What is a Correction?

Whenever I engage in conversations online about the use of corrections in dog training, someone inevitably poses the question, "what is a correction?"

This question is not typically asked because somebody is genuinely interested in knowing exactly what each of the discussion participants really mean by the term.  Usually, the person posing the question is trying to make a point of his or her own about the use of corrections in training.

But . . . honestly . . . I believe that this is a question that we should be actively discussing!  This word can mean different things to different people.  I would love to see more trainers - both those who choose to incorporate correction into their training and those who do not - put their understanding of the term, in as much detail as possible, out on the table for all to consider.

When you use the term "correction", what, exactly, do you mean?

In this post, I am going to explain my answer to that question, and how I have come to formulate this definition.

When I speak of a "correction" in a dog training context, I mean:

A deliberate application of something that is aversive to the dog that is intended to communicate that a particular action or behavior is wrong or incorrect.

The aversive used can range from mild to harsh, but its use is intentional on the part of the trainer.

More than "Making the Dog Correct"

It is often claimed in online discussions that a "correction" is anything that "makes the dog correct".  That would be a general definition, indeed.

If it were true that a "correction" really were simply anything that "makes the dog correct", then some things that are very pleasant to the dog could be classified as "corrections".  However, in practice, those who choose to incorporate correction into their training typically do not classify the use of things that the dog likes to make him or her correct as "corrections".

I have presented these scenarios, or similar ones, on several occasions to groups of trainers who use corrections in their training, and have asked if they consider these to be examples of "corrections":

The dog is cued to do a trick and the dog does not comply.  The handler gets out a treat and lures the dog through the trick, gives the treat.  The handler then cues the behavior again and the dog is successful.

The dog is sent over a jump but runs around it instead of taking it.  The handler ignores the error and puts a target out on the other side of the jump and sends the dog again.  This time the dog takes the jump correctly.


The dog is called but does not come when called.  The handler goes to the dog with a neutral demeanor and, upon getting to the dog, says the dog's name cheerfully while patting his or her leg.  As the dog starts to walk along with the handler, he or she continues to praise the dog for returning as they walk along.

The answer to my question, "do you consider these choices, on the part of the handler, to be corrections?", has consistently been a resounding NO!

That begs the question . . . why not?  In each example the dog was "made correct". 

Obviously, there is more to a "correction" than simply making the dog correct.  And, clearly, a "correction" is not intended to be something that the dog enjoys.

That is why I define "correction" as aversive.

More than "Information"

Another very common explanation that I see for the term, "correction" is that "a correction is just information".

I do not dispute that a correction is, in fact, information.

However, I would like to see those who define "correction" this way go on to give more detail about the type of information that is intended to be conveyed through the correction.

There are many ways to give a dog information.  Praise is information.  Delivery of treats is information.  The click of a clicker gives the dog information.  Being on a leash provides information.  A well conditioned target is information.

Is use of a "correction", by virtue of being "information" exactly the same thing as use of praise, treats, a clicker, a leash, or a target?

In order to answer that question, we must consider the nature of the information that is being conveyed.

Information is not some neutral concept.  Information is communication.  By providing information, we strive to say something to our dogs.

Through use of praise, I intend to convey, "I am pleased with what you just did".  By the click of a clicker, I am clearly stating, "that exact behavior in that instant is IT!"  When my dog and I are joined by a leash, I am conveying, "we are sharing this space together now and we need to cooperate in our movement".  And a target is telling the dog, "go there" or "lay down here" or "paw there".

What information does a correction give to the dog?

That is the question that I would like to see more trainers who incorporate correction into their training answer.

I would say that a correction conveys this information: "you are wrong" or "that is incorrect".

The merits and disadvantages of communicating that message to a dog in training is a matter that is up for legitimate debate.

But for those who define a correction as "information", the time has come for you to provide more information about the exact nature of the message that you are looking to convey through the use of corrections.

A Uniform Definition?  Probably Not!

