When I enter my house, I’m always greeted by Tiramisu, a small black dog with big brown eyes that shine up at me. “How’s my baby girl?” I usually ask in a squeaky play voice and she jumps up to lick my face, her tail wagging a mile a minute. I pet and fuss over her and then she runs to get a toy to show me and she is just so obviously delighted to see me. Similarly, I can look across the room at Tiramisu and say “Hey sweet girl!” and her ears prick up and her tail s
wishes from side to side. There are no treats or clickers involved here but my dog clearly enjoys the verbal interaction I give her. But other dog owners struggle to even get their dog’s attention after repeatedly calling their names or making all kinds of noises. Why is my dog different than that?
The answer is both complex and simple. It is simple in that Pavlov’s laws of Classical Conditioning are always operating and always influence any relationship between me and my dog. Yet the answer is complex because the associations my dog will make as a result of Classical Conditioning — about me and the words and actions I use — can be a complicated. They will appear even more complicated if I don’t realize or direct how these associations are happening.
Let’s Start at the Science
Pavlov’s famous experiment was actually a happy accident. In 1927, he was researching the digestive physiology of dogs when he noticed that his subject dogs would salivate in the presence of meat powder presented for food. After a few feedings, Pavlov then discovered that the dogs would salivate in the presence of the lab assistant who would normally bring the meat powder even if no meat powder was present. Pavlov then went on to repeat the experiment using a bell to prove that even a sound could be conditioned as a stimulus to get the dogs to salivate. And so we have the story today of Pavlov’s dogs and the bell.
Pavlov went on to document what we now call, in psychology, the science of Classical Conditioning. The process to create a meaningful association between something intrinsically relevant to an animal and something that previously had no meaning, such as a sound or other signal. Why is this important to us as dog trainers?
To start with, it is the principles of Classical Conditioning that make Mark and Reward training (i.e., Clicker Training) work. Through repeated presentation of a marker signal and a reward, the dog begins to associate the marker as a reliable predictor of a reward coming soon. The marker soon becomes a “good news” signal that a treat is on the way.
Behavioural trainers call this a secondary or “conditioned” reinforcer. The marker, although of lower value than a primary reinforcer like food, takes on reinforcing properties just like the presence of the lab assistant or the bell in Pavlov’s experiments became associated with the meat powder. And this leads to some very interesting possibilities!
Conditioning All Over the Place
Most dog owners I know have an endearing habit. They frequently say “Good Boy!” or “Good Girl!” or “Good Dog!” when delivering a treat. And why not? We usually give our dogs treats when we are happy with their behaviour. There is usually a big smile that comes with this ritual as well. So is it any wonder that most dogs wag their tail when they hear “Good Dog”?
We have “conditioned” them using Pavlov’s Classical Conditioning to have a positive response to our words of praise. Most trainers, from the time they first begin training their dogs, habitually accompany the delivery of a treat with some words of praise or physical affection. And our dogs respond more and more to those verbal or physical signals over time whether or not we actually deliver the treat – just like the salivation in Pavlov’s dogs.
But there is another side to Pavlov’s discoveries. We cannot control when and how our dogs make these associations. We can set up situations to influence the associations our dogs make but there are no guarantees. Our dogs will decide if something is “good for dogs” or “bad for dogs” by what happens around that predictable signal. If you wonder why your dog looks “guilty” sometimes, consider Pavlov.
Practical Conditioning — Cues and Commands
All dog trainers have been using Classical Conditioning for a long time. We used to call them “commands” but today many of us prefer to use the term “cue” instead. When I raise my hand or say “Sit”, my dog will put her butt on the floor. She does this in response to Classical Conditioning.
“How does that work?” you may be asking yourself.
Well, during the training process your dog somehow determined an association between that signal and getting some kind of reward for sitting. In the beginning the training is clumsy because the association has not been formed yet. We say “Sit” and the dog stares at us blankly. After several repetitions an association begins to form: “When mom makes that sound (“Sit”), if I put my butt on the floor I get something good.”
So the Cue or Command is a signal that the opportunity for something good (or the opportunity to avoid something bad) is presenting itself if you do the right thing after seeing or hearing the signal. It’s rather like the green light in the rat’s cage in the psychology experiment that indicated that the food lever would now operate.
It’s not hard to see why my little dog perks up as soon as I give her a cue. It’s a clear signal for her that she has an opportunity to earn a reward. And what’s even more interesting is that every new cue I give her becomes reinforcing because she knows from experience that these behaviours will lead to a reward. The Premack Principle is based on this concept: the opportunity to do one behaviour can be reinforcing for a different behaviour.
Even More Conditioning
So there’s lots of conditioning going on all the time, if we believe Pavlov. The presence of the toy is a signal for fun, the leash is a signal for an outdoor adventure, food bowls signal dinner, the bright green jacket that means it’s off to dog class, and even the hand in the pocket can be a signal for some dogs. With all this conditioning and association going on, how are we supposed to know which things our dogs value most?
Well, only our dogs can tell us that, but a general rule is that “primary” reinforcers like food will always have the highest value since our dogs need to eat to survive. Although even food is not a sure thing if your dog is well and frequently fed. Food is not such a big deal since it shows up all the time. But generally food is the tops, and the better the treat, the more value it has (e.g., roast beef beats kibble most of the time).
Just like Pavlov, we can condition other reinforcements like verbal praise or toys by pairing them with high-value reinforcement like food. It’s such a useful practice that most dog owners do it without realizing that’s what they are doing. If we say “Good dog!” and give the dog treat, we’re conditioning the praise as a reinforcer. If we throw the toy for a fun game when our dog brings it back to us, we are conditioning the toy (even the dropping of the toy!) as a reinforcer.
