An Experimental Study of the Behavior of Rats with Special Reference to the Principle of Reinforcement

Abstract

This experiment was performed in an attempt to clarify some of the basic preblems arising from the “paradox of reinforecement”.
In the situation of reinforcement at a fixed ration including the continuous reinforcements, the ratio and the absolute number of reinforcements, were changed step by step systematically with different experimental groups in accordance with the following plan:
Thus, there were twelve groups altogether in the two dimensional experimental design.
In treating the results, I examined the time of reaction in acquisition and extinction, the member of reactions in extinction and the latency in acquisition and extinction in their relation to one another.
The results showed that only under the condition in which the same number of reinforcements was given, the resistance to extinction with partial reinforcement was stronger than that with continuous reinforcement. On the other hand, Hull's principle of reinforcement was confirmed with respect to partial reinforcement when the number of reinforcements was taken into consideration. This is a blind spot which has hitherto been overlooked.
As the proportion of reinforcement became less and less, the resistance to extinction increased paripassu-within the range of comparatively law rations, provided the same nember of reinforcements was given. However, at a certain high ratio (a critical ratio), it began to decrease aud continued so with the increase in the proportion of reinforcement.
It would seem that at night ratios the extinction process goes on and partially counteracts the process of acquistion.
This is another blind spot that has been overlooked not only in the principle of reinforcement but also in the very principle of expectancy.
Generally speaking, on the one hand, the gradual change in the ratio of reinforcements produces a bidimensional curve like a parabola whose crest represents the point of criticalratio, and on the other hand, the number of reinforcements may function as an independent variable of the exponential curve in a given ratio group. Then, there is a question why the resistance to extinction with partial reinforcement is stronger than that with continuous reinforcement. In my opinion, this difference is due to secondary reinforcement. It seems that responses preceding the reinforced one bring out the secondary reinforcement.
From this point of view, Hull's equation should be re-examined and revised. There still remains a problem of ‘unit’ hypothesis.
In the reinforcement at a fixed ratio, the latency, due to reinforcement, changed periodically in the sequence of acquistion. The latency immediatelly after the reinforced response was eminently longer and thereafter it gradually diminished until the next reinforcement. Thus, the series of latency made a kind of goa gradient, This ‘unit’ of responses was formed in the early period of acquisition and it made some difference in extinction in the different ratio groups.