If you’re even vaguely familiar with positive reinforcement training, at some point you have received the advice to ignore unwanted behavior.
As with the advice to redirect to a preferred behavior, this advice is rooted in good science, but has become so divorced from the understanding of the underlying principle that it sometimes does more harm than good.
The process people are trying to use when they ignore behavior is called extinction. Extinction is nonreinforcement of a previously reinforced behavior. It’s not the withdrawal of a reinforcer that’s already been offered, and it’s not something taken away contingent on the performance of the unwanted behavior. Whatever was reinforcing the behavior is simply is no longer available for doing that behavior.
Reinforcement is what makes a behavior stronger, and removing reinforcement, permanently, will eventually make it weaker. But it’s not a straightforward or tidy process—and beyond that, ignoring is only nonreinforcement if your attention was the reinforcer your dog was behaving to get. Your attention is not generally the reason your dog starts pulling on the leash, barking at strangers, or peeing in the house (though through accidental training, it can certainly come to be).
And even if your attention is the reinforcer, you need to be aware of what will happen when you stop giving it.
Extinction is best used in conjunction with reinforcement of another behavior—ideally one that serves the same function as the unwanted behavior, and definitely one that will provide at least the same quantity and quality of reinforcement.
That’s because the problems with using extinction by itself are many. The ones I’ll discuss here aren’t even all of them. Extinction is different from punishment (where something is added to or removed from the learner’s environment, contingent on the behavior, to decrease the behavior), but that doesn’t mean it’s more pleasant for either the learner or the teacher. In fact, it can have some of the same side effects as punishment, including the emergence of frustration-related behaviors and aggression.
Before a behavior starts to decline due to nonreinforcement, it will predictably flare up in what’s called an extinction burst. If you and your dog live in a condo or an apartment, or you have a headache or are tired, or you have a guest with whom you are trying to have a nice meal, an extinction burst of barking will be very hard to ignore. You’re more likely to end up reinforcing it—thus teaching your dog to bark more or more intensely.
This Family Guy clip has probably been shown at least once at every dog training conference I’ve ever attended, and it is a rather perfect example of reinforcing an extinction burst:
Extinction isn’t fast, either. A single session “is often not enough to extinguish behavior . . . even when the extinction session lasts for several hours and involves hundreds or even thousands of unreinforced acts,” writes Paul Chance in the textbook Learning and Behavior. And once it does go away, it can come back. In the phenomenon known as spontaneous recovery, Chance continues, “what usually happens is this: The rate of the previously reinforced behavior declines and finally stabilizes at or near its pretraining level. Extinction appears to be complete. If, however, the animal or person is later put back into the training situation, the extinguished behavior occurs again, almost as though it had not been on extinction.” Chance wraps up his chapter on extinction by noting that “there’s considerable doubt, in fact, about whether a well-established behavior can ever be truly extinguished.”
Another extinction-related phenomenon is called resurgence. When one behavior is no longer reinforced, other, previously reinforced behaviors tend to emerge. In other words, when the behavior doesn't work, the dog tries other behaviors that have worked before.
Many dog owners have seen this in a scenario like the following: You’ve stood up and collected your training or walking gear, but get temporarily distracted or fumble with the equipment. Or you’re in class, with your dog in front of you, but the teacher is talking. Your dog first offers an expectant sit—a behavior that probably has been reinforced a lot by you in this context. But you’re listening to the teacher, and so there’s no reinforcement. Your dog then offers a paw, goes into a down, rolls over, then barks and paws at you.
And when you do finally turn your attention to him, because pawing your leg hurts, what have you just reinforced?
But if we stay focused on our learner, we can actually take advantage of resurgence.
One of my favorite brief training articles to send clients is this item by Dr. Caryn Self-Sullivan, a KPA CTP in Virginia: Stop, Watch, Wait, Reward. It’s just a quick hit, so she doesn’t go into extinction or resurgence or any other technical business, but I think it’s revelatory to read it with the science in mind:
For example, if your dog jumps or barks when you enter your home:
STOP: Stand perfectly still and be absolutely silent.
WATCH: Observe your dog out of the corner of your eye and watch for a behavior you want to reinforce.
WAIT: Wait, wait, wait for a desired behavior, such as a sit or even just eye contact with four-on-the-floor.
REWARD: Mark (click) the desired behavior, and then toss a treat. Proceed into the house.
I love this. I also love thinking about what we can do, before we actually find ourselves in this pickle, to make it likely that when we do, the dog will offer a previously learned behavior we’d like to see more of, instead of another one we don't like.
One simple thing we can do is to make sure our dogs have a big, fat repertoire of other, frequently and recently reinforced behaviors to call upon. We know that animals, when they have a choice, tend to allocate their behavior in direct proportion to how much reinforcement those behaviors have received in the past. Teaching simple but acceptable behaviors and reinforcing them regularly in a variety of contexts will make it more likely that the dog will go to one of these when another behavior isn’t working. Orienting to the handler, sitting, lying down, and settling on a mat or bed are behaviors that a dog can offer in many situations to the delight of their humans.
Here's a quick video of Horton—a dog with a history of barking long and loudly at humans who are holding food without giving him any. (He's also pictured at the top of this post.) Trying to ignore him failed, and may have contributed to the development of this persistent behavior in the first place. But here he is choosing to lie down (without being "redirected" or even cued) instead because he has recently had a ton of reinforcement for doing so.
To help give your dog lots of good alternatives, pay attention to all the acceptable behavior you don’t explicitly teach or ask for. Many times behaviors we like are happening right in front of our noses, while we’re absorbed in something else. Meanwhile, behaviors we don’t like rarely fail to get our attention. The dog walks on a loose lead near us for five steps—a completely unreinforcing activity by many dogs’ standards—then pulls ahead, which is when we call him back for a treat. What behavior is he probably going to do more of? If instead we notice and give a treat for those five steps, we will get more like them. When the dog pulls ahead, we don’t go along—but should the dog check in of his own accord, we have something else we can reinforce.
We can also sometimes rig the environment to make the right choice more likely and the wrong choice less so. Take the jumping dog in Caryn’s example above. We don't even have to wait for him to jump and then wait for other behaviors. Placing a gate between the dog and the door could prevent both jumping and reinforcement of jumping by family members or visitors who pet the dog when he jumps. It's also easier to watch for preferred behaviors and reinforce them when you're not getting jumped on.
After all, humans respond to the laws of behavior just like dogs. You can make the right behavior easier for yourself, too.
This post was originally written for One Tail at a Time. It has since been revised and updated.