It is a fact that there is not one standard vocabulary in the discipline of dog training that all trainers, regardless of personal training philosophy or chosen methodology, recognize and adhere to.

Many lament over this, and express a desire for there to be such a standard.  But the fact of the matter is that dog training in the United States is not a regulated industry, and we, as trainers, really are free to define the terms we use as we wish to define them.  This may make for some confusion, but it really isn't something that any of us have the power to change at this time.

I propose that the next best thing is open, honest, and respectful dialogue among trainers about the terms that we use.  We may not agree on how we choose to train, but we can take steps toward understanding one another better.  I believe that thorough discussion of how common terms are used between different trainers would be a big move in that direction.

 

Dealing With Errors

One of the most common questions that I hear from trainers who choose to incorporate correction into their training is: "what do you do when your dog doesn't do what you want?"

This is a great question.  Some people actually have the impression that +R trainers allow their dogs to do whatever they want, whenever they want, and that if the dog does not respond correctly to a trained cue, we do absolutely nothing.  This is a misconception.  Our dogs would not be trained if we did not do something to address errors.  The fact that I choose not to employ some kind of correction does not mean that I choose to do absolutely nothing.

So, what do +R trainers do to deal with errors?

First, allow me to clarify that I am speaking specifically about addressing errors that occur with fully trained behaviors with which the dog has demonstrated prior and consistent understanding.  Also, I am not addressing situations where the dog is choosing not to comply to a cue due to physical pain or discomfort, or because of significant fear, or circumstances where non-compliance would put the dog (or anyone else) in danger.  In those instances, training is not the issue at hand, and the pain, fear, or safety issue must be addressed directly.

Dealing with Errors from a +R Point of View

1.  Ignore the Error

Sometimes this is the best way to handle an error, especially if it is a very random or out-of-character mistake.

There are times when I mean to turn left and I turn right.  In those cases, I don't need information from anyone else to tell me that I went the wrong way.  I realize it the second I make the wrong turn.

I approach occasional random incorrect responses from my dog from this perspective.  No living being is 100% all the time.  There are times when I conclude that the best response to an error is to simply ignore it.

Example:  I ask my dog to jump up on the sofa and sit so I can put on his collar, and he jumps up and lays down.  I proceed to put on his collar in spite of the fact that he is laying down, not sitting.

In this instance, the fact that he is laying down, not sitting is of no consequence, so I ignore the error and go on with what I need to do.

2.  Provide positive feedback for the effort, but not the highest value reinforcer

This is the most common way that I choose to handle errors.  And this is always my approach when my dog and I are working in a new context, or in a situation where the dog might be experiencing some level of stress, or if I see confidence wavering.

I might say, "good try!", but not give a treat.  Or I might give a tiny bit of a treat, but not a jackpot.

In these cases, I do let the dog know very clearly that I am pleased with the efforts made, but I leave something bigger for the dog to score when the correct response is given.

Example:   I cue a tunnel on an Agility course, and the dog takes a jump.  I say, "good try" and toss a treat behind me to set the dog up to try again.  I reserve a game of tug (highest value to this particular dog) for the correct response.

Some trainers object to this approach, claiming that the dog will persist in the error if any measure of reinforcement is provided when an incorrect response was offered.  This has not been my experience.  I have found consistently that providing some degree of positive feedback for effort has resulted in the dog remaining engaged, keeping his or her mind in the game, and actually builds understanding of correct responses much faster than ignoring incorrect responses altogether.

However, if the dog continues to repeat the error, I do break off and provide extra assistance to help the dog see how to be correct and get the higher value reinforcer.

 

3.  Reset the Exercise/Cue dog again/Jackpot correct response

This approach will follow either of the above choices - I have ignored the error, or I have provided some positive feedback for effort.  Or, I have broken the dog off mentally from the first attempt by tossing a treat or giving the dog a chance to sniff or take a moment of down time before setting up again.

