Extinction is a familiar concept to most animal trainers. A behavior that previously produced reinforcement no longer does, and over time the behavior decreases. On the surface, that sounds fairly straightforward. But anyone who has observed extinction in practice knows that what happens along the way can be anything but simple.

Behavior may initially increase. Animals may persist, vary their responses, return to behaviors that worked in the past, or demonstrate behavior we might describe as frustration or aggression. And even after a behavior has decreased substantially, it isn't necessarily gone forever.

For me, this raises an important question. If we know extinction can produce predictable fallout, how often do we really need to rely on it to solve behavior problems?

Increasingly, I find myself interested in what we can arrange instead.

What Actually Happens During Extinction?

Extinction occurs when a behavior that previously produced a reinforcing consequence no longer produces that consequence. The contingency has changed. What worked before no longer works now.

An important detail is that extinction doesn't erase behavior from an animal's repertoire. A response that has decreased substantially can reappear when conditions change, something we describe as spontaneous recovery. We can also see resurgence, in which previously reinforced responses reappear when a more recently reinforced behavior no longer produces its expected outcome.

Reinforcement history matters as well. A behavior with an extensive reinforcement history may be quite persistent, and behaviors maintained on intermittent schedules can be especially resistant to extinction. From the animal's perspective, continuing to respond may make perfect sense. Sometimes persistence has paid off in the past.

This is one reason I think it is useful to move away from thinking about extinction simply as “ignoring” an unwanted behavior. What matters is the contingency. What consequence did the behavior previously produce, and what happens when that consequence is no longer available?

Once we frame the question that way, the animal's response to extinction becomes much easier to understand.

Extinction Can Have Predictable Fallout

One of the best-known effects associated with extinction is the extinction burst. A behavior may temporarily increase in frequency, duration, or intensity when it stops producing the outcome that previously reinforced it.

We may also see greater variability in responding. If one behavior no longer works, trying something different can be quite reasonable. Previously reinforced behaviors may return. Emotional behavior may also occur.

This becomes particularly important when the behavior we're trying to reduce is already problematic. If aggression has historically resulted in increased distance, for example, arranging conditions in which aggression no longer produces distance may result in more intense or persistent responding. From a procedural standpoint, that may fit the definition of extinction. From an animal welfare and safety standpoint, however, we should be asking whether creating that situation was necessary in the first place.

That doesn't make extinction inherently “bad.” Extinction is a behavioral process, and it occurs naturally as contingencies change. It is also involved in many training procedures. But understanding its possible effects should influence how deliberately we choose to use it.

Extinction Is Part of Shaping, Too

There is an interesting complication when we talk about avoiding extinction because extinction can also occur during shaping.

Suppose I am shaping an animal to touch a target. Initially, I reinforce orienting toward the target. As the animal becomes successful, I begin reinforcing a small movement toward it instead. Orienting alone no longer produces reinforcement.

Technically, the previous response has been placed on extinction.

But good shaping doesn't necessarily produce the dramatic fallout we often associate with extinction. If approximations are small, criteria are clear, reinforcement is frequent, and the environment makes the next response easy to emit, the animal can move through the process with very little disruption.

That distinction is important.

The question isn't simply whether extinction is technically present in a procedure. We can also ask how much behavior has to fail before the animal discovers what works next.

Careful shaping can minimize that failure.

What Can We Arrange Instead?

This is where constructional thinking has significantly influenced how I approach behavior challenges.

A traditional reduction-focused question might be:

How do I stop this behavior?

A constructional question is more likely to be:

What repertoire would I like the animal to have instead, and what conditions will help build it?

That shift can lead us toward very different training plans.

Rather than deliberately withholding the consequence maintaining an unwanted behavior and waiting for that behavior to decrease, we can often arrange conditions in which desirable responses are easy to emit and effective at producing meaningful outcomes.

This requires us to understand function. What is the animal currently accomplishing through the behavior? What consequence is maintaining it? Can that consequence be made available through other, less costly responses?

It also encourages us to examine antecedent conditions. Sometimes changing the environment makes the behavior we want much easier while making the behavior we're concerned about less likely to occur in the first place.

The goal isn't merely to suppress an unwanted response. It is to build a repertoire that works better for the animal under those conditions.

Rethinking Our Reliance on Extinction

There will always be changes in contingencies, and extinction is an important behavioral process for animal trainers to understand. But understanding a process doesn't mean we need to arrange it deliberately every time we want behavior to change.

When I encounter a behavior problem, I am increasingly interested in asking what can be built rather than what needs to be eliminated.

Can we change antecedent conditions? Can we reinforce smaller approximations toward a useful alternative? Can we provide more than one response that produces the critical consequence? Can we arrange the environment so that desirable behavior is easier and more effective?

These questions don't guarantee that extinction will never occur. But they can greatly reduce our reliance on procedures in which an animal must repeatedly experience that something that previously worked no longer does.

And that is one of the things I find most valuable about a constructional approach. It moves our attention away from simply making behavior stop and toward building repertoires that give animals effective ways to navigate the situations they encounter.

Sometimes the best way to address unwanted behavior isn't to become better at extinguishing it.

It is to become better at building something useful in its place.