What Predicts What: A Key Question in Dog Training
By Kiki Yablon, KPA CTP, CPDT-KA
Dog training is not unlike cooking. Whatever kind of food you’re preparing, there is a science to how ingredients respond to heat, cold, pressure, acidity, each other, and so forth. You can follow a recipe and get the same end result as the chef who wrote it—but if something goes wrong, suddenly you’re lost. To be a truly masterful cook, you need to understand the scientific princples so that you can apply them in many situations and even on the fly.
A frequent debate in professional dog training circles is how much of the science, if any, to lay on the average dog owner. On one end of the spectrum, you have those who argue owners just want their problem solved—they don’t care why the dog is doing what they’re doing, or why the solution works. On the other are those who believe that no behavioral solution will hold up if the person responsible for maintaining it doesn’t have any concept of the underlying principles.
I fall toward the latter end. This might be because as a learner, I don’t like to be told what to do without being told why. But it’s also because I’ve learned that behavior is always a response to environmental conditions. In fact, in the literature, individual behaviors are often referred to as “responses.” You can’t have a response without something to respond to.
The standard definition of learning is a change in behavior due to experience—experience being interaction with the environment. And what constitutes your dog’s environment? You. You are a huge part of your dog’s environment. He learns from every interaction with you and with the world you give him access to. You might want to exert some informed influence over that.
How to get started? Well, there are two basic types of learning, and they can both be boiled down to what predicts what:
- What items, events, or conditions in my environment predict good things? What predicts scary things?
- What in my environment predicts that a given behavior will produce good things or scary things?
You’ll be familiar with first type, called respondent learning, if you’ve ever seen so much as a gag cartoon about Ivan Pavlov. Pavlov taught dogs that certain signals (though probably never the bell with which he’s so closely assocated) predicted the appearance of food. When the dogs learned the association between the two, the signal that had heralded the arrival of food would elicit the same responses as the food—drooling, for instance.
This kind of learning is often invoked when we hope to change someone’s emotions—if a bell routinely predicts delicious food, pretty soon, in addition to salivating, we’re happy to hear the bell. Likewise, if a tone routinely predicts a shock, the tone will soon produce the same fear response as the shock. The behaviors elicited during this kind of learning are reflexive, involuntary, unlearned—like drooling. What the dog learns is a new trigger for them.
The second kind of learning, known as operant learning, is what we usually think of when we’re training dogs. If the first kind of learning is “If bell, then food,” and then “if bell, then drool,” the second kind is “When bell, if sit, then food.” This “when-if-then” unit is called a contingency. Here the bell is a signal that says “Hey, this would be an excellent time to sit, because usually if you sit when you hear the bell, you get delicious food.” What the dog learns is that the cue is an opportunity to act and produce a desired outcome.
The two kinds of learning are often discussed as separate processes, but in reality they’re always happening at the same time. The bell that signals the dog to sit may also elicit drool. Dr. Susan Friedman, psychology professor emeritus at Utah State, refers to the relationship as a “Gordian knot” in her behavior science course for animal professionals. And Dr. Jesus Rosales-Ruiz, a behavior analysis prof at University of North Texas, said in a seminar for trainers earlier this month that changing the contingencies changes the emotions. In other words, situations where the animal can behave to get something it wants produce happy emotions, and situations where the animal must behave to avoid what it doesn’t want produce unhappy ones.
I still can’t decide if this is utterly mind-blowing or a total duh.
What can dog owners take away from this? So many things, but here are a few to start:
If you want reliable behaviors, your cues need to be reliable predictors for the dog. I’ve written about this here before.
Be mindful of your dog’s emotional state when you’re training. I taught my dog to fetch a couple years ago, at the ripe old age of 8 or 9. She had absolutely no interest, so it was on me to make it worth her while. I shaped the behavior up from scratch using a recipe (hey, they’re helpful sometimes) adapted from Hannah Branigan’s Obedience Fundamentals. One thing she said that stuck with me was that the emotions the dog is experiencing while you train become associated with the behavior. I wanted an enthusiastic behavior, so I only worked on this behavior when my dog was most excited to see me, which was when I got home from work.
On the flip side, if your dog is communicating through body language that he is stressed or afraid or just plain tired, it might not be the best time to start teaching a behavior that you want to be exuberant.
Use positive reinforcement when working with fear. The goals of sound behavior modification plans for fear (and its cousin, aggression) are generally to change the dog’s emotional response (respondent learning) and replace the unwanted behavior with an acceptable one (operant learning). Any new behaviors you’re going to use should be trained with positive reinforcement—that is, they should have a clean history of earning things the dog really likes. If the behaviors and their consequences are more delightful than, say, a man with a hat is frightening, you may create positive associations with the man with the hat.
But if you ask the dog for behaviors that are prompted with an implied “or else,” you may be creating negative associations with the cues—even if you also reward correct responses. And then when you cue those behaviors in the presence of the man with the hat, you may be associating the man in the hat with the threat of punishment. That’s counterproductive at best.
When dogs learn what predicts a good thing, they do good things to get the good thing. I copped this phrasing from behavior blogger Eileen Anderson, who gives a great example in this post about how she taught her puppy to come to her whenever her “reactive” older dog barked. Spoiler: she simply offered the puppy cheese whenever the older dog barked. Her goal was to create a positive emotional response to the older dog’s barking, but the puppy started coming to find her whenever the older dog barked.
No training “recipe” can tell you exactly how to apply these lessons to an individual dog. Training is an exchange between the teacher and the learner, with each adjusting to the other. But understanding how dogs learn, and what they learn, will guide you toward more successful interactions.