I love operant conditioning and the science of behavior change. I love the jargon, the details, the studies. What researchers find in their studies really impacts what I think about in behavior change.
That’s why I rely primarily on one quadrant of operant conditioning: positive reinforcement. We’ll get into that more in a second.
There’s a lot of ideology running rampant in the dog training world because of the competing ideas that:
1) We should use positive reinforcement based training only.
2) We should use all tools available to us to get the training results that we desire.
This can lead to a lot of confusing behavior advice, which is really frustrating for owners (and trainers).
Dog trainers draw lines in the sand and scream at each other. They accuse the other side of causing behavior problems that lead to euthanasia of shelter animals. Yes, it escalates that far that quickly. I’m not disagreeing with that.
As a shelter worker, I often see dogs where I think that the training methodology was a factor into that dog’s behavioral concerns. But training is just one factor, and remember that this study showed that punishment-based training was still more effective than no training at all.
But here’s the thing.
We can’t eliminate environmental punishment from our dog’s lives. No matter what I do, Barley might be punished for trying to steal hot food from the stove by burning himself. That’s punishment, and it happens every day to our animals. Just last week, Barley was punished for going swimming by freezing his furry @ss off for the rest of our 28* run.
We also can’t eliminate tiny punishers from our own interactions with our dogs. Yes, I’m a primarily positive reinforcement trainer. (The fashionable term for that today is “relationship based,” though at times I’ve also called myself “force free” or “R+ based.”) But that doesn’t mean that I don’t ever punish my dog intentionally in training. Let’s get into the nitty gritty of training science!
THE QUADRANTS OF OPERANT CONDITIONING
Hang onto your panties, here comes some science.
Behavior is based off of its consequences. Reward a behavior, and it will increase. Punish a behavior, and it will decrease. For now, we’re going to skip over classical conditioning (read on for that).
There are just four basic ways of modifying behavior. These are known as the four quadrants of operant conditioning. There are two ways to change a behavior (punishment or reinforcement) There are likewise two types of things to add to your environment (add or remove). Multiply two times two, and you get the four quadrants.
To get our terms straight, remember that positive just means something is added. We’ll use a + sign to make that easier. Negative means removing something from the environment. We’ll use the – sign to help you keep that straight. As a reminder, punishment decreases behavior and reinforcement increases a behavior.
+ Punishment: this means adding something unpleasant to the animal’s environment to decrease a behavior.
Example: a collar gives your dog some electrical stimulation (shock) when he tries to steal food from your plate. He’s less likely to try to steal food again in the future.
+ Reinforcement: something pleasant is added to the animal’s environment to increase a behavior.
Example: you toss a treat to your dog to reward him for staying on his bed while you’re eating. He’s more likely to stay on his bed next time you’re eating.
– Punishment: something pleasant is removed from the animal’s environment to decrease a behavior.
Example: you remove yourself (and your food) from the room when your dog tries to steal food from you. He’s less likely to try to steal food from you in the future.
– Reinforcement: something unpleasant is removed from the animal’s environment to increase a behavior.
Example: you turn off an ultrasonic trainer only when your dog is in his bed. He’s more likely to stay in his bed in the future.
Many trainers largely focus on the + half of the quadrants, because it’s much easier to add something to your dog’s day in a systematic way than it is to remove something. Owners use a variety collars that are meant to stop behaviors – these an all corrections are +punishment. Owners also use a variety of treats, praise, petting, and toys to reward good behavior. These methods are all +reinforcement.
Note: if the punisher or reinforcer that you’re using is not changing behavior, it’s not actually a punisher or reinforcer. For example, I often see clients yelling “NO” at their dog when he jumps – yet he jumps on people every time I see them. Is the “NO” actually changing his behavior? It’s not. So it’s not a true punisher.
The – half of the quadrants is trickier. You see lots of -reinforcement in bridle and bit training for horses and some leash training for dogs. Give into the pressure, and the pressure goes away. Next time, you’re more likely to turn your head and body when you feel that pressure.
I also use -reinforcement when training very fearful animals by rewarding them for touching a stick by leaving. My presence is so scary that I can reward them for doing what I want by leaving! -Punishment is also tricky. I often use this as a way to work with dogs that bark incessantly to get their owner’s attention (read more about fixing this problem here). We punish their barking by removing people from the room if they bark. The barking should decrease.
So, you see, I frequently use 3 of the 4 quadrants of operant conditioning at Journey Dog Training.
CLASSICAL CONDITIONING: HOW DOES THAT FIT IN?
I promised we’d circle back to classical conditioning, because this gets confusing. Buckle in, we’re getting on the science train again.
