Learning What Predicts What:
I am a scientist who studies animal learning and behavior and I am also a dog trainer. As you might expect, I am interested in how to apply what science knows about animal learning and behavior to dog training. The scientific study of animal learning has provided a lot of useful information for dog trainers. For example, these days a lot of dog trainers are fairly well versed in Operant Conditioning. That is, many trainers use reward (and perhaps punishment) as a way to modify a dog's behavior. I have previously written articles about Operant Conditioning as well as its variations.
Another type of conditioning (or learning) is called Classical Conditioning. Dog trainers don't talk about it as much, however, I believe that an understanding of it will lead to a better understanding of why our dogs do what they do as well as to more effective training. Thus, the purpose of this article is to provide the reader with a basic understanding of Classical Conditioning and how it relates to dog training and behavior.
Classical Conditioning is concerned with events in the environment (called Stimuli) that predict the occurrence of biologically relevant events (or "special" Stimuli). The basic paradigm (or setup) is:
|S -----> S*|
|Which should be read as:|
|a Stimulus predicts a biologically relevant Stimulus|
The asterisk indicates that the stimulus is biologically relevant (that is, "special"). To make a long story short, the dog learns "what predicts what". This learning enables the dog to prepare for the biologically relevant stimulus. Ivan Pavlov was a Russian physiologist that discovered Classical Conditioning in the early 1900s. He found that if the sound of a tone (a stimulus or S) reliably predicted the occurrence of food (a biologically relevant stimulus or S*), the dog would learn to salivate to the tone. The salivating helped prepare the organism to receive the food. Classical Conditioning is also called Pavlovian Conditioning and it often makes people think of drooling dogs. In this article, I hope to make clear that Classical Conditioning involves a lot more than just drooling dogs.
Let's consider some examples. A lot of people that own dogs walk them with a leash. As a result of the leash (a stimulus) predicting the walk (which I would argue is biologically relevant because of bathroom behaviors, etc.), the dog learns through Classical Conditioning. That is, when the dog sees the leash, it will get excited and will orient to the door you typically exit from. Or, perhaps you are in the habit of asking the dog, "Do you want to go for a walk?" before you do so. If you do this regularly, I bet the dog will get excited when you ask this question.
Other examples concern meal time. If you feed the dog at the same time(s) each day, the dog learns this. There are structures in the brains of animals that regulate various circadian (about a day) rhythms. Thus, the dog has a way of telling time. In this case, time serves as a stimulus that predicts the meal. When it comes that time of day, the dog may get restless. In some cases, it may even try to take you to the food bowl or bring the food bowl to you. With my dogs, I have a meal time ritual where I ask the dogs to sit for a moment or two while I prepare their meal. In this case, they often drool just like Pavlov's dogs. The meal preparation routine serves as a stimulus that predicts the meal which is biologically relevant.
For much of my life I enjoyed jogging and I would take a dog or two with me since they seemed to enjoy it at least as much as I did. I always put on a special pair of running shoes prior to heading out for the jog. Over time my dogs would become especially excited when I would bring out the running shoes. The shoes (a stimulus) predicted the run (which again, I would argue is biologically relevant to the dog).
Still other examples include toys and traveling to training/play areas. I often play retrieve, tug, and other games with my dogs to give them exercise and as a part of their training. They enjoy this type of activity a lot. Since the toy repeatedly predicts the activity, they get pretty excited as soon as they see the toy. I train my dogs at a local facility. Since this is something we do with regularity, when we get about a half mile from the facility my dogs typically begin whining with excitement. During the warmer weather, I often bring my dogs to certain lakes/rivers to go swimming. As a result of going to the same places on a regular basis, my dogs will often show their excitement before we actually get there.
I hope these examples of Classical Conditioning that occur in the normal course of the dog's life make clear that this type of learning involves more than drooling. Now that you understand Classical Conditioning, I will talk about two characteristics of it that are worth being familiar with. First, Classical Conditioning has an emotional component. Second, in Classical Conditioning, the animal comes to treat the predicting stimulus as if it were the biologically relevant stimulus.
Let's first consider the emotional component. If a stimulus tells you that something good is going to happen, you feel good about it - one might call the emotion joy. It is probably what dogs feel when you tell them to get ready for dinner. If on the other hand, a stimulus predicts something bad is going to happen, you feel bad about it - one might call the emotion distress (by the way, this emotion can be an undesirable side effect of the use of punishment). An example is that dogs are often hand shy (that is, they duck their heads when a person raises their hand) when they have heavy handed owners (that is, owners that punish the dogs by hitting them with their hands).
Things get a little complicated because a stimulus can also predict that a biologically relevant stimulus is not going to happen. For example, if a stimulus tells you something good you were expecting to happen is not going to happen, you would likely feel disappointed. When I have to go someplace and my dogs will stay home, I tell them to "wait". They sometimes express there disappointment with a verbal sigh. On the other hand, if a stimulus tells you that something bad you were expecting to happen is not going to happen, you would feel relief. The table below summarizes these four possible emotions that occur in Classical Conditioning.
Now let's consider how the animal comes to treat the predicting stimulus as if it were the biologically relevant stimulus. This one gets complex, but I will try to keep it interesting and relate it to dogs quickly. Let's start by talking about rats playing basketball. Students in Dr. Alliston Reid's Behavior Analysis class spent the Fall 2009 semester training rats to play basketball using positive reinforcement. They shaped and rewarded the rats for putting a ball through a hoop. Below is a YouTube video of the highlights of a rat basketball game they had at the end of the semester:
Notice how the rats seem to have a tough time letting go of the ball when putting it through the hoop. In fact, they appear to be trying to nibble on it (that is, eat it). Because putting the ball through the hoop predicts they will get a food pellet, they begin to treat the ball as if it were the food pellet. In this case, treating the ball as it were food actually slows down the rate at which the animal gets the food. Anyhow, what does this have to do with dogs? The best example comes from a seminar that I attended some years ago by Pamela Reid, Ph.D. She is the author of the book Excel-erated Learning: Explaining in plain English how dogs learn and how best to teach them.
During the seminar, someone in the audience who was having trouble training her dog to retrieve a dumbbell for obedience competition asked Dr. Reid about training a dog to retrieve (that had no interest in retreiveing) without using force. She suggested creating a pulley system to enable a dumbbell to be lowered into a crate that contained the dog. There would also be a tube/chute that enables a food treat to be delivered to the dog. The training would go like this. Lower the dumbbell into the crate and then deliver a food treat. Do this about once every other minute for a dozen or so trials. Repeat this once or twice a day for a week or two. Since the appearance of the dumbbell predicts the occurrence of the food, the dog will eventually begin treating the dumbbell as if it were the food and will begin to mouth it. This mouthing of the dumbbell can then be encouraged and rewarded through Operant Conditioning, and you are on the road to teaching a retrieve without force.
In summary, Classical or Pavlovian Conditioning is concerned with what stimuli in the environment predict the occurrence of biologically relevant stimuli. It involves emotions and the dog may begin to treat the predicting stimulus as if it were the biologically relevant stimulus.