neurophysiology: an integration 



'957 



Clearly, therefore, without additional regulation 

 there can be no uniform maintained cell activity nor 

 regular waves traveling through the mass. Waves 

 could arise here and there from active foci and either 

 spread a way before dying out or rise to maximal 

 intensity and spread indefinitely (convulsive or ictal 

 activity). Beurle later notes (personal communication) 

 that inhibitory components are essential for stability. 

 The wave will decrement or increment sharply, 10 

 time units (iot) will suffice for it to rise to saturation, 

 or fall to zero, if the threshold is, respectively, only 

 one per cent above, or below, the critical value. 



Since used cells lie in the wake of a wave front, 

 there are always more sensitive cells ahead than 

 behind, and the wave passes forward as a true exci- 

 tation wave. The used cells recover, however, so 

 saturated waves can recur at a frequency determined 

 by the recovery period, and unsaturated ones can 

 recur even faster. Waves can thus return to or even 

 cross in a given region. Moreover, since only some 

 cells are used by unsaturated waves, each wave 

 engaging its own unique set, waves of the same size 

 and course can travel through the same cell mass and 

 yet differ in detail in the actual cells used, and so 

 possess individual identities. The detailed shape and 

 movement of a wave would depend, further, on the 

 actual shape of the cell mass. A wave originating at a 

 point would tend to spread sphericallv in a homo- 

 geneous mass; but on reaching the surfaces of a 

 sheet, it would continue to spread in it as a cylindrical 

 or linear wave front. Clearly, for such activity a thick 

 flat cortex would be far from equivalent to a thin 

 convoluted one. 



Controls. If a special set of excitatory (E) axons 

 penetrated the cell mass, each able to contribute i 

 partial pre-excitation of cells in the mass, ,1 traveling 

 wave would be facilitated when such axons were 

 active. Comparably, a set of inhibitory (I) fibers 

 which would predischarge a few neurons (or a 

 subset of I cells, especially easily discharged) would 

 reduce the density of responsive cells and so tend to 

 inhibit a wave. Even simpler, I fibers may directly 

 inhibit, aside from predischarge of cells, an assumption 

 Beurle was earlier advised to eschew but now prefers 

 (personal communication). Either or both mech- 

 anisms would suffice for regulation, the situation is 

 svmmetrical. Finally, with a set of axons from the 

 main cell mass to the special E or I cells, acting as a 

 servo or feed-back control, a wave could be stabilized 

 so that it would travel through the mass at some 

 constant value well below saturation. Such a wave, 

 now entering a cell region where travel is more 



difficult, will progressively attenuate until only one 

 final cell is stimulated after the wave penetrates some 

 distance into the region. But, since each wave is 

 specific in the neurons it uses, even though all waves 

 are of the same total size, each will activate a different 

 terminal cell; in fact, which terminal cell is fired 

 could serve to define the wave. 



The point at which a wave arises clearly could 

 correspond to input to a central mass; the terminal 

 cell or area, to the output; and the region in which a 

 controlled constant wave moves, to the cortex or to 

 any other central nervous mass of neurons. If, now, 

 the assumption is introduced that each activation 

 leaves a minute lasting fall in the threshold of a cell, 

 a given wave will pass through the neuron mass more 

 rasilv each time it is evoked. Any other wave, even 

 through the same total neuron population, will not 

 be significantly enhanced, which easily accounts for 

 discriminative learning. 



All sensor) messages teach the cell mass and can 

 start waves. The further assumption of a discrimi- 

 nator, such that sensations which produce effects 

 desirable to tin- organism activate E while those 

 producing undesirable ones activate 1, helps explain 

 adaptive learning. In a familiar environment, with 

 usual imputs, waves will travel along established 

 facilitated 'paths' and give the usual motor responses. 

 In a new environment, with novel excitation, new- 

 waves will travel, a new motor unit be reached and 

 some different act follow. Il this leads to satisfactory 

 results, so producing a desired input, 'E' will be 

 activated; the wave will be reinforced for future 

 travel, and the response, 'learned. 3 II the result is 

 unsatisfactory, T will be activated , the wave will be 

 blocked or deflected to a different output cell; and 

 the 'trial and error' behavior will continue until a 

 satisfactory response is obtained. Since the presence 

 of a goal normally is implicit to the very process of 

 [earning, the only real assumption here is as to the 

 mechanism of the E or I effects. 



Reactivation properties. The model has further 

 richness, without additional assumptions. The regu- 

 lated wave, controlled by a feed-back circuit which 

 has a time lag, can reflect from the boundary of a 

 region of difficult travel — one with a lower cell 

 density or higher cell thresholds. Two waves can cross 

 without dying out; since activation is nonlinear, the 

 proportion of cells active will be more than doubled 

 where the waves meet. Best of all, waves meeting at 

 an angle will form a line of the loci of crossing of the 

 advancing wave fronts, a line at which excitation was 

 excessive. The cells in this track of the crossing point 