What if we could harness the power of this conditioned reinforcement to use to train our dogs? Good news! Most dog owners already do without realizing it. I’m sure you may have recognized some of the scenarios I’ve presented here whether or not you know anything about classical conditioning or mark and reward training.
Myth of Slot Machines
This is where it gets interesting, at least for me as kind of a “training geek.” It has been said that the most powerful training results come from what is called “Intermittent Reinforcement.” That is, you don’t pay the dog EVERY time for a behaviour. In fact, you reduce the frequency of payment over time.
Trainers like to call this a “Slot Machine” analogy. The same random chance of a pay-off that keeps a gambler pumping quarters into a slot machine will keep a dog performing for you without any payment. But is this really true? With all of this Classical Conditioning going on, are we really NOT paying the dog after behaviours? In order to be truly “Intermittent” in our payment system we would have to provide NOTHING at all as the alternative to deliberate payment.
But that’s not how we usually train, is it? There’s always a “Good dog!” in there, or a toy or a quick bit of physical affection. So what we are really doing is varying the value of the reinforcement rather than withholding reinforcement altogether.
This brings us to an important point.
Rewarding is Easy if You Do it Right
What’s remarkable to me about all of this “conditioned reinforcement” stuff is that it satisfies most trainers across the methodology spectrum. Frequently I hear opponents of Mark and Reward training or Positive Training complain that the dog won’t work without food present. This isn’t necessarily true if we condition verbal praise or some other alternative as a reward. In fact, most traditional trainers do this kind of associative conditioning automatically whether they realize it or not.
Positive trainers who are concerned that they won’t always have their marker or treats with them can use conditioned reinforcement in place of those tools and continue to train as they always have. A verbal “good” becomes the marker, and a toy or pat on the head becomes the conditioned reinforcer. Suddenly our training options become MUCH broader!
The whole concept of value and reward can be somewhat confusing. Not all dogs respond to the same things. Similarly, not all dogs have the same conditioning history with all things. For example, a dog who has been frustrated in play with a handler won’t find toys very reinforcing. But this can be manageable. With some careful observations of your dog’s preferences and by carefully creating associations with new signals or environments, we can find Classical Conditioning to be a powerful ally.
Pavlov and his dogs didn’t just make an interesting discovery in the field of psychology, they made our lives as dog trainers SO much easier. All we have to do is make use of the information he has given us.
Until next time, have fun with your dog!
Photo credits
Tiramisu – Petra Wingate 2009
Licky – TB2011 from Flickr 2010
Play – Christina Spicuzza from Flickr 2010
Happy – H. Michael Karshis from Flickr 2010
Recent Eric Brad CPDT-KA Articles:
- The Dog That Changed My Life
- Keeping Your Dog Informed
- Teaching Your Dog to Not-Behave
- The Simple Complexity of Dog Training
- What The Dog Wants
One of my previous dogs responded with enthusiasm to the phrase Are you a Bad dog?
For him had the same connotation as the Good Dog phrase
All in the delivery and context.
I realize I am a little late to this party, but I just finished up my certification in a mostly positive based training program. I am toying with the idea of coming up with a slightly different set of “conditioned reinforcers” because of the very tendency and temptation to overuse the “good boy.” Do you think this is even an issue? I mean, have you ever had issues training using the CR “good” because “good boy” is such a commonly (and not always appropriately) used thing to say to dogs?
Thanks for your comments and question.
I must confess, I’m not exactly clear what you are suggesting but I’ll take a stab at it. A verbal marker such as “Good Boy!” can become a conditioned reinforcer over time if consistently paired with a primary reinforcer enough times. It reliably predicts the presence or delivery of a “good thing” and that can form a positive emotional response to the previously neutral sounds “Good Boy!”
That said, I think it’s important to understand the nature and role of conditioned reinforcers in any training regime. I do NOT use conditioned reinforcement to teach new behaviours. In fact, I use them mostly for maintenance of learned behaviours in those situations where it is not practical or possible to use primary reinforcement (i.e., food). There are a variety of situations where it is beneficial to use conditioned reinforcers and even to help your dog understand that high value primary reinforcers may not always be available. The trick is to understand the training economy and not let performance suffer due to improperly conditioned (weak) reinforcers or a radical shift in the dog’s expectations (i.e., “variety” of reinforcement should be trained so that the dog is not surprised/disappointed when the reward protocol changes).
The only issue I can see with over use of a conditioned reinforcer such as “Good Boy!” is that, over time, if the pairing with primary reinforcement stops, the strength of the conditioned reinforcement might diminish. When this happens, the dog’s performance begins to diminish and, if nothing is done to shore up the primary reinforcement paring with “Good Boy!”, it may come to signal NO REWARD to the dog instead of invoking the positive emotional response it previously held.
The trick with conditioned reinforcement is that the conditioning, the pairing with primary reinforcement, must be maintained at some level in order for it to remain effective.
I hope that gives you some answer to your question. Please feel free to post again or contact me directly if I didn’t address what you were asking.
Thanks for reading,
Eric
Thanks for all of your nice comments. I’m glad people are finding my thoughts useful. They are just things that rattle around my head and I hope people can take something from these articles to use with their own dogs.
Thanks for reading and thanks for the encouragement. If there are any topics you would like me to address, feel free to post them here or send me private email.
With appreciation,
Eric
Insightful article, Eric. and easy to understand. I have seen the results you describe frequently with my own dogs, and at the shelter I volunteer with.
They are amazing creatures, thanks for helping us better understand each other.
annette
Great article, Eric.
You do a good job here explaining some pretty complicated topics.
Mary
Awesome article!!!! Engaging and useful as always!
Hi Eric,
Thanks for the dog training lesson. How interesting it is that it’s also a lesson in psychology. Sometimes I think I overuse “Good Dog” with my Pounce. Hmm.
-Ashley