It has been argued that Thorndike’s experiments are not typical of the learning process, since the cat cannot «see the connections» between the movement of the valve and the «exit of the box». But in operant conditioning, fixing connections is not essential. Both during and after conditioning, a person often talks about the relationship of his behavior to his environment. His report may turn out to be suitable for scientific explanation, and his reaction to his own behavior may even prove to be an important link in certain complex processes. But in simple processes of operant conditioning, such reports or reactions are not required. This is because the person may not be able to describe the circumstances of the reinforcement that had a clear effect. There is also no need for any permanent links between the response and its reinforcement.
We make food intake dependent on the reaction of the pigeon by organizing mechanical and electrical connections. Outside the laboratory, the situations in which the connections between behavior and its effect arise are determined by various physical systems. But this does not affect the body in any way, and there is no need for this. As far as the organism itself is concerned, the only important property of the reinforcement contingent is its temporary character. Reinforcement follows reaction. How this happens is irrelevant.
We must assume that the presentation of a reinforcer always reinforces something, since it inevitably coincides with some behavior. We know that just one reinforcement can have a significant effect. If only a random connection is established between a response and the appearance of a stimulus, the behavior is said to be «superstitious.» We can show this with the example of the dove by accumulating the effect of several random circumstances. Let’s say we feed a pigeon a small amount of food every fifteen seconds, no matter what it does. When the food is presented for the first time, he performs some behavioral responses — unless he stands still — and conditioning takes place. It is then more likely that the same behavior will be observed again when food is presented. If this happens, then the operant will be further strengthened. If it doesn’t, then some other behavior will be enforced. Usually this behavior reaches the frequency with which it is reinforced. It then becomes a permanent part of the bird’s repertoire, even if the food is presented at a time that is not related to the behavior of the bird. Visible responses that have been established in this way include jerky head tilts, stepping, stooping and shuffling, pivoting, unnatural gait, and head tilts. Later on, the topography of behavior may shift to other reinforcers, as subtle changes in response form may coincide with food intake.
When producing «superstitious» behavior, the intervals at which food is served are important. At sixty seconds, the reinforcement effect is largely lost before another appears, and another behavior is more likely to occur. Therefore, «superstitious» behavior is less likely to occur, although it may occur if the experiment is run over a long period of time. After fifteen seconds, the effect is usually almost instantaneous. Once a «superstitious» response is established, it will persist even if it is reinforced occasionally.
The dove is no exception. Human behavior is also largely «superstitious». Only a small part of the behavior, reinforced by unforeseen circumstances, becomes a ritual that we call «superstition», but the principle itself remains in force. Suppose you find ten dollars while walking in the park (and suppose that this event has a significant reinforcing effect on you). What we were doing when we found the money is backed up. Of course, it is difficult to give strict evidence, but it is very likely that we will again go for a walk in the same park or in a park similar to this one, and it is more likely that we will look at the exact place where the money was found, etc. e. This behavior will vary depending on the state of deprivation to which money is relevant. It should not be called «superstitious», but it is generated by a situation that rarely turns out to be «functional».
Some situations that produce «superstitious» behavior are not completely random. Sometimes a reaction is followed by an effect, which nevertheless does not «produce» it. This can best be illustrated by examples of the type of stimulus that reinforces after it has been eliminated. The cessation of short-term stimulation of this kind can reinforce the behavior generated by its onset. When an unpleasant stimulus appears, the body enters a state of activity; the stimulus stops, and this reinforces a certain part of the behavior. Certain diseases, lameness, allergies are such that any measures taken to cure them are likely to be reinforced when the condition improves. It is not necessary that the measure is actually related to the treatment. The rituals used in non-scientific, folk medicine can probably be explained in terms of this characteristic of many forms of disease.
In «superstitious» operant behavior, as in conditioned «superstitious» reflexes, the conditioning process fails. Conditioning gives the organism great advantages by equipping it with the means of behavior that allows it to operate in new conditions, but there is no way to prevent the learning of inefficient methods of random behavior. Curiously, this difficulty seems to have increased as conditioning progressed through evolution. If, for example, it always took three reinforcers to change the probability of a response, «superstitious» behavior would be unbelievable. Only because the organism has reached a level at which already one contingent brings significant changes in behavior, they can coincide.
In human society, «superstitious» rituals involve verbal formulas and are passed down as part of the culture. In this sense, they differ from the simple effect of operant reinforcement. But they probably originate from the same process, and they must be supported by random conditions that have the same pattern.