How do you un-ring Pavlov’s bell?
We’ve all heard of Pavlov, his bell, and his dogs, but what did he really discover, and how does it apply to dog training?
Ivan Pavlov noticed that the dogs that he was already experimenting on (he was using the dogs for research on digestion) would salivate when lab technicians arrived at the lab to feed them — before their food was actually given to them.
This seems kind of obvious to us now, but this discovery was the beginning of Pavlov’s pivotal research on what he called conditional reflex and what is commonly called classical or respondent conditioning today.
What is classical conditioning?
Classical conditioning is learning by association. Pavlov’s dogs learned that the technicians predicted the food. Our dogs make similar associations all the time, for example:
- The sound of a crinkly plastic bag may predict treats.
- The smell of birch oil predicts food.
- The sight of another dog while on leash may predict trouble.
These associations are remarkably powerful. A new (conditioned) stimulus is being associated with an existing (unconditioned) stimulus. In the case of the first 2 examples a new stimulus has come to mean “food is available!” while in the third a new stimulus means “danger is near!”
How strong is classical conditioning?
Being able to make these kind of associations is a critical survival skill for any organism – even humans. We aren’t born knowing what the sound of an approaching car is and what it means if the car hits us. But anyone who has suddenly heard a car while crossing a street makes an association and has a reflex — Evade danger!
So it makes sense that these associations have a couple of important properties:
- They generalize very quickly. Learning that a dog approaching on leash is dangerous after one event will very often become associated with all dogs approaching on leash from that event on.
- They are very durable. Classical associations don’t tend to go away on their own, even if the associated stimulus is not encountered for a very long time.
- They tend to override other types of learning. Attempting to punish way a conditioned response or reinforce an alternative rarely works, unless they are combined with counter-conditioning and desensitization (CC&DS). (And punishment is never a good idea.)
How are these associations made?
The obvious way that these associations are made is through experience. If a dog is attacked by another dog she may develop a negative associations with all dogs, dog while on leash, dogs that resemble the attacker, or some combination of factors.
Similarly a pleasant experience, such as a specific person and good treats or a fun activity can create a positive association.
You have probably heard about the importance of early socialization for puppies. Effective early socialization helps puppies deal with new and unexpected things, while an under-socialized dog will often respond to anything new with fear and create yet another negative association.
How are these associations changed?
As I explained above, classical associations are strong. Stronger than the reinforcers and punishers that tend to increase or decrease the frequency of behaviors.
Caffeine is ready for action. Or a nap.
When dealing with a classical association punishment and reinforcement are ineffective in addressing the underlying problem. Classical associations are very powerful, and must be addressed directly by created a new association.
Changing the Association
When a dog (or again — most any other species) has an emotional response to a stimulus the best approach is to change it rather than trying to eliminate it. This is done by combining the stimulus with something the dogs likes. This is what counter-conditioning and desensitization is.
Counter-conditioning and desensitization takes time. How much time depends on a variety of factors, such as the nature of the association, the individual dog’s history and genetics, and our ability to control the environment. For example, CC&DS for traffic noise in a rural area is going to be a lot easier that it would be in a crowded city.
Let’s break down a common problem.
On Leash Aggression Example
We’ve all see something like this: when approaching (or approached by) one dog (let’s call him Buddha) another dog (let’s call her Caffeine) reacts in what can be called, at least colloquially, aggressive behavior. This is often, but not always, on leash
The first step is to figure out how close Buddha can get before Caffeine is completely out of control. What we’re looking for is a “sweet spot,” where Caffeine is aware of Buddha but not completely out of control.
So we might start with Buddha a full city block away. When she sees Buddha her ears might go back, her breathing might increase, and she may orient in his direction. We’ll feed her for a minute or so and then have Buddha move back out of sight.
Ears back. Breathing increases. She orients toward him. Yes, we’re watching the dog very closely when we are performing CC&DS.
Starting the process – feed as distance gradually decreases.
The key here is to avoid going over threshold. Over Threshold
When a dog goes over threshold training is not going to happen. This is the biggest problem I have with some of the “as-seen-on-TV” dog training. Skillful behavior modification rides that edge between too-far-under and over threshold (note that it’s not too-far-over — it’s not over at all.)
Find a spot where the dog recognizes the stimulus but is not “over the top.” If you can’t find it, err on the side of caution. Manipulate the environment. Put up a visual barrier. Move very far away. Whatever it takes.
In our example a sign of her being over threshold might be when Caffeine starts to bark, lunge, and try to pull to get to Buddha. If we tried to grab her collar she might even turn to snap at the person grabbing her collar. In addition to being a state in which behavior modification would not work, it’s also just not a safe state to work in. (In real-life dog training getting bit is a sign of an error, not a badge of honor.)
CC & DS is a Process, Not a Technique
Over time we can close the distance between the dogs, but very gradually.
This is not a linear progression: it’s not five feet a session, ten feet a week, or some another predictable pattern. As a matter of fact sometimes it’s two feet forward and then one foot back.
And each session is also relatively brief! This process can be very stressful for the dog, even when carefully managed under threshold. Sessions should be short, and as always, observe Kenny Roger’s Rule of Dog Training.
Eventually we will get to a point where the response has been diminished enough that we have some choices. Do we want to start to train a specific behavior in response to seeing a dog? Are we happy with a diminished response at a safe distance? After we have made significant progress in addressing the underlying emotional response, we have these options.
Generalizing This to Other Situations
This process can be generalized to other situations. To do that, figure out what the gradual steps are. How do you present the stimulus in gradual increments? It could be, similar to this example, by adding distance and then decreasing it, or by controlling the amount of time exposed to it, by controlling how loud it is (which is often a way of simulating distance) or some other way in which the fearful stimulus is presented gradually to your dog.
- If a dog does not like being handled, start slowly and pair it with something he likes.
- If a dog doesn’t like having his nails trimmed, start with showing him the clippers and work step by step before actually touching the clipper to his nails, while pairing the clippers with a treat.
- If a dog is afraid of siren, try a recording of a siren at a very low volume and slowly increase it each session while feeding your dog treats.
A lot of this is trial and error, and sometimes you really do need someone with a lot of problem solving experience to help set things up and identify the trigger(s).
Basic Graduation 2-14-12
Deciding that an association is causing your dog to behave a certain way means making assumptions about what is going on “inside” the dog. These kinds of assumptions are not always right. As a matter of fact, these kinds of assumptions are what can lead to describing a dog as stubborn, dumb, or even the dreaded (and horribly misused) “dominant.” Which is not a personality attribute dammit. But I digress…
With the understanding that we are making judgements based on our dog’s body language and behavior there is a general rule we can follow. We use CC&DS to change an undesirable response to a stimulus that seems to be driven by a negative reaction to the stimulus. Let’s consider three possible responses to a human stranger approaching a dog:
- The dog attempts to escape.
- The dog lunges, growls, barks, in what we would characterize as an aggressive manner.
- The dog attempts to jump up and greet the person.
In numbers one and two the dog’s reaction is negative. Both reactions are likely driven by fear. In number three his reaction is positive – he is happy to see the person and wants to greet them, albeit in an inappropriate manner.
We need to change the emotional response in scenarios one and two. A dog that is attempting to flee or attack cannot be taught to greet someone politely, and even if it were possible, he would probably still be distressed. We want to make him more comfortable. This is job for CC&DS.
In scenario three the dog is happy to see people! We certainly don’t want to change that. We have a training problem: we need to teach him how to greet people politely.
In situations where we need to make something “bad” become something “good” (or at least a lot less bad) we use CC&DS. In a situation where something is already good but the response is what is “bad” we use training.