Looking for the recording? Click here
To understand what a Conditioned Emotional Response actually is and how to design a training plan to build one.
What do we need to know?
- What is the role of classical conditioning?
- What is the role of operant conditioning?
- How can we arrange things to get the picture and behaviors that we want?
Operant conditioning (OC) and classical conditioning (CC) are natural processes that are constantly occurring at the same time. We can’t ever separate them. And that can work both for and against our training. So it seems reasonable that we’d like to use all the tools at our disposal to use both processes to our advantage.
When we are applying a CC procedure to achieve a “positive conditioned emotional response” (+CER), these conditioned emotional responses manifest as respondent (aka reflexive/involuntary) and operant (voluntary) behaviors. The very things we call “emotions” are a combination of both respondent and operant behaviors.
Let’s take a look at some examples.
While CERs come in all shapes and sizes (some we would call positive, and some not so positive), I particularly enjoy watching “positive” examples. So let’s look at a few.
In this video, you’ll see clips of dogs who have completed a food guarding protocol and are exhibiting the fruits of that labor… completely different behaviors than when the training started.
Look at how each dog’s behavior changes as the trainer approaches the food bowl.
Most trainers would point to the wagging tails and bouncy body movements, and be happy to say that these dogs are all showing a positive CER to the trainer’s approach.
Great! But how did we get here? What are the principles in play that are giving us this result? And how can we deconstruct them to develop CERs in other contexts?
Let’s start with a quick review of classical conditioning.
In a nutshell, classical conditioning is a process where an initially meaningless (neutral) stimulus or event (like a bell ringing) is immediately followed by something (like food) that is already meaningful (the dog already has “feelings” about it).
After repeating this association over and over, the bell becomes a “Conditioned Stimulus” and responses/behaviors that were initially displayed in the presence of the food start being displayed in the presence of the bell by itself.
One of the most common examples that’s actually relevant to us is the process by which the clicker becomes predictive of food. We all know this story already, but let’s play it out and unpack it a bit from the classical learning perspective.
Initially, the click is a neutral stimulus (NS) without meaning to the learner.
Through the repeated pairing of click followed by food (UCS: Unconditioned Stimulus), the learner develops an association between the click and the treat and demonstrates behavior we term a Conditioned Response (CR).
When the behavior(s) appears particularly “emotional” to us, we label it a Conditioned Emotional Response (CER).
The behaviors may include salivation (evident to us by drooling or foaming in some breeds or a lip lick in others), head lift or head turn toward the source of food (handler), bouncy movements, wagging tails, and so on.
The important piece here is that any way we slice it, we’re still talking about behavior.
Here’s a clip of a dog’s response to the sound of the clicker (and yes, she’s blind). What are the behaviors you see here?
There’s an evident change in her expression from before the click to after the click, right?
And expression *IS* BEHAVIOR!
There are muscles and body parts moving. Those movements count as behaviors.
So when we say Ella demonstrates a positive CER to the clicker, we mean we see a change in her behavior after she hears the click.
And we define the CER based on these observable behaviors!
She looks expectant and anticipatory of a treat, doesn’t she?
The behaviors we label as “expectant” and “anticipatory” include: head movement, body shift forward, and orientation toward source of reinforcement (her head turns toward food source).
She may not be able to see her handler, but she responds to the click by snapping her head up toward the position in which she usually receives reinforcement.
So, let’s keep playing this one out.
If we’ve established that we’re talking about behaviors here – behaviors that are overt and observable – then some of those observable behaviors are likely to be under the animal’s control. After all, we can see a great many of them; there’s more going on here than reflexive, unobservable, internal events (behaviors), right?
If some of these behaviors are under our learner’s control (and they are!), then this puts us squarely in the land of operant learning!
Before we take this further, please know I’m not suggesting that classical learning isn’t at play here; it almost certainly is. I am suggesting that operant learning is also at play. So let’s unpack it from the operant perspective now.
The clicker (our Conditioned Stimulus in our classical procedure) is also a cue, right? It’s an antecedent that facilitates the learner’s performance of a behavior.