Classical versus Operant Conditioning
In operant conditioning, the animal is the “operator.” His behavior dictates what happens next. Do A, and you get B. Do C, and you get D. Don’t believe me? Double check the examples of all four quadrants above. The punisher or reinforcer only changes after the animal does something.
Classical conditioning relies on pairing one object with another. This is not contingent on the animal’s behavior. Instead of relying on the animal to do something that we can either reward or punish, we’re relying on pairing one environmental condition with another.
Let’s talk in concrete examples, that’s easier.
1) A bell rings and food appears. Your dog starts to salivate at the sound of the bell. Sound familiar, Pavlov? The bell used to be meaningless. Now it’s a cause for excitement! This is the basic idea of clicker training: click = treat. Mind blown?
2) A dog sees another dog and meatballs rain from the sky. Meatballs rain from the sky even if Fido is barking, lunging, and screaming at the other dog. He learns that seeing other dogs is pretty cool. He’s being classically conditioned to like the sight of other dogs. It’s pretty cool that we can use this to change how your dog feels about another dog!
This process takes longer than with the bell because your dog likely already has some complicated feelings about seeing another dog. The bell is a neutral stimulus, so it’s no biggie to pair it with food and make it awesome. If your dog already gets a surge of adrenaline and negative emotion when he sees another dog, we’ve got more work to do to make him happy to see that dog!
This is how we start out working with reactive dogs at Journey Dog Training. Instead of punishing your dog for barking and lunging, we teach him that seeing another dog makes awesome stuff happen. Over time, because we’re using classical conditioning, he isn’t so freaked out when he sees another dog. Then we can start using positive reinforcement to teach him what to do when he sees another dog. We’re changing his emotional state first, behavior second.
3) Every time a snake crosses your path, a collar zaps your dog. This happens regardless of if he’s walking calmly or chasing the snake He learns that seeing snakes is really, really bad. He’s being classically conditioned to be afraid of the snake. This is a common way of teaching dogs to avoid snakes.
I don’t like Option 2, but I’d entertain the idea of it if avoiding snakes was a very important skill for a dog to learn very quickly. I’d rather avoid using this technique because it runs the risk of teaching my dog that hiking = shocks, butterflies = shocks, or my scream (SNAKE!) = shocks. I don’t want to accidentally teach my dog to be scared of anything.
On the other hand, it’s ok if my dog accidentally learns that white picket fences = meatballs, my voice (Oh, hi Fluffy!) = meatballs, or walks = meatballs.
So those are the five basic training tools available to us. Nature uses them all. We’ve got four types of operant conditioning, plus classical conditioning.
REMINDER: THIS IS INDIVIDUAL
The crazy thing is, not all animals love and hate the same things. You’ve got to get to know the animal you’re working with before embarking on your training odyssey. Let’s explore a few common examples where things get messy.
Maya, my 13.5 year old lab, will sell her soul for a hearty scratch on the butt. When I first adopted Barley, I tried to dig my fingers into his butt the way Maya likes. He literally ducked away from me. Clearly, petting him this way was not going to work as a training reward. I work with many dogs at the shelter that are so distrustful of me that trying to pet them would likely result in stitches in my hand.
Don’t assume all dogs love petting. For many dogs, petting is at best a weak reinforcer. They might agree to sit in your kitchen in exchange for a pat, but you’re going to have to pay better if you want them to listen in a more challenging environment. At worst, petting can actually be a punisher! I can actually get Barley to leave me alone by petting him on the butt in the exact same way that Maya loved. It’s how I get him to bugger off when I’m writing.
Let’s use Maya and Barley as demo dogs again. Maya is a hearty, happy Labrador. If you yell at her, she wags her tail and looks at you. It’s a game, right? Yelling “NO” at her did nothing as a puppy. We either had to use stronger punishment or teach her what to do instead of jumping. Yelling no simply didn’t work for her, because she didn’t find it scary. In order for yelling to work, it’s got to be sufficiently unpleasant.
Barley is a whole other creature. If Andrew and I get too animated in reimagining our days at work, he crawls under the bed. If he puts his paws on the counter, a firm, “Barley, OFF” sends him off the hide under the sofa. I don’t even have to yell with him. That’s a sensitive border collie for you.
Many, many dogs do not respond to yelling as a punisher. You’re going to have to punish them more severely if you want them to stop the unwanted behavior. This can lead to lots of frustration and increasingly painful or scary training techniques. On the flip side, many other dogs find yelling absolutely petrifying. There are dogs at the shelter that likely would pee all over themselves in fear if I raised my voice at them. Not exactly the best way to build a relationship, eh?
