Feature Article

Extinction:

What It Is & Why It’s Important

 

Last month on Melissa Alexander’s popular discussion group, Clicker Solutions, there was an invigorating discussion about extinction.  Of course we’re talking behavioral extinction, not the kind that means an entire species exists no more.  It was mentioned during the lengthy and vigorous discussion that there is at least one popular trainer who is teaching that there is no such thing as extinction.  Um, really?

Extinction is real, but the confusion is understandable.  There’s a punishment process that seems quite similar until you dissect things a bit.  It’s important to make the distinction between these processes, so for clarity’s sake I’m going to go back to the fundamentals and start from there.  Follow me. 

In order to understand extinction, it’s important to understand what reinforcement is, so that’s where I’ll start.   It might also help to review the archived article, Positive Punishment and Other Oxymorons at: http://www.behaviorlogic.com/id110.html .

 

Reinforcement basics

Reinforcement is a process through which a behavior is followed by some event and the future rate of that behavior increases or is maintained.  The process is only reinforcement if the rate of the behavior increases or stays put because that event followed it.  That event is called a reinforcer – if it works.  The event is only a reinforcer if it follows the behavior and the rate of the behavior increases or is maintained.  The reinforcing event might be food, beverage, a chance to escape an unpleasant situation, some social interaction, or a variety of other things. 

The reinforcing event must happen contingent upon that behavior, too.  By contingent, I mean that the reinforcer must happen if and only if the behavior of interest happens.  The reinforcer and the behavior are related to each other. 

In clicker training the most common reinforcer is small portions of food.  An  animal performs a behavior and the trainer clicks the clicker and produces a treat, which results in the animal doing more of that behavior in the future. 

 

How extinction relates to reinforcement

            Although extinction refers to a reduction in the rate of a behavior, it can only happen when reinforcement has already been holding that behavior in place.  Extinction is sometimes described as just ignoring a behavior until it goes away.  That’s not exactly it, although it can seem that way. 

Extinction is a process through which the rate of a behavior decreases because a reinforcer that has been delivered in connection with that behavior is no longer delivered.  In other words, a behavior is reinforced for some period of time and its rate increases or stays stable, but then the reinforcer stops being delivered when the behavior happens so the rate of the behavior diminishes.  In the extinction process the behavior declines because the reinforcer is no longer delivered when the behavior happens.  The reinforcer and the behavior no longer have that contingent relationship.  They’ve broken up.  Jennifer no longer shows up where ever Brad is happening.  Things have changed. And the behavior decreases as a result.

 

Where the confusion comes in

            So, that’s what extinction is.  But there’s another process that sounds a lot like extinction.  Negative punishment sounds, at first blush, so much like extinction that some people have decided that extinction is really the same thing.  But they dismiss it too soon. 

Negative punishment is a process through which an event is removed from the environment contingent upon a behavior and the future rate of that behavior decreases.  Yeah, that sounds very much like extinction which is the process through which a reinforcer is no longer delivered and the rate of the behavior decreases.  In both situations something isn’t there any more and in both situations the rate of a behavior decreases.  But there are some important distinctions.

In both cases an operant behavior is happening.  That means the behavior is happening because the reinforcer is connected to it.  (Operant is a term B.F. Skinner came up with that indicates that a behavior operates on the environment in some way.  In these cases the behavior operates on the environment to produce some reinforcer.)  So these behaviors are going along being reinforced when the change happens.

 

Extinction

            The behavior is operating on the environment to produce some reinforcer.  But then the reinforcer stops being delivered when the behavior happens and the behavior goes away.  The reinforcer was not readily available in the environment until the behavior happened, so the behavior operated on the environment to produce the reinforcer.  Then the behavior stopped having the effect of producing a reinforcer, so the behavior decreased.  It is important to note that the thing that is stopped in extinction is the very thing that originally kept the behavior happening—the reinforcer. 

 

• When Fido first came to live with Janie, every time he stood on his back legs she gave him a treat, so Fido did this in Janie’s presence very often.  After a few weeks Janie ran out of treats and didn’t bother to get any more.  For a while Fido continued to stand on his back legs around her, but gradually he did it less and less often.  His standing behavior was extinguished when the treats were no longer delivered for it. 
• The treats made him stand on his back legs, and when the treats went away he gradually no longer bothered to stand on his back legs. 