Remember the ABCs of operant conditioning?
The click is the cue that tells our learner to access reinforcement. We have an A!
And we have a B! What are the behaviors we observe in response to the click? We tend to see movement toward the source of reinforcement.
In Ella’s case, we see her head lift and weight shift to the source of reinforcement (her handler).
Ella receives food when orienting to the source of the food. Sure, the food isn’t contingent on her behavior in this case (because the trainer is giving her the food no matter what), but that doesn’t mean operant learning isn’t happening!
How can we tell? Look to see if see if the behaviors increase in frequency!
In this clip, she continues to orient in subsequent repetitions. Since we see increased frequency of behavior, we can suggest that the food is reinforcing her head movement. We have a C!
ABC. Right smack in the middle of a classical learning example.
Okay, okay I hear you: CC is supposed to be about reflexive/respondent/involunary behaviors.
And here I am saying that a “classical” procedure can be operantly unpacked. Well, yes. Yes, I am.
Here’s the thing.
We can think of any behavior as on a spectrum from voluntary to involuntary. Some behaviors like a freeze or an eyeblink tend to be closer to the involuntary side. Others like a lip lick or even a bark seem to fall somewhere in the middle (but sustained barking would fall on the voluntary end).
Orientation like a head turn or head lift also fall toward the voluntary side of the spectrum. If the behavior is voluntary, then it is under the animals control.
And if it’s under the animal’s control, it can be reinforced (or punished) by the consequence that follows it. EVEN IF WE THINK WE ARE APPLYING A CLASSICAL PROCEDURE!
Let’s look at another example:
In this video, look at both dogs’ response to the sound of the clicker.
The border collie on the left whips her head toward the source of reinforcement. The red dog on the right snaps her head around and licks her lips in anticipation of reinforcement. We’d say these dogs demonstrate a “positive CER” in response to the sound of the clicker, right?
And we define that CER based on these observable behaviors: head snap and lip lick!
Now, let’s unpack it from an operant perspective:
B: Head snap / lip lick
Here we have it again!
ABCs smack in the middle of the classical learning paradigm. The two learning processes are so interwoven that even when we think we are applying one procedure or the other, we can’t separate the learning processes at play. Both Classical Conditioning and Operant Conditioning occur simultaneously and can’t be separated in real life (beyond of our selection of which procedure we are basing our training plan on).
Why does this matter so much?
Well, I think we get focused on wanting our dog to have a +CER to something. And to get there, we try to apply a CC procedure. And this may work really well – or fall totally flat. Not because of some flaw in the learning process, potentially because of a flaw in our procedure, and also potentially because we didn’t consider the impact of both processes.
First, let’s remember that when we say we want him to have a positive CER to something, what we’re really saying is we want him to feel good about some cue or behavior or event or AND we define the CER based on the observable behaviors we want to see.
And since we are interested in training or changing voluntary, observable expressions of behavior, we might use either procedure (classical or operant) to get us where we want to go: a +CER defined as observable behavior.
Often the behaviors we want to see are fluent behaviors in response to some cue, stimulus, or event, right? We want our dog to respond immediately to the cue (low latency), perform precise behavior, and do so reasonably quickly (speed). We also want to see behaviors we define as “happy” like a sweeping tail wag, relaxed mouth, and a relaxed body.
So what do we do? From an operant perspective, we need to define all of those behaviors, use a thoughtfully constructed shaping plan (with many splits), and train the behaviors to fluency with the use of consequences that are of positive value to the learner.
From a classical perspective, we still need to define all of the behaviors we want, use a thoughtfully constructed CC or Classical Counterconditioning plan (with a sound desensitization hierarchy – think splitting!), and teach the behaviors to fluency (yes, I’m moving right on in to OC here). Again, we also want to see behaviors we define as “happy” like a sweeping tail wag, relaxed mouth, and a relaxed body.
From Classical to Operant
Next, let’s look a little bit at how we can “build a +CER”
Step 1: Define what a +CER looks like in terms of observable behavior.
