Positive reinforcement-based trainers never use positive punishment, right? At least we certainly try not to. But it can sneak into our training all the same.
Punishment, in learning theory, means that a behavior decreases after the addition or removal of a stimulus. In positive punishment (the addition case), the stimulus is undesirable in some way. It gets added after the dog’s behavior, and that behavior decreases in the future. Some examples of that kind of stimulus would be kicking the dog, jerking its collar, shocking it, or startling it with a loud noise. You can see why positive reinforcement-based trainers seek not to use positive punishment.
In contrast, in negative punishment, the stimulus involved is desirable (appetitive). It gets taken away after the dog’s behavior, and that behavior decreases in the future. Examples of negative punishment are pulling the treat away from the dog’s mouth if she lunges for it, and leaving the room if a puppy plays too roughly. (Here are more examples of the processes of operant learning.)
In positive reinforcement-based training, we try to use only negative punishment. But even negative punishment can be unfair sometimes, as I explain in this post. Not only that, but it’s possible to slide straight into positive punishment inadvertently from negative punishment.
Positive Punishment: A Note About the Definition
Just because something hurts doesn’t mean that it will punish behavior. It is possible to administer an unpleasant stimulus (repeatedly!) and have no behavior change. For instance, I give allergy shots to both my dogs once a week. They get a whole CC of fluid injected under the skin on the back of their necks. I can tell it doesn’t feel great. But from the very beginning, I have followed the shot with a little box of fabulous treats, different every week. I’ve tried to determine whether the shot acts as a punisher. I’ve watched for decreases in behavior that might result from the shot. I’ve found no such decreases. The dogs come eagerly for their shots and take the position I ask and stay still. The shot event is happy overall, even though there is some brief pain involved.
So, keep in mind the “second half” of the definition of punishment. A behavior must decrease. It’s not only that you did something icky to the dog. It had to have an effect on behavior over time. Positive punishment can actually be difficult to employ successfully. The unpleasant stimulus must be applied at the right magnitude, with good timing, and consistently.
Even with these caveats, I have seen accidental positive punishment happen several ways.
Examples of Accidental Positive Punishment
Side effects of “leave it.” Many trainers begin the training of “leave it” (a.k.a. “it’s your choice” or ” doggie Zen”) by holding a treat in their hand. Some start with the hand open; some start with the hand closed and work up to it being open. When the dog moves forward to take the treat, they close their hand. The goal of closing the hand is negative punishment. When the dog moves toward it, the treat (appetitive stimulus) disappears and becomes unavailable. If the training mechanics are good, lunging for the treat will decrease over time. But there is a danger of positive punishment here. If the dog is fast, then the trainer has to close her hand fast. (Most trainers recommend against pulling the hand away.) Suddenly closing your hand on a dog’s muzzle can be startling or unpleasant for the dog. If the behavior of lunging subsequently decreases, what happened? You may have used positive punishment rather than negative punishment.
What’s the Fallout?
The examples I gave above don’t involve scaring, hitting, or kicking the dog. They don’t sound as bad as that. A hand snapping shut, a collar grab, or a leash jerk. Not so terrible, right? Can even these milder sounding aversive stimuli create fallout? Oh, yes. If you snap your hand shut on a puppy’s snout, or right next to it, you can cause the puppy to be wary of hands. A very unfortunate lesson for a pup. Likewise with collar grabs: if you do them without conditioning first, you will create a dog who dodges away from humans. And while some dogs habituate to leash jerks, your next dog might be the one who shuts down from the jerk you create by moving backward. My pressure-sensitive dog got positively punished when I charged up at her to “help” change a prop setup.
Of course, it’s not the theoretical change from “minus” to “plus” that creates a problem for the dog. It’s that when we set out to follow a training plan, we often fail to notice the dog’s response to different parts of it. We don’t see the dog saying, “Hey, you pinched my nose! I hate that!” We are probably concentrating on our own mechanics. So I could have written these cautions without any reference to learning theory, and just said, “Watch the dog!”. But then they would just be scattered incidents. Using learning theory helps me see the pattern so I can head off future problems.
Some people claim to train without the use of aversives. That’s a goal of mine, as well, but unless we are vigilant, they can sneak in anyway. Just wait until I write a similar post about negative reinforcement. Evil grin.
Have you ever used positive punishment by accident? I promise I won’t let anyone hassle you if you want to comment. These examples are super useful for all of us to be aware of.
Copyright 2018 Eileen Anderson
|↑1||A third option is to call the dog, but most trainers don’t want to call the dog to a negative consequence.|