Negative Punishment

            With negative punishment the behavior is operating on the environment to produce some reinforcer to start with, just like in extinction.  But then something else that was readily available in the environment is taken from the environment when the behavior happens.  The reinforcer continues to be available if the behavior is performed, but the loss of this other thing causes the behavior to happen less often in the future.   It’s not the discontinuation of the reinforcer that causes the change as happens in extinction, it’s the withdrawal of something else. 

 

• Every time Jimmy pulled his sister Sally’s hair, she squealed, so he pulled her hair a lot.  Their mother got very tired of this, so one day when Jimmy pulled Sally’s hair, she took his Game Cube system and locked it in the trunk of the car.  Jimmy eventually got his Game Cube back, and he pulled Sally’s hair less often (at least in the presence of Mom) so that he wouldn’t lose it again. 
• Unlike in extinction, the reinforcer for pulling Sally’s hair was still available… she would have still squealed when he pulled her hair.  But something worthwhile to Jimmy was taken away when he pulled her hair.  Even though the Game Cube had nothing to do with him pulling her hair in the first place, its loss was sufficient to cause him to pull Sally’s hair less often in the future.     

Why it matters

            The behavioral processes, negative and positive reinforcement, negative and positive punishment, and extinction happen in nature.  They are natural processes just like those that make up physics and chemistry and other sciences.  They don’t happen merely because trainers “do” them to learners.  If a bear goes to a particular berry patch for days and days and going there is reinforced with berries to eat, but then the berries run out and he gradually stops looking for berries there, that’s extinction.  And it’s also extinction if Bruce stops petting his dog, Milo, every time he jumps on him and Milo gradually stops jumping on him.  It’s a natural process trainers can learn to employ to their advantage, just as pharmacists employ chemical laws in making medicines.

            The same is true with negative punishment.  If a baby fox runs off catching bugs but realizes that when he turns around his mom is no where in sight, he may go bug catching less often in the future, at least until he’s a bit older.  His Mom, who has been readily available all along is gone because he followed a bug.  The reinforcer for bug catching… either the fun of the chase or the eating of the bug, is still available, but his bug chasing is negatively punished by the withdrawal of Mom’s presence.  Likewise,  the human Mom can take away Jimmy’s Game Cube to get him to stop pulling Sally’s hair, and assuming that works, it’s negative punishment. 

            The important matter for trainers is that we have to know what process is in effect before we can effectively change it in order to efficiently change behavior. 

By eliminating extinction from our training lexicon we eliminate a possible explanation for behavior that falls apart, and we eliminate an important part of an extremely valuable behavior building tool, differential reinforcement. 

 

Shaping and differential reinforcement require extinction

Differential reinforcement makes up the component parts of shaping, which is at the heart of clicker training.  In differential reinforcement, you reinforce the new approximation toward the target behavior while letting the previous approximation go on extinction.  Shaping is the differential reinforcement of successive approximations toward a target behavior.  And what is extinction?  It’s the process through which a behavior’s rate is decreased when a reinforcer is no longer delivered with it.  It would be a real challenge to try to find negative punishers that would effectively help you build toward new behaviors through shaping.  With extinction, you already know what to do.  You know your learner’s behavior rate increases when you use a certain treat or other reinforcer, and knowing that you know that if you stop delivering that reinforcer the rate of that behavior is going to diminish.  So you carefully change the criterion for getting that reinforcer.  The magic is in the reinforcer… and in the trainer’s skill in the shaping dance.   

Differential reinforcement is also the tool of choice for diminishing unwanted behaviors.  A common technique is to reinforce the heck out of a behavior you prefer, and extinguish the behavior you don’t want.  It’s particularly effective if you use the very same reinforcer that was maintaining the problem behavior to reinforce the preferred behavior.  So if Fido jumps on family members when they get home in order to get attention and petting, the family can stop providing attention for jumping and only provide attention when Fido sits at their feet.  If they are consistent, Fido will sit at their feet far more often than he jumps on them as the training process progresses. 

One of the beauties of differential reinforcement… the combination of reinforcement for one behavior and extinction for another … is that you often can reduce or prevent the problematic extinction bursts that can happen if you use extinction alone.  By providing a desirable way for your learner to earn the reinforcer he has already proven he will work for, you can often eliminate or dramatically reduce extinction bursting.  However, during shaping you still get enough of a burst that you get some variable behavior giving you more approximations to work with.

 

The bottom line is that extinction is a real process, and one that you will benefit from understanding … and believing in! 

 

Kellie Snider

Copyright 2005