Clarifying Behavior Science: Classical Counter-Conditioning

This blog post is the first in a series about clarifying behavior science.   First, I am posting a quiz on the Canine Behavior Science Facebook group that uses real-world examples of misinformation about various dog training and behavior analysis topics — misinformation often written by and spread by professionals.  Next, I follow-up with a blog post like this to explain the answers.

The goal here is not to embarrass anyone, and that is why I never mention names.  The goal is to build a library of common misinformation addressed with valid, credible correct information so that all of us — professionals and pet owners alike — can be more efficient and humane living together as dog and human.

Why Address Classical (Respondent) Conditioning Confusions?

I so often see confusion around what respondent conditioning is, and how to apply it effectively.   This confusion leads to the common—and mistaken—assertion that classical counter-conditioning doesn’t work.  A typical reason why it doesn’t work is that many implement it poorly.  The strawman of “classical counter-conditioning doesn’t work” has created an explosion of pseudoscientific protocols as an attempt to replace.

Another problem with poor quality implementation of classical conditioning (or counter-conditioning) is that, as professionals, our clients deserve effective and efficient solutions and treatment plans.  If we are not doing a procedure well because we don’t understand it, it will negatively impact treatment/behavior plan effectiveness.  I believe our clients deserve better than that.

Finally, classical conditioning is a powerful tool when applied effectively.

For a general overview of what Classical Conditioning is, read Does Counter-Conditioning Bribe Our Dogs.

Confusing Operant and Classical Conditioning

One of the first things I ask when people tell me they’ve been doing classical conditioning and it doesn’t work is to describe what they are doing.

Typically, I get responses like answers B and C, which are both examples of operant conditioning.  Eileen Anderson, who writes the blog Eileen and Dogs, already did a blog post dissecting the difference between classical and operant conditioning, Are You Really Performing Classical Conditioning?  It’s a great read with great examples.  If you thought that either B or C were good examples, it’s worth the time to read Eileen’s blog.

Continous Treats Create Good Feelings — But Not Necessarily For the Stimulus You Want

The rest of the examples are various forms of tossing treats to the dog while in the presence of the trigger.  Seems to make sense, right?  Treats make dogs feel good and “Pavlov is always on your shoulder.”

Let’s take Kona playing find-it in the reactive dog class.  What is the dog associating with the treats?  Often, as humans, we assume the association is being made with the other dog because, after all, that is our intent.  But what about the classroom?  The instructor?  The treat bag?  The other students in the class?  How does the dog differentiate all these other stimuli from the one that matters?

Now let’s look at Bob and Niles.  Every night, Bob takes treats and sits on the kitchen floor, feeding them to Niles.  Niles will take the treats, but he is still leery of Bob at all other times.  What’s going on?  Doesn’t this mean classical counter-conditioning doesn’t work?  Well . . . just like with the Kona example, this procedure has no way to clarify for Niles that Bob is what is making the treats happen.

But Isn’t Pavlov Always on Your Shoulder?

As I said at the beginning of the blog, all of these examples came from professionals.  Their rationale for these being good examples of classical counter-conditioning is that “Pavlov is always on your shoulder.”  This response is a misinterpretation of this quote by Bob Bailey (who is one of my mentors).

Pavlov is always on your shoulder does not mean that tossing treats around to create a “feel-good” moment is an effective classical conditioning protocol.  It means that you can’t separate out the emotional aspect of how you train when you use operant conditioning.  It does not mean that the animal will be able to make the right association because they are feeling good in the moment.

Make the Pairing Clear

The goal when doing classical counter-conditioning is to use great mechanical skills and eliminate all the noise for the dog so that he understands, “when I see that stranger, food appears.”  Here are some best practices for classical conditioning procedures to ensure that the dog is making the right associations:

• Order of events.  The stranger appears first.  Then the treats come out and in the dog’s mouth.  (Remember, the treats are delivered no matter what the dog’s behavior is when the stranger appears.).
• Offset.  To make that association, we also must clearly show that when the stranger goes away, the food stops.  This is call offset.  The onset of the stranger — food appears.  Offset of the stranger — food stops. The reason that examples A, D and E are not good examples of classical counter-conditioning is that none of them had an offset.  They were just one long continuous feeding fest.
• Longer intertrial intervals.  Rapid cycles of “here’s the stranger, food, stranger goes away, no food” can also interfere with creating the association.  A constant cycle of here it is and here it isn’t can muddy up the order of events — what is coming before the other?   Also, a pattern is set up making it hard to set up the prediction of stranger arrive equals food is coming.  Best case is if there can be irregular longer intervals — up to hours — of the presentation pairing and the offset.  This, of course, is challenging for those of us professionals who do in-home or classroom sessions as we have time constraints.
• Varying all variables EXCEPT the two stimuli you are pairing.  The location, the time of day, the clothes you are wearing, the treat bag you have on, the other people present — all these can potentially get paired with the food instead of the stranger.
• Make sure nothing else is predicting the treats.  Plastic bags crumpling.  Hands moving towards the treat bag.  Looking towards the stranger.  All of these could end up being predictors of the food arriving in place of the stranger.

It’s a lot of stuff to keep track of, but as professionals, we should know how.  Even pet owners who understand these success factors can build great plans and implement them well.  Here is Pamela Rhodes, a member of the Fearful Dogs Facebook group, who is working with her dog Buster on various veterinary procedures.  With a bit of knowledge gained in the group and a telephone consult with Debbie Jacobs of Fearful Dogs, Pamela created and implemented this lovely classical conditioning plan for her dog, Buster.  You can view it here:  Classical Conditioning for Husbandry.

Is Tossing Treats Ever Helpful?

Absolutely.  Providing you understand what is happening when you use that approach and don’t confuse it with classical conditioning.  It is useful for problems that happen “in the moment,” typically as a distraction.

In the reactive dog class example, playing find-it in the classroom is likely acting as a competing reinforcer for the dog.  Either eat treats — or get reinforced for “reactive” behavior by making the other dog go away.

I often use it to prevent the occurrence of problem behavior in that while we are training a new behavior and trying to prevent the occurrence of a problem behavior.  For example, I’ll tell clients if they are at the vet office and a rambunctious dog trots in and moves towards them to feed their dog and move away.    I’ll have clients with dogs that jump up on people to toss treats to prevent the dog from jumping up.

This approach is helpful with a variety of reinforcers —  toy play, treats, noseworks — anything the dog enjoys doing and has a higher reinforcement value than the behavior we are trying to reduce.  It is a management-type procedure only during that transition period until the training is proofed and generalized.

The Quiz Answer and Responses

The correct answer was 6 and almost everyone got the quiz correct!  Great job.  Everyone was able to identify the difference between operant and classical conditioning, and most of you mentioned that lack of offset was the reason why none of the examples were good.  Darn, I was trying to trick everyone!

If you want to make sure you get to see the next quiz, like the Canine Behavior Science Facebook page.  It will be posted there at least monthly.