Positives of Negatives & Negatives of Positives
Thanks to all who have commented so far in answer to the questions “Are you a 100 % positive trainer?” and “Would you sign a pledge to only use positive reinforcement and never use punishment?” I appreciate the thoughtful discussion that the questions have generated. I’ll jump in now, with the caveat that this topic deserves an all day seminar (at least) and I can’t begin to say all I’d like to in one post. I’ll start however, by summarizing some of my thoughts on the issue.
Let me start by saying that I consider myself to be an overwhelmingly “positive” trainer. I would imagine that those who have seen me work would agree that I am a kind and gentle trainer, and primarily use positive reinforcement when working with dogs. That said, I’d never sign a pledge saying I’ll only “use positive reinforcement” in all my dealings with dogs. I wouldn’t consider it for a moment. Here’s why:
1. As most of us are well aware, the word “punishment” itself refers to two completely different concepts, depending on whether you are talking to the public or to professionals in animal training or psychology. I virtually never use the word when talking to the general public or a client, unless I define it first in operant conditioning terms. That’s because the public and generic dog owner usually defines “punishment” as something inherently aversive, and often as something involving pain or fear. But in operant terms, punishment simply means anything that decreases the frequency of a behavior, and that should be well understood by any professional group involved in dog training. If someone in the dog training field asked me to sign a pledge that I’d never use “punishment,” I’d assume they were aware of its precise meaning, and decline, being unable to say “I’d never do anything to decrease the frequency of a behavior.”
2. I love “Negative Punishment.” What? Trisha loves “negative” and “punishment?” Oh my, say it isn’t so. But remember: in learning theory terms, “negative” means to take something away and “punishment” means to decrease the frequency of a behavior. Period, that’s all. Nothing about aversive, or even “good” or “bad” for that matter.
Here’s an example: Say I’ve been working with a young dog for six months on sitting when asked. We’ve proofed the behavior in a million contexts, and Fido has gotten 10 gazillion pieces of chicken and 5 gazillion other times he’s got to run and play with another dog when he did as asked. Now we’re at home, there’s little going on, but Fido didn’t sit when I gave the cue. I’m as sure as one can be without speaking ‘dog” that there’s nothing physically bothering him, he just seems more interested in ignoring me and going somewhere else to sniff. I’ll take a piece of said chicken, let him sniff it and then withdraw it, saying, melodramatically, “Too bad…. ” and possibly, depending on the state of the chicken and my stomach, “Mmmm, this sure is good! Too bad you don’t get any.”
That’s “negative punishment:” I took something away (food) to decrease a behavior (ignoring my cue). I didn’t learn this technique until I’d been in the business for quite awhile, but I use it, in the right context, and have found it to be really and truly effective. (Leslie Nelson, for her great Reliable Recalls, uses a similar technique in which one dog ends up watching other dogs get treats when he didn’t come when called.) In the example above, once I’d withdrawn the food, I ask for a sit again, and then I’d back up so that my feet don’t get squished by the dog’s hindquarters hitting the ground. That’s how effective it is. I only use this once a dog has received positive reinforcement literally over and over and over again, and when I feel confident that he understands the exercise and is capable of performing it (not feeling poorly that day, not overwhelmed by a new environment, for example). But let’s be clear: it’s punishment, if you are going to use the term correctly.
3. What’s Positive for the Trainer may not be Positive for the Dog: As the opposite of “Punishment,” “Reinforcement,”is something that increases a behavior, period. If it’s “positive” (I’ll use +R here for Positive Reinforcement) then you have added something to the system, if it’s negative (-R), you’ve taken something away. In either case, you are looking for a behavior to increase. So how would you evaluate these scenarios:
Scenario One is, regrettably, astoundingly common: A shy dog is greeted by a person, whether it’s a vet tech or a neighbor, with shrieks of joy and looming hugs and/or kisses to the nose. The person is being “positive” in their eyes, but the dog is being punished for its very existence, terrified as it is by the rude and overwhelming approach by the stranger. A perfect of example of +P to the dog and +R to the human.
In Scenario Two, a trainer is waiting for a dog to raise it’s paw so that she can use +R and give it a treat, on her way to shaping a “high five.” The dog, having no clue what the trainer wants, tries sitting, circling, and lying down. The trainer stays still and quiet, an atypical posture for her, and turns her head away very slightly. She has just used +P to communicate to the dog, adding in an unnatural posture and an obvious turn of the head (obvious to the dog anyway) to decrease the frequency of the dog’s response in that context.
4. Positive Punishment (in which something is added to decrease the frequency of a behavior) isn’t always aversive. For example, after watching herding dogs influence the behavior of sheep without touching them, I took a page from their lesson plan and began using what I called “Body Blocks.” For example, while teaching Stay, I’ll give a dog infinite quantities of treats for staying still when asked, but also move forward to block her movement if she starts to get up. “Taking the space” I’ve called it, and I’ve found it to be incredibly useful in helping dogs understand what you want. (This is similar to the Psych study one commenter noted, in which students were “trained” to perform a new behavior by either 1) only being told “Yes” when they did right, 2) only being told “No” when they did wrong or, 3) being told both “Yes” and “No.” The students who were told both what was right and what was wrong learned fastest.)
