What do the following training descriptions have in common?
- “My dog’s afraid of strangers. But when she stops barking and makes eye contact with me, I give her a treat.”
- “I hold her foot. Then I give her a treat after I clip each toenail, as long as she stays in place and doesn’t pull her foot away.”
- “When we have guests, I wait for him to show some calm behavior like stretching, breathing more deeply, or lying down. Then I give him a treat.”
- “We play LAT (Look At That). I say ‘Look at the dog’ and she does. I mark, then give her a treat.”1)
- “When the cyclists go by, I cue my dog to sit, then I treat.”
These are all training methods designed to help a dog cope with something uncomfortable, undesired, or scary. But they are not classical conditioning.
These five descriptions are all operant methods. How can we tell? It’s because the food is given as a consequence of a behavior. In each case, a certain behavior is required before the dog is given the food morsel. There is a contingency. If the training is successful, the trainer reinforces the behaviors of making eye contact, staying still, stretching, breathing deeply, lying down, looking at a trigger, or sitting.
Each of these follows the operant model:
- an antecedent (usually the trigger appearing);
- a behavior as specified above;
- and a consequence (food).
The goal is for the dog to learn to perform these specified behaviors instead of being reactive or tense. Any of these could be a successful method, especially if the dog’s unease is not extreme.
Classical conditioning involves a different type of learning.
The Real Thing: Classical Conditioning
First, a little about respondent behaviors. Respondent (involuntary) behaviors include reflexes like the following:
- blinking when a puff of air is directed at the eye;
- sneezing because of a bright light or an irritant in the sinuses; or
- salivating at the sight or smell of food.
Respondent behaviors follow a two-part model: Stimulus/response. In general, respondent behaviors can’t be reinforced or punished. Most of them aren’t under our control. (There are some exceptions.) Think about it this way: if you got praised or got a chocolate chip every time you got goose bumps (cutis anserina), would that happen more? Nope. Goose bumps are a response to a specific stimulus, usually cold or something that causes a strong emotion. Every respondent behavior likewise has a stimulus or stimuli that will cause it to occur.
Classical conditioning is a name for a procedure where we “attach” the respondent behavior to a new stimulus.
The Oxford Dictionary defines it thusly:
A learning process that occurs when two stimuli are repeatedly paired; a response that is at first elicited by the second stimulus is eventually elicited by the first stimulus alone.
We can cause respondent behaviors to occur in response to a new stimulus by pairing them as described above. Pavlov’s dogs are the standard example. They were conditioned to salivate at the sound of a buzzer that meant the food was coming. The buzzer (first stimulus) reliably predicted the appearance of the food (second stimulus).
In dog training, we use classical conditioning to change the dog’s physiological and emotional response to a stimulus. For example, if a dog is afraid of the sound of delivery trucks we can consistently feed the dog roast chicken after the sound. The dog’s attitude towards delivery trucks will likely change. It will go from fear to, “Yay, chicken is coming!” The truck sound itself will come to trigger the body’s preparation to ingest food and the happy feelings that can accompany that. The happy feelings and behaviors are why we do this. We aren’t trying to teach the dog to want to eat delivery trucks. We are attaching a positive conditioned emotional response (CER+) to something that was formerly scary.
So how is it different from the five examples above? Here are two examples that outline the basic process of classical conditioning done correctly.
Classical Conditioning: Two Examples
- “After my dog sees the bicyclist, I wait just a moment, then start feeding her. As long as the bicycle is passing by, I keep feeding. Then I stop feeding a moment after the bicycle disappears from sight.”
- “My washing machine makes a certain beep if the load get unbalanced. Whenever it beeps and my dog hears it, I give her a treat right afterward.”
Notice that nowhere in those two descriptions is there any mention of a required behavior. Trigger happens; dog gets food. We are so accustomed to asking for a behavior that this can seem quite foreign at first. There are other important issues involving the mechanics of the process, such as timing, that I haven’t described above. Someone who described their training with the phrases above could still be making mistakes in the training. But those are the bare bones descriptions that generally mean that the method is classical, not operant.
