GENERAL DISCUSSION 657 



the lowering ot threshold is non-specihc tor all synapses on a neurone. This is the 

 problem of redundant conditioning, as it is appropriately called by Shimbel. 



The hypothesis that I proposed in 1952 was built upon properties ot neuronal 

 networks that derived troni the convergence of synaptic excitatory action on 

 neurones and also upon the postulate that synaptic use resulted in an enduring trace 

 ot increased etiicacy. At that time there was little direct evidence for this latter 

 postulate, but very convincing experimental evidence has now been obtained in 

 mv laboratory, and presented earlier in this symposium. The essential tcatures ot 

 the hypothesis are displayed in this diagram (Fig. iC), but it is important to realize 

 that the actual neural events occur in the neuronal networks of the diagram which 

 are shown as black boxes but are of unimaginable complexity. Hence it is import- 

 ant to make brief reference to the properties that these networks are believed to 

 have. 



hi order to generate the discharge ot an impulse trom a neurone it is usually 

 necessary that there be convergence onto it of several excitatory nerve impulses 

 within a few milliseconds. Thus eiiective neuronal action does not occur for a 

 serial one to one arrangement of neurones. Rather the propagation in a neuronal 

 network resembles an advancing front of traliic with many cells activated in 

 parallel at each synaptic linkage in the chain. With such an arrangement it can be 

 showni that two completely different inputs into a neuronal network can involve 

 many neurones in common and yet emerge as completely ditFerent outputs, i.e. 

 the same neurones can participate in countless specific patterns of activity. A further 

 feature is that closed loops of activated neurones are possible in which impulses 

 circulate again and again over approximately the same neuronal pathway being 

 thus responsible for the reverberatory activity that has otten been postulated. 



In the diagram (Fig. iC) there is firstly the unconditioned retlex pathway US- 

 RC-Rjust as in A, but this pathway is assumed also to branch so that it converges 

 on to the same complex centre of neurones as does the conditioned stimulus, CS. 

 hi this convergence centre (CC) it is postulated that two changes are brought about : 

 there is synaptic facilitation so that man) more neurones and neuronal pathways 

 are excited to discharge impulses than would be by either CS or US alone; and 

 there is increased etiicacy of synaptic knobs on account of the excess use, those 

 specially activated as a consequence ot the summation of CS and US being 

 particularly important. As a consequence ot these two factors there will be a 

 specific pattern of activation of CC as shown diagrammatically b\' the hatched 

 pathway and this will be further elaborated in the neuronal network NN that 

 leads to the etFerent pathway, RC to R. Since the spatio-temporal pattern oi 

 neuronal activation in CC is dependent in part on US, it would be expected that, 

 as shown in the diagram, it would in part achieve ctierent expression in the same 

 RC as occurs tor the unconditioned reticx evoked by US. It is evident that, if the 

 discharges of impulses into NN are set up by CS and US in close temporal sequence 

 and with an adequate number of repetitions, the increase in synaptic etiicacy in the 

 hatched zone both in the convergence centre (CC) and in the neuronal network 