Step 2: Choose your procedure to train the shit out of those behaviors.
Here’s an example of a shelter dog who demonstrated the behaviors typically associated with “resource guarding”: accelerated eating, snarling, growling, and even biting when in possession of a bowl of food and approached by a person.
Step 1: Define what a +CER looks like in terms of observable behavior.
My goal behavior is a head lift out of the food bowl. I also want to see a relaxed body, tail wag, relaxed facial muscles, and so on. All of these behaviors would fall into the bucket I call +CER.
Step 2: Choose my procedure and train the shit out of the behaviors I want to see.
The training plan here is a Classical CounterConditioning plan that takes advantage of the operant contingencies at play.
Stick with me; I’ll show you what I mean in a moment. A desensitization plan is constructed to incrementally adjust the conditions, starting from “least difficult” (no food in bowl) working up to “most difficult” (really delicious food in bowl).
The first two repetitions you see in the video are straight Classical Counterconditioning procedure at the easiest desensitization level (no food in bowl). The trainer’s approach is paired with delivery of food into the bowl.
Hooman walks up = food appears!
The food is not contingent on any specific behavior from the dog. The dog gets food dropped in the bowl no matter what he does. In these first two reps, you’ll see Carlin’s relaxed body, tail wag, and orientation toward the source of reinforcement (trainer).
Watch him look over his shoulder in hopes his trainer is coming his way; I love that moment. All of this looks good, right? I have the relaxed body I wanted. What I don’t yet have is a head lift out of the bowl.
So how can I build that behavior?
Arrange the conditions so the behavior occurs! Watch the 3rd approach on the video again; here’s where things get interesting.
When the handler approaches this time, Carlin is still eating food that’s inside the bowl (bowl is no longer empty – a change in condition/increase in desensitization step).
When the handler approaches, Carlin lifts his head out of the bowl. Bingo. There’s the behavior I really want. And it’s well under this dog’s voluntary control.
Welcome to Operant Land.
The A and B are well in place. Now, it’s time for to reinforce the behavior since I know I want a lot more of it. Treat placement is a key strategy here.
This is the first rep in which you see the trainer deliver the food straight into the dog’s mouth instead of dropping it into the bowl. That strategy is deliberate – the goal is to reinforce the head lift and use our treat placement in the most effective way. There’s our C.
What did we just do?
We seamlessly shifted from a classical procedure to an operant procedure to leverage both learning processes. We recognize that although we’re applying a classical procedure, operant learning still occurs. And we’re ready to jump on the operant behavior we want and reinforce it.
How is this even happening?
Because of the strict, artificial dichotomy we’ve been taught regarding CC and OC, we tend to regard CERs (taught in the scientific lit as a product of classical conditioning) as a nebulous concept: something intangible, not readily observable, some internal emotional event that may or may not occur inside the animal.
But if we can stay focused on our lens (observable behavior) – the only lens readily available to us as trainers and behavior consultants, then we can move beyond CERs as a murky concept, take the term for what it is (a label taped onto a box of behaviors), and focus our energy on observable behavior.
Many of the behaviors we consider indicative of a CER are under the animal’s voluntary control.
And if the behavior is under voluntary control, whatever follows it may reinforce or punish it even when we think we are applying only a classical conditioning procedure.
These behavioral responses – these emotional behaviors– are subject and susceptible to reinforcement and punishment – to operant learning – learning that occurs even when we think we are applying a classical conditioning procedure.
If the animal demonstrates a behavior under his control, then whatever happens after that behavior is expressed may reinforce it – even when we are only focused on the CC procedure.
What’s the big take-away?
We can call these behaviors CERs. But CER is just a label on a box with a bunch of behavior dumped in it!
We are always looking at observable behavior.
I’m not suggesting there aren’t respondent behaviors occurring as well. I am suggesting that those respondent behaviors are not within our available lens as trainers and consultants.
At our level of analysis, CERs *are* behaviors – behaviors we can see and behaviors we can reinforce!