Barley would jump into the Grand Canyon if you threw a tennis ball over the edge. I have no doubt of that. He’s a total toy maniac. He’s actually so obsessed with toys that for the first six months of owning him, I couldn’t even use toys to train him. As soon as I pulled out a toy, his eyes totally glazed over and it was like he couldn’t even hear me. Now that he’s learned that the if he listens to my cues, he gets the toy, we’re off to the races. Toys are his kryptonite.
Compare that to Duncan, a white American bulldog I worked with recently at the shelter. Duncan is a young, rambunctious dude. He scrambles around the shelter wagging and play bowing at people. He looks like he needs a squeaky toy. But when I first tossed a toy to him, he ran away from it and hid. Toys were not a reward for him. In fact, the squeaky toys were downright scary. I work with countless other dogs that look at you when you tempt them with a toy. Their eyes look like those of a teenager who’s being offered to wear a birthday hat at her sweet 16. “Really? You’re going to insult me with that?”
To make matters more complex, don’t forget that not all situations are created equal.
Will I pick up my friend from the airport for $5? Sure! Will I drive her from Denver to Connecticut for $5? No. That’s laughable. The same goes for your dog.
While your dog might be happy to sit in exchange for a Cheerio at home, he’s likely to totally ignore you if you try to get him to sit when you’re at the dog park.
The bottom line is that you can’t assume what your dog likes or doesn’t like.
For Barley, toys are a reward, yelling is a punisher, and petting is neutral to slightly punishing. Meanwhile, toys are a punishment to Duncan. For Maya, yelling is neutral while petting is a reward. Keep in mind this does not mean that certain breeds require harsh punishment to train. All dogs can work with all four quadrants.
If you’re struggling with teaching your dog to do something or stop doing something, reevaluate your reinforcers and punishers. You might be surprised by what you find.
I USE ALL FOUR QUADRANTS, SO WHY IS +REINFORCEMENT MY FIRST CHOICE?
+Reinforcement is my go-to.
I rely heavily on +reinforcement (adding good stuff to make behaviors happen again). This is where my treats and toys come into play. I’ll also use petting and happy talk with dogs that are into that sort of thing. This is my tool of choice. Reward the dog with good things, and two awesome things happen:
1) He’s more likely to do that thing that I liked again, because it made something that he liked happen.
2) He’s likely to be really excited to see me next time because I’m The Bringer of Good Things. Isn’t that how I want my dogs to think of me? Not some pack leader or Big Brother type figure!
-Reinforcement is tricky, but usable.
That said, I also use some -reinforcement. I don’t use much -reinforcement (removing bad stuff to make behaviors happen again), but it comes into play with two areas of my training.
1) As stated above, I often use -reinforcement with very scared animals. I teach a terrified cat that if he touches his nose to the eraser of my pencil, I’ll go away. Next time I offer my pencil, he’s more likely to press the button that makes the scary human go away. This is great for animals that are so scared that they won’t eat or engage with any other +reinforcement that I offer.
2) I also used -reinforcement to teach Barley “right” and “left.” I’ll be honest: I don’t carry treats or toys with me on my runs with Barley. Shoot me. Instead, I yell “right!” and then start turning my body to the right. He feels some pressure from the leash on his harness. That pressure is relieved when he turns. His behavior of turning to the right is rewarded by the removal of the unpleasant tugging from the leash.
I don’t rely heavily on -reinforcement because it means that first I must add something negative. I find it’s usually much faster to reward my dog than remove something bad. It’s a bit trickier to remove something than add something, so that’s that.
-Punishment has its uses, too.
I use some -punishment, too. Removing good things from my dog’s life is a valid way to train him. As yet another example of why knowing your individual animal matters, consider that I often punish dogs for barking by leaving the room. The same action by me (leaving the room) is a punisher for one group and a reinforcer for another.
The problem with -punishment is that this can be very frustrating for your learner. At Journey Dog Training, I rarely use -punishment alone. Instead, I pair it with +reinforcement. Take the barking dog for example. When she starts to bark, I leave the room. She stops barking, so I reappear. We then engage in some training or play and teach her something to do to get my attention instead of barking. She quickly learns that if she sits, she gets what she wants. If she barks, I leave the room and the fun stops.
+Punishment is usually my very last resort.
I do not want to do anything that harms my relationship with the animals. That’s why I’ve never recommended that a client purchase or use a choke chain, prong collar, e collar, citronella collar, bark collar, electric fence, or any other form of correction. The potential for fallout is too big.
That said, I do use do some +punishment. As in the example above, I’ll sometimes pet Barley roughly on the butt as a way to get him to leave me alone. I also will use Bitter Apple spray to stop dogs from chewing on furniture. But beyond those two examples, I have a hard time thinking of +punishment in my training sessions.
