J. OLDS AND M. E. OLDS I71 



this finding emphasized what we should have known at first: response 

 learning involves, hrst, cessation of wrong behaviours, secondly, com- 

 mencement of correct ones, and thirdly, repetition of correct ones. A 

 stimulus causing repetition of behaviour would only promote the thu-d 

 stage of the process, which in a sense is not learning at all but its antithesis. 



But again there was still room for doubt. Our stimulation of the large 

 hypothalamic system causes repetition and inhibits adaptive changes in 

 behaviour. In a sense, this is one and the same thing; by repeating, the 

 animal cannot be adaptive, changing. This is consistent; but we must 

 sometimes beware of consistency. 



Our animals could repeatedly get the brain stimulus (which is all the 

 repetition they want) and yet adcipt the correct response to get food at the 

 same time; and they appear to be trying to get food. Why does the 

 response fail to become stereotyped down one lane? One might say there 

 is a reinforcement ot the antecedent so per cent probability. But then why 

 in the memory tests where the learning occurs first does the stimulus cause 

 return to 50 per cent probability: Or even more strikingly, why, after the 

 animal learns in reinforcement-interference tests, can he run successfully 

 under stimulation, to food and back, with no errors even though the 

 stimulation continues? The same animal running currently in the memory 

 test starts and continues to make errors upon introduction ol the stimulus. 



The possibility we want to suggest is this: our stimulus might cause 

 some confusion in addition to the output of positive motivation. In the 

 reinforcement-interference test, the strong positive motivation emanating 

 from hypothalamic stimulation keeps behaviour in line. But minor prob- 

 lems cannot be solved because of confusing effect. It might be, then, that 

 total activity in this hypothalamic-rhinencephalic system determines the 

 positive-reinforcement function; but the pattern of activity in the same 

 system is precisely the more or less lasting process that is altered by learning. 



Giving up, then, the idea that positive reinforcement promotes learning, 

 we are left with alternative notions: either it is a simple inhibitor of 

 learning or else it is somehow a function of the same neural aggregates 

 which are chiefly involved in learning; but quantity is most important 

 from a standpoint of reintorccmcnt, the patterning most important from 

 the standpoint of learning. 



From the latter point of view, stimulation of the system could be said to 

 augment the positive reinforcement and the repetitiousness of the whole 

 system, but to disrupt the patterns formed by association of mild cues. 



The final experiment we have to report searched for ways of causing 

 learned changes in neural patterns by operant-conditioning techniques. 



