658 BRAIN MECHANISMS AND LEARNING 



(NN) will cause impulses from CS alone to be effective m evoking a pattern of 

 impulses in NN that leads to the response R, i.e. a conditioned reflex has been 

 established. Of course the distinction between CC and NN in the diagram is 

 arbitrary. It is done merely to separate the predominantly afferent function of CC 

 and the more integrative and efferent function ot NN, where we can think of the 

 eventual response as being synthesized. There is no more time for elaboration ot 

 the postulate so that it can be made to account for many of the phenomena of 

 conditioning. Possibly this will be possible when attempting to answer criticisms. 



Gerard. Before we consider Dr Eccles's model let me put before yoti the one 

 by Beurle to which I have referred several times. The two are basically alike: 

 Beurle's is worked out in rather more detail. (For a fuller exposition, and the 

 physiological evidence bearing on the model, see my summary chapter in the 

 Handbook of NctiropliysioU\^y). Beurle assumes a population of neurones, randomly 

 placed in a layer (such as the cortex) and randomly connected with each other, 

 except for a progressive falling oft of connections with distance of separation. Each 

 neurone receives impulses from many and passes on impulses to many. Several 

 incoming impulses are needed to fire a cell : thresholds at specific synapses tend to 

 tall slowly with repeated activation: and thresholds may be temporarily and 

 reversibly raised or lowered by the action of inhibitory or facilitatory elements. 



A wave of activity, entering such a cell mass, does not engage all cells in its path; 

 but it does tend to engage a progressively large fraction until all cells are active, or 

 a smaller fraction until the wave dies out, unless there is regulation by a negative 

 feedback loop. With this, a wave ot constant intensity can travel through the mass, 

 the fraction of cells engaged in any vollime depending on the past history of the 

 system and on the current action of positive and negative feedback systems. 

 Moreover, two or more individually characteristic waves may pass through the 

 same neurone mass, each engaging its own special group of neurones. Two such 

 waves might, therefore, travel simultaneously through the neurone pool, cross one 

 another, reflect upon each other or from some other discontinuity in the mass, 

 and facilitate one another at the trajectory of crossing wave fronts. 



hidccd, because of the more than linear building up oi excitation with the 

 number of active neurones, the locus ot intercepting waves can serve as a source of 

 new waves, reflecting back through the cell mass along the route of either of the 

 incoming waves. With sufllcient threshold lowering by repeated activity, the 

 characteristic waves can originate at such loci even without specific incoming 

 stimulation. Here is the beginning ot a neural mechanism tor memory association, 

 recall and even imagination. If recurrent connections carried messages back from 

 the output of the mass to the input region, internal testing of alternate behaviours, 

 rather than external testing, becomes possible, and the model accounts for reason 

 based on imagination and memory, and is able to choose the 'good' response for 

 its first act. 



All the assumptions here are based on legitimate neurophysiological knowledge; 

 even the existence of a feedback loop (through the diftiise neural system of retictilar 



