories applied to more or 

 less ciiftuse neuronal networks are to be preferred, hi this respect, the mathematical 

 model by 13eurle, as recalled by I3)r Gerard, is certainly a remarkable and liighK 

 suggestive one. 



I also think that some useful qualitative and unsophisticated evidence must be 

 ascertained before any mathematical treatment is attempted, when one considers 

 the most fimiliar connectivity patterns, especially those exhibited by the wideK- 

 distributed network-like structures This is quite apparent when we consider the 

 illuminating diagram I3)r Eccles has just presented to us. I made use of similar 

 notions in a report with Dr Gastaut (Symposium on 'Conditionnement et Appren- 

 tissage, Strasbourg, 1956; Presses Univ. France, 195H, p. 216) where wc had first 

 gathered all the experimental evidence then available which appeared to show the 

 important role played by subcortical polyneuronic nuclei — namely, mesencephalic 

 reticular formation and so-called 'diffuse thalamic system' — in the formation of 

 new conditioned reflexes. This justifies our belief that some speculation should be 

 made on the general properties of neuronal networks. By their very nature the)' 

 are sites of election for convergences and interactions of heterosensory signals. Our 

 schema differs from the one of Eccles onh' in that he makes a difference between a 