This is in line with the humane hierarchy. This is a road map of training options from least aversive or unpleasant to most. I adhere to it in my training, and there’s a lot more in there than just the four quadrants. Oh, yes! Training also encompasses antecedent arrangement and much, much more.
If nature is a balanced trainer, and we’ve got all these tools at our disposal, why don’t I use them all equally?
Let’s think about a toolbox. I’ve got a screwdriver, a saws-all, a sledgehammer, and a chainsaw. If I wanted to remove a door from my home, I could use any of these four options. If I were a skilled chainsaw artist, I might even be able to remove the door without harming the rest of my house. But let’s be real, that’s not me. I’m going to be much more successful at keeping my house safe if I use the plain old screwdriver to remove my door’s hinges. Does that make me a bad carpenter? No! It just means that I chose the technique with the least potential for harm.
The same goes for dog training. As we saw at the beginning of the article, I can use any of the four operant conditioning quadrants to teach my dog not to steal food from my plate.
I’m most likely to choose to reward my dog in order to get what I want for six reasons:
1) +Reinforcement shows him what I want him to do. Life is easier when we know what we’re supposed to do in a given situation. If I were navigating New York City, I’d much rather be told exactly where to go and what to do than be told every single route and attraction to avoid. I’d rather if my dog knows that when he sees another dog, he should look at me. It’s easier to work on that than to teach him every single possible response to another dog that is undesirable.
2) +Reinforcement actively builds our relationship. +Reinforcement is a great way to bond. In the food-stealing example, Barley learns that he can still rely on me for food and fun things – if he does what I ask. His trust in me and desire to be near me can only increase if I reward him for good behavior.
3) Punishment has major potential fallout. One of the scariest things about punishment heavy training is the potential for fallout. You might be able to teach a dog to stop acting aggressively towards dogs by giving quick and controlled corrections via a prong collar. But you also might teach him he was right – dogs are really bad, and we have to work really hard to keep them from coming near us. We’ve just taught the dog to be more aggressive because we just used classical conditioning (dog=pain) instead of operant conditioning (bark=pain). The trouble is, it’s very challenging to control which lesson the dog learns.
4) Punishment creates negative emotions. As a best case scenario in the example above, we’ve suppressed the behavior. Remember, punishment can only reduce or stop unwanted behaviors or create negative emotions. It cannot improve emotions or increase a desired behavior.
Doling out corrections via the prong collar might reduce the unwanted behavior of barking, lunging, and growling at other dogs. However, the prong collar cannot make negative emotions go away. The calm dogs that you see are not acting calmly because their owner has calm, assertive energy. Instead, they’ve learned that looking calm makes punishment stay away.
I’d much rather teach my dog that Scary Things = meatballs (classical conditioning). When my dog is now calmer around the Scary Things, I can turn back to +reinforcement and teach him that when he sees Scary Thing, he should look at me for a meatball. Cool!
5) +Reinforcement is really hard to overdo. You can really ruin a dog with +punishment. Just read about Emma Parson’s dog Ben, whose aggression towards other dogs got much worse when a trainer hung him from a prong collar around another dog. It took years of training for Emma to teach Ben to be calm around other dogs again. There are horrible stories of dogs being harmed or even killed by trainers using punishment.
Yes, +reinforcement can have some negative side effects. You can make your dog fat if you’re not careful about proper exercise and portion sizes. If used incorrectly, you can reward your dog for bad behavior. I see this all the time when people try to push dogs off of them while the dog happily jumps at them. What they fail to realize is that pushing and yelling at the dog is actually a reward for the dog, so the dog is more likely to jump on them again! Oops!
6) It’s easy. I started with +reinforcement as a dog walker because it’s so darn easy to bring some treats with me on a walk. Since I hated the idea of accidentally yanking too hard on a prong collar, I started teaching my dog walking clients to use treats to walk their dogs calmly.
I quickly found that dogs pranced along next to me, happily looking up for a cookie when they sat at crosswalks. Wow, that’s pretty cool! That’s how I got started as a dog trainer. Using treats to train dogs was fun for me as I saw my relationships with these dogs blossom.
At the end of the day, training dogs is fun.
Since I’m not a mother dog or mother nature, I largely stick to the operant conditioning quadrant that’s least likely to have bad side effects and most likely to build my relationship with the animal I’m working with. It’s safest, and I’ve found it’s very effective. My clients agree!
At the same time, I don’t stick my head in the sand. There are times where all living animals run into punishment in their lives. There are times where I use punishment to control behavior. Focusing on the behaviors I want and the neatest tools I have seems like the smartest way to go.