In this column, I’ve already addressed a few well-intentioned training chestnuts that can really go south if you don’t understand the underlying behavioral principles. This month, I’ll talk about another one: the advice to “redirect” unwanted behavior. E.g., when your dog jumps, or puts his paws up on the counter, or barks at you while you’re watching TV, you direct him to a more appropriate behavior, then reinforce that behavior.
If your dog is already jumping on some poor unsuspecting grandpa right this very minute: by all means redirect him. But know that if you make a habit of this, your dog will probably make a habit of jumping, at least a little. Here’s why:
Cues are signals that predict reinforcement for a certain behavior. If you have trained your dog (as I hope you have) that sitting on cue will produce a treat, petting, play, or something else he wants—then through Pavlovian conditioning, your sit cue has acquired some of the value of that treat, petting, or play, becoming what’s known as a conditioned reinforcer. Money is a powerful conditioned reinforcer for humans—we work for scraps of paper or numbers on a screen because they are permanently paired with food, shelter, and other stuff we need and want.
Cues trained with positive reinforcement can be used to reinforce other behaviors. Which is fantastic news if, say, you’re trying to maintain all the behaviors in an agility sequence, where food or play can only come at the end.
But it also means that if the dog throws his paws up on the counter, and you give him a cue that means sitting will be reinforced, you will reinforce putting paws on the counter. And behavior that is reinforced is repeated.
Don’t leap from here, as so many do, to the conclusion that we must therefore punish unwanted behavior. Unfortunately things just aren’t that black-and-white. Punishment, especially (but not only) physical punishment, carries the risk of side effects including fear, aggression, apathy, and escape-avoidance behavior. The same Pavlovian conditioning that creates such good associations with cues trained with positive reinforcement can create bad ones with people, places, animals, and cues that may predict something aversive. (Read about “poisoned cues” here.)
The fact is, if the dog is already doing the unwanted behavior, you’re starting out behind the eight ball. All your choices are less than ideal. You may be obligated for safety reasons to reinforce the behavior (e.g., dragging the dog away from whatever he’s using aggression to get distance from). If there’s no immediate danger, you may be able to simply make reinforcement unavailable (e.g., by continuing to scroll through Facebook as your dog barks in your face) or remove it (e.g., stepping away from the dog when he jumps to greet). With either of these, though, you risk frustration and escalation to more intense and emotional behavior. If your dog has a rich repertoire of other well-reinforced behaviors that you like, he may try one of those next—be sure to notice and reinforce it.
Or you can redirect. Yes, you might reinforce the unwanted behavior. Is it the worst thing in the world? No, and sometimes it’s the best of your crappy choices. Probably anyone who lives with a dog will settle for it in some situations. I’ll confess right here that I redirect my dog from barking pretty much every night—when my husband is cooking dinner, and I walk down the stairs into the kitchen, she play bows and barks. I then cue her to her mat, and when she lies down (which for her tends to be incompatible with barking), I pull a fish-skin chew out of the treat basket and hand it to her. Think she’ll bark again tomorrow? I do.
If you do need to redirect, do it as quickly as possible, so that you only reinforce a little bit of the unwanted behavior. Then make sure the dog is not set up to immediately repeat the cycle—after all, setting up rapid repetitions and reinforcing each one is exactly what you do when you want to teach a behavior. Prevention of repetition might involve leashing, gating, or crating the dog, or just giving him something more appealing to do (say, chewing on a fish skin).
Now, if it’s important to you to reduce this behavior in the future, your job is to figure out how to predict it, so that in the future you can preempt rather than react. If you want to teach a new response to the problem cue, you need to identify the problem cue.
Ask yourself: What are the conditions that set the stage for the behavior? What time is it? Who’s present? What are they doing? From how far away? What seems to be the most immediate cue for the unwanted behavior? Can you prevent any of these conditions from even happening in the first place? (If the blinds are closed, the dog can’t bark at stuff he sees out the window. No “training” required.)
Next, what do you want? If the cue does happen, what would you like the dog to do instead? (This answer shouldn’t involve the word “not,” as in “not bark.”) Pick something easy. When the cue for the unwanted behavior occurs, that’s the best time to ask for (or just catch) a desired behavior. With repetition and reinforcement, the problem cue can begin to evoke the new behavior instead of the old one.
Finally, what does the dog get out of the unwanted behavior? That’s the “reason” he does it again and again. Can you provide that same consequence for a more acceptable behavior? Can you provide something the dog likes as well or better? If the new behavior produces more reinforcement, that’s what the dog will choose first.
This systematic way of looking at the behavior in the context of its environment is called functional assessment. It lets us figure out when and why a behavior is probably happening, and gives us the specific information we need to get ahead of it—instead of waiting for it to happen and then trying to figure out what to do about it.
Now you’re ready to train. You may need to strengthen, or even teach from scratch, the behavior you want the dog to do instead. Best to do that in low distraction first, then introduce it into the problem scenario. And you may need to introduce it into the problem scenario in small, achievable increments.
But sometimes it’s much simpler than that. If your dog routinely jumps to greet, and you routinely cue him to sit before petting him, can you instead reach down and pet your dog while his feet are still on the ground, or invite him up on a bench or chair so he can get closer to your face without jumping? If your dog puts his paws up on the counter to get tidbits while you’re cooking, can you put his bed nearby, ask him to lie on it before you start food prep, and toss him bits of what he wants straight from the chopping block?
As for me, I could direct my dog to the mat while I’m still on the stairs, before she starts barking, and have a chew with me to toss her as soon as she goes to it. As with most plans to change a dog’s behavior, this will require some behavior change on my part, namely moving her mat slightly, stocking fish skins in my nightstand, and remembering to grab one before I head downstairs. I’ll let you know how it goes.