I realize that some people consider Body Blocks to be highly aversive to dogs, and don’t use them. One commenter noted that she never used Body Blocks, feeling that they were too aversive to use on her dogs. Two things come to mind here: one is that I’ve done Body Blocks on one or two dogs now (or maybe 5,000), and can tell you that they respond in a myriad of ways. Some field-bred Labradors seem to think it’s the best game in town, and try their best to beat you, eyes shining, until they figure out soon enough that something even better happens if they just stay still for a moment. They behave as though, if they could, they’d say “That was fun! Got any more cool games up your sleeve?” However, super soft dogs, let’s imagine a melty little Shetland Sheepdog, need a quiet little forward lean to be influenced, and if someone moved too fast and too abruptly they could indeed scare them. This is a perfect example of how important it is for a trainer to be able to ‘read’ a dog, no matter what method they are using I would argue, and adjust their behavior based on the personality of the dog him or herself.
5. However, this does raise the question, the elephant in the room really, of using “aversives.” Are “Aversives” always bad? Ah, here’s where the rubber hits the road, isn’t it? We can all debate about what is +P and -P and +R and -R to our heart’s content, but isn’t the issue really “Is it ever acceptable to purposefully respond to a dog’s behavior with something that they perceive as aversive?” My own answer is another reason why I’d never sign a pledge to never use punishment, even as defined by the public. Life is just too complicated to be summed up in simple categories of black and white.
Do I think that we have a responsibility to be kind and gentle to our dogs? Yes.
Do I think that Positive Reinforcement is overwhelmingly the most effective method of training? Yes.
Do I use it 99.99% of the time? Yes.
Have I ever done something to a dog that I knew he would think was aversive to get him to stop doing something? Yes. Would I again? Yes.
Here’s an example:
When Willie first started working sheep, he had a bad habit of dashing into the flock and scattering them as if he was playing pool. Alisdair McRae, a brilliant trainer and teacher, explained that I simply had to prevent it from happening during the early stages of training, because there was nothing I could do that was more reinforcing to Willie. Not only did he get to watch the sheep bolt away (look what I can do!) and then chase them (wow is this fun!), he got to disperse the tension inside of his own body (and boy do I feel better!). So I set up practice after practice in which I was between him and the sheep, and just my presence was enough for him to stay back where he should when working. But once I had to move back away from the flock to begin short outruns, he began doing it again. We went back to working in closer, but every time I backed away far enough he’d eventually dash in, scatter the sheep and turn around, body relaxed, eyes shining, mouth open, having gotten the best reinforcement he could possibly get.
I began walking him away in response: you bolt in, session over. This helped a great deal, but not enough. Eventually, after several months of work, Willie charged in, for what I believed to be the simple joy of it. I responded a gruff voice (“Cut it out!”) and a fast and direct march toward him. I stopped a long way away but looked directly at him and said again, in no uncertain terms “You cut that out!” Willie, an extremely biddable dog, backed up and looked absolutely shocked . . . and didn’t do it again. He now has the most gorgeous outrun you can imagine, and he works right on balance 99.99% of the time. Every once in a while, when he’s very tense, he’ll begin to dash in and I’ll say his name low and quiet, and he’ll curve back out again. Do I feel badly about raising my voice in that context years ago? No, not at all. Does that mean I use aversives often in training? Not at all. I quite literally never use them in any trick or “obedience” training, and primarily use +P and “Premack principle” methods to solve behavioral problems. (For example, Willie learned to lie down while working sheep because lying down on cue became the window to getting to work more.)
I’ll talk more next week about what’s critical to do or not do IF one is going to use punishment, but right now it’s time to go let Willie out to pee. No doubt relieving his bladder will be +R for him!
MEANWHLE, back on the farm: We’re pretty much at storm central here, being pummeled by ice and sleet at the same time that the politics of the area are swirling around in a social and legislative blizzard of epic proportions. (I’m right outside of Madison WI, and teach at the University, which is pretty much the eye of the storm here. FYI, for those of you out of the country, there is a huge political debate going on here, involving almost 70,000 protesters at our state capitol over the weekend. Enough said about that, except that everyone I’ve talked to agrees that the energy of the entire area is palpable, and not so much in a good way. I wonder if the dogs can sense it?)
Poor Willie injured his left foreleg again on Friday, darn. He’s been on leash restriction since then and he’s improving nicely, but not enough to let him off leash yet. If it’s not better in a day or so I’ll take him in to my sports medicine specialist vet. So Willie is bored and Sushi is disgusted–last week the warmer weather had her happily outside for hours at a time, now she’s sitting at the window slashing her tail. After I slide my way to the barn and feed the sheep we’ll do a bunch of trick training tonight a perfect time to exercise their brains instead of their bodies!
Here are some lovely clouds from a few mornings ago: