Reading M11 – Can you unlearn a behavior?


Reading M11 – Can you unlearn a behavior?

Have you ever suggested that a behavior can be unlearned? It’s an easy slip-of-the-tongue, not apparently much different from saying that we can get rid of a target behavior. And if we’re successful, isn’t that the same as unlearning the behavior? Can we say that the individual has learned to not engage in that behavior – effectively unlearning it?

Can a behavioral history go backwards?

We know we can arrange contingencies that make a particular behavior very unlikely to occur, but we misunderstand the nature of operant learning if we imply that a behavior can actually be unlearned. As a particular behavior becomes well established in a person’s repertoire, it accumulates a reinforcement history. This history cannot be erased, no matter what future contingencies may unfold regarding the behavior. The behavior may at some point no longer generate any environmental support, for example, which means it would eventually no longer occur. It may even result in punishing consequences, with the same eventual outcome. Its function could be more effectively served by a different behavior, thereby reducing the chances of it showing up again.

In any event, we know that just because a certain behavior has not occurred for a very long time does not mean that it cannot reappear. Once well established in the repertoire, it always has the potential to show up uninvited. Depending on the complexity of the behavior and other factors, it might only reoccur in a weak or deficient form, but the fact that it was once established means it can always reappear, at least in principle.

This possibility is certainly not random. Behaviors occur for all of the reasons that make up operant learning processes. For instance, a behavior not seen in some time might reoccur because other behaviors that better served the same function and replaced it in the repertoire are for some reason no longer effective. It is easy to see this in shaping a new behavior. Shaping involves gradually shifting reinforcement contingencies to increase support for better approximations while reducing support for approximations that were previously reinforced. If the trainer isn’t careful and fails to provide adequate support for a new approximation, it is common for a previous approximation to reappear instead. Although this earlier approximation was in some sense dormant in the repertoire, the right combination of operant processes can call it forth.

Careful how we talk

If we seem to be successful in getting a target behavior to no longer occur, we should be careful how we explain this change. We certainly don’t want to say that the person has learned to not engage in this behavior, for instance. The problem with this familiar everyday description is that it implies that what has been learned is “not doing something.” Of course, “not doing something” is an awkward way of expressing the fact that a particular behavior is no longer occurring, even on particular occasions, because “not doing something” is not a behavior (see Box 4.5 in Johnston, Pennypacker, & Green, 2020 for more discussion of this point). What is happening is that other behaviors are occurring instead. When a behavior seems to disappear from a repertoire, there is no hole left by its absence. A repertoire is always full. The time previously spent engaging in that behavior is now filled by other behaviors.

In sum, once a particular behavior is learned, it is possible that it can recur. It cannot be unlearned because we cannot revise any part of a behavioral history. We can only build new history, which can reduce the likelihood that a behavior will occur again, but we shouldn’t offer guarantees.


Johnston, J. M., Pennypacker, H.S., & Green, G. (2020). Strategies and tactics of behavioral research and practice, Fourth edition. New York: Routledge.