Yoked Control Procedure

The yoked control procedure is a research design used in operant conditioning experiments in which matched research subjects are yoked (joined together) by receiving the same reinforcement but with different contingencies. Operant conditioning is said to have occurred when the frequency of a class of behavior is changed by its consequences. Reinforcement occurs when the rate of the behavior increases; punishment occurs when the rate of the behavior decreases. These behavioral changes represent the operation of the law of effect, first formulated by Edward L. Thorndike.

Consider the most celebrated case of operant conditioning. Burrhus F. Skinner placed a hungry rat into a small box containing a lever and a food cup. During Phase 1, the rat's rate of lever pressing was quite low; this unconditioned rate of responding is called the “operant level.” During Phase 2, Skinner arranged a contingency in which each lever press delivered a small pellet of food into the cup; now, the rat's rate of lever pressing regularly rose—so-called acquisition via reinforcement. During Phase 3, Skinner removed that contingency by discontinuing feedings after presses; now, the rat's rate of lever pressing progressively fell toward operant level—so-called extinction via nonreinforcement.

The lever press −food contingency might or might not be responsible for the increase in responding in Phase 2. Another plausible possibility lurks. Perhaps intermittent feeding made the hungry rat more active; this upsurge in activity might have increased the rat's lever pressing and many other behaviors that Skinner did not choose to measure.

One way to disentangle this interpretive confounding is to deploy the yoked control procedure. Imagine two conditioning boxes instead of only one, with a different rat in each. The research project would unfold for the experimental rat just as described. But, the yoked control rat would be treated differently in Phase 2; there, the yoked control rat would experience the same number and temporal pattern of feedings as the experimental rat, but there would be no contingency between its lever presses and pellet deliveries. Only the experimental rat could produce food by lever pressing in Phase 2; the yoked control rat could not. Hence, it would be expected that the experimental animal would respond more than the yoked control animal in Phase 2 if the lever press −food contingency were contributing to the experimental animal's observed rate of responding.

Pay special attention to the possibility that—even though there is no programmed contingency between lever pressing and food delivery in Phase 2—the yoked control rat might sometimes experience a chance conjunction between pressing the lever and food presentation. Such chance conjunctions might be sufficient to increase the incidence of lever pressing, what Skinner termed “superstitious” conditioning. Such superstitious conditioning would tend to minimize the disparity between the behavior of the experimental rat and the yoked control rat.

Also pay attention to the possibility that individual animals might differ in their reactivity to the presentation of food—whether delivered contingently or noncontingently upon behavior. Such possible disparities have been thought by some theorists to compromise the utility of the yoked control procedure. Although there are reasons for concern, the expected effects of such disparities might be small compared with the many merits of the procedure. Furthermore, yoking need not only be accomplished between subjects; yoking can also be accomplished within subjects.

...