Operant behaviors can change as a result of classical counterconditioning. Former behaviors that were prompted by the fear can extinguish when the dog is happily anticipating food. The dog will likely stop panting, pacing, and barking at the delivery truck if we condition him that truck noises predict chicken. Instead, he’ll be salivating, wagging his tail, and looking for the chicken source.
Classical Conditioning vs. Classical Counterconditioning
“Classical conditioning” is a general term. But we generally use the term “counterconditioning” when we know that the dog already has a fear response to the trigger. We aren’t starting from neutrality; we are attempting to “counter” a negative emotional response. In that case, we usually include desensitization as part of our method as well. That’s beyond the scope of this post.
But I do have other posts and videos with examples of both classical conditioning and counterconditioning/desensitization. Check out the following. The first two entries are about classical conditioning and the third and four entries are about counterconditioning/desensitization.
- “Classical Conditioning: Creating a Positive Response to Barking” (blog)
- “Conditioning a Positive Response to Another Dog Barking and to Other Distractions” (movie)
- “Stinky Stuff on My Back: DS/CC for Flea Treatment“
- “Successful Desensitization and Counterconditioning”
“Pavlov On Your Shoulder”
Bob Bailey says that whenever we are training, Pavlov is sitting on one shoulder and Skinner is on the other. As one grows in importance, the other shrinks. What this means is that even while we are teaching a dog with operant conditioning (Skinner), classical conditioning (Pavlov) is going on. If you train with food and toys and other fun, the dog usually gains a positive emotional response to you, the activities you do together, and even the place where you typically train.
The converse is also true. When we do classical conditioning and pair food with a stimulus, we can quickly start to reinforce related operant behaviors. For example, when the dog comes to expect the food after a stimulus, he will start to turn to or approach the source of the food, usually the trainer. Those orienting behaviors occur in between the stimulus and the food, so they get reinforced.
I’m including this section about the interplay of two learning modes because some people use the “Pavlov on the shoulder” comment to claim that the operant training I described in the five comments is classical after all. They will say that all training has elements of classical conditioning, since associations are being made. This part is true. But while both processes are usually going on at the same time, our methodology targets one or the other. The methods are different. Understanding the differences can help us be more effective trainers.
I hate to tell you this. Just to make things a bit fuzzier, there is a term called “operant counterconditioning.” It’s not used that often. But it’s the reason I have been specifying “classical” conditioning and counterconditioning all through this post. Veterinary behaviorist Dr. Sophia Yin defined operant counterconditioning as follows:
Operant counterconditioning is when you train an alternate, incompatible behavior. For instance, if a dog lunges and barks every time he sees other dogs across the street, you can train the aggressive dog to watch you and go through other obedience exercises when he sees dogs. —Rapid Reversal of Fear and Aggression in Dogs and Cats, Sophia Yin, DVM, MS
Some of my five “not classical conditioning” examples above could qualify as operant counterconditioning.
Why Does It Matter?
I realize that not everybody is a nomenclature nut like I am. But if we want to learn about different techniques, know their strengths and weaknesses, practice them, and discuss them, we need to know the correct concepts and terminology. I have seen dozens of people say that they were performing classical counterconditioning when they were using an operant method. I’ve mixed up the two myself. That usually indicates more than an accidental terminology problem. It usually means that the person really doesn’t understand what classical counterconditioning is.
These short movies show fun examples of conditioned responses. When we perform classical conditioning, we look for the moment where the dog starts anticipating the food after the new stimulus. In the first movie below, Zani gives a clear, “Where’s my food??” look when I pause with the food delivery after touching her back with a plastic syringe.
In the second movie, Hazel goes beyond the double take. She’s wagging her tail but she also licks her chops (a sign of salivation) when she notices the nail file.
Has anybody else’s dog gotten as far as actually salivating as a result of classical conditioning?
Copyright 2016 Eileen Anderson
Thank you to Lori Nanan of Your Pit Bull and You and the wonderful Hazel for allowing me to use the cool movie and photo.
|1.||↑||Leslie McDevitt first described the Look at That game under that name in her book Control Unleashed. She also includes a classical conditioning protocol by the name of Open Bar/Closed Bar in the book.|