Poisoned Cues, Fallout and Shutting Down

Or, Why Stephani Really Really Dislikes Aversives!

When a trainer combines positive punishment (P+) or negative reinforcement (R-) with Positive Reinforcement (R+, mostly food), it can really screw up a dog. Not just the dog, but the relationship between the dog, its owner and the environment around it.

Terms like: poisoned cue, fallout, and shutting down all refer to cases where both food and punishment have been used.

For the sake of this post, I am going to use "food" for positive reinforcement and "punishment" for positive punishment. Please note that there are other instances that also cause fallout.

So, a typical scenario in positive reinforcement based training might be: dog sits, gets a click and then a cookie. Happiness all around! If the dog does not sit, then we wait a few seconds and do something else (like a down or loose leash walk) and then re-cue. If the dog sits, click then treat. If the dog doesn’t sit, then no cookie. As easy as that.

In a scenario that combines food with punishment it looks like this: owner cues the sit, dog sits, gets clicked and treated. Owner walks a few steps and asks for a sit. Dog doesn’t sit, so owner leash pops the dog (or smacks the dog’s butt or pushes the dog into a sit), the dog sits and then is clicked and treated. The cue no longer means something good. It becomes poisoned. The dog has no way of knowing if the cue will mean something good or something bad, so it avoids the bad and sits.

You can poison a cue quite easily with a soft (sensitive) dog if you are training while frustrated. You can also poison a cue by introducing a shock collar into the mix and shocking the dog.

Retraining a dog that was trained with punishment or a combination of punishment and food is a long, long road to recovery.

If a trainer tells you that the only way to “proof” your dog for certain behaviors is to start using punishment, then I say bullshit (proofing is another traditional trainer term, but I use it in class and mean the same thing, but without punishment). Proofing means testing the dog in a number of scenarios to see if they really understand the cue. You don’t need to correct (punish) the dog if they screw up - you need to back up your expectations and train more. In other words, you have not trained enough!

 
Note: Part of this post was originally written in another forum. 

Operant Conditioning

I am a positive reinforcement based dog trainer. That means that I strive to teach my dogs appropriate behavior and to do it without using applied an aversive.

What is an aversive? Well, in my world view, an aversive is something that a dog doesn't like. It can be something that causes pain or fear or discomfort. Granted, there are plenty of things in life that causes pain, fear and discomfort, but I do not want to be one of those things. And, I can train without them.

Most dog training is based on Operant Conditioning: Positive Reinforcement (R+), Negative Reinforcement (R-), Positive Punishment (P+) and Negative Punishment (P-). I use R+ and P- when training my dogs (I have listed them in green below). These are the two categories that do not contain aversives.

Quick guide:
Reinforcement  = increasing behavior 
Punishment  = reducing behavior
Negative = taking something away
Positive = adding something


Positive Reinforcement: This category means you are adding something in order to increase behavior. The thing that you are adding can be: food, attention/praise, and play.

Negative Reinforcement: This category means you are taking something away in order to increase behavior. This category uses aversives.

A classic example of this is when a rat is being shocked and it pushes on a bar to stop the shock. The shock increases the likelihood that the rat will press the bar. In dog training, an example is holding a choke collar tight until the dog sits and then loosening the collar when the dog sits.

Positive Punishment: This category means that you are adding something in order to decrease behavior. This category uses aversives.

An example of this is a shock collar to make the dog stop barking. Or, an electric fence to reduce the dog's wandering behavior.

Negative Punishment: This category, you are removing something in order to reduce behavior. This category can have aversives, but most often, aversives are not needed.

Examples of non-aversive uses in this category is a time out (you are removing freedom). Or, turning one's back on a jumping dog and walking away (you are removing attention).

Some other terms that I frequently use:

Desensitization and counterconditioning: This is, essentially, classical conditioning (for instance, pairing a feared object with food) while slowly decreasing the distance and amount of time the dog spends with the feared object. This technique is used a lot with fearful or reactive dogs.

Poisoned Cue, Fallout, Shutting Down: see the next post for more information about these.