Extinction has two meanings in behavior analysis. It first refers to the procedure of withholding or eliminating reinforcement. A technical aside is that it also can refer to the procedure of, instead of removing the reinforcement, simply eliminating the dependency between a response and the reinforcer that normally would follow it. Thus, the reinforcer still occurs, but it does not depend on the response. This second procedure has found rather widespread use in applied behavior analysis under the unclear label of “noncontingent reinforcement” (the reinforcement is actually contingent on the passage of time). The present commentary will focus only on extinction as reinforcement elimination.

The second meaning of extinction is the actual elimination of behavior that is observed when reinforcement is eliminated. Thus, behavior that is no longer reinforced and no longer occurs is said to be extinguished. Sometimes the two definitions are used in the same breath, which can be confusing to people unfamiliar with the relation between the two.

Extinction procedures can remain in place for short or long periods of time, depending on the circumstances. The duration often depends on the behavior being extinguished. If, for example, problem behavior is eliminated, an ironic situation can develop: there is no need for the extinction procedure because the target behavior no longer occurs, but because reinforcement is no longer occurring, the extinction procedure in a sense still remains in effect. Extinction also can occur in concert with other, reinforcement, conditions. For example, if Dad never reinforces whining to get to do something and Mom usually does, then whining extinguishes (is eliminated) when Dad is around and whining continues to occur when Mom is present. In such a case, a discrimination is said to be formed based on the reinforcement or absence of reinforcement of whining. 

Extinction also forms the core of two frequently used procedures for reducing or eliminating problem behavior: the differential reinforcement of other behavior, or DRO, and timeout from reinforcement. Both of these procedures involve the implementation of relatively short periods of extinction (periods of nonreinforcement of the target response In the case of DRO, the periods of extinction typically are not associated with a stimulus change, that is, they are unsignaled, and in the case of timeout the extinction periods typically are signaled by some stimulus change (such as the therapist turning around to ignore the problem behavior). With both DRO and timeout, the extinction periods are response-dependent, that is extinction does not occur unless the target response occurs. When it does, a short period of extinction immediately follows. With both DRO and timeout, reinforcement of responses other than the targeted one(s) continues to occur.   

With both DRO and timeout, extinction is said to be response-dependent in that the periods of nonreinforcement are initiated by a response. In fact, the same is true of the procedure simply labeled “extinction.” Instances of the target response are not reinforced, but other responses continue to be eligible for reinforcement if other reinforcers are available for such responses. In the laboratory, because typically only a single response is measured, extinction usually is taken to mean that all behavior in that environment undergoes extinction. This often seems to be the case in that laboratory animals experiencing the discontinuation of a previously available food reinforcer following some response such as lever pressing often simply become immobile and fall asleep during the extinction session. It is possible, however, that other responses such as preening or grooming can continue to be reinforced and thus maintain to the extent that reinforcers for such behavior remains available. In nonlaboratory settings with humans, there are multiple sources of reinforcement for many different classes of behavior, so, unsurprisingly, extinction becomes rather pinpointed to the target response(s).

This is the first of three commentaries on extinction. In the next two, we will explore some of the technical and ethical implications of using extinction to eliminate behavior.