Now the dog has a new opportunity.  The behavior is cued again.  If the dog is correct, a jackpot is given to emphasize that the correct response has been given.

Example:   Dog is cued to sit and stay.  As handler walks away, dog gets up.  Handler returns to dog, says, "good try" and tosses a treat to break off the exercise, then sets the dog up again and cues the stay, walks away (dog stays).  Handler returns, gives dog five treats in succession as a jackpot.

Frequently, after the dog has responded correctly and received the jackpot, I will move to something else for a time.  In doing this, I allow the opportunity for latent learning to occur after the jackpot given for the correct response.

4.  Provide more, or clearer, information

Sometimes when an error is made, the dog needs more information.  Perhaps a reminder of what is desired, or a clearer cue.

At times, I will go out of my way to provide this information.  I might show the dog a large visual cue, or I might bring a target into the picture, or I might use some kind of physical guide (ex. a ring gate or platform), or even a food lure to set the dog up for success and to give the dog extra information, or the needed reminder.

Example:   I cue my dog to down from a stand and my dog offers a bow.  I get a mat and repeat the exercise, stopping so the dog is standing on the mat when we stop, and then cue the down.  Since the dog is strongly conditioned to down on the mat, the dog lays right down.

I am honestly not concerned, in these instances, with the fact that the dog is "supposed to know" what I want.  If the dog is not responding correctly, something has, in fact, broken down.  The dog may honestly be confused, or the dog may not be able to generalize the behavior in that particular situation, or the dog might be stressed or distracted.  Unless I have reason to be concerned about my dog's well being, I am not really concerned with any of that.  I simply provide the reminder or additional information to see if that will get the dog back on track.

5.  Take it back to training

Sometimes the error is not simply a matter of genuine mistake, or temporary "brain fade", but the behavior truly has broken down, or it needs more work to become fully fluent.

In those cases, I ignore the error and make a plan to take the behavior back to training.

Example:  At an Agility trial, the dog does not read the handling provided and takes the wrong piece of equipment in a discrimination.  I move on in the course, and then go on later to on discrimination skills in training.

When working on the behavior in training, I will lower criteria, raise the rate of reinforcement, and give the dog ample opportunity to practice the correct behavior.  In the context of training, I will also focus on fluency building and make sure the dog has the opportunity to generalize the behavior adequately. 

What about NRM's?

I realize that many who are +R trainers use NRM's (Non-Reward Markers) to address errors, but it is not my choice to do so.  When working with my dog, it is always my direct intention to communicate "how to be right" to my dog, rather than "that is wrong".

NRM's mark mistakes.  They do not, in themselves, provide information that will assist the dog in choosing to respond to cues correctly.  Therefore, I do not consider the use of NRM's to be in conformity with my approach to training dogs, so I do not use them to handle errors. 

When Management is Needed

There are times, of course, when management is needed.  If the behavior that has broken down is a recall, it is always my choice to work on that behavior in a fenced area, or with the dog on a long line for safety.

If the dog suddenly begins counter surfing, I will keep my counters clean and free of tempting items as the dog is learning a default "four on the floor" to greater fluency.

Generally speaking, management is temporary and is needed only until the training "catches up" and the dog, once again, becomes reliable.  Management is "an ounce of prevention" that helps facilitate the "cure".

A Multi-Dimensional Answer

The answer to this question, "what do you do to deal with errors if you don't use corrections?" cannot really be given in a short answer.  As +R trainers, we have a good many options at our disposal for handling error, and it is up to the trainer to discern the best course of action, given the dog and circumstances at hand.

But I can say this simply, we do not simply "do nothing".  We have effective options at our disposal and we proactively use them.

 

 














 

No Need to be Insulted!

Years ago I had a very close friend who was a vegetarian.  She chose to be a vegetarian because she believed that it was wrong to kill and eat animals.  That was her reason, and she stated it freely.  There was never a need for her to sugar coat it, or try to express it in a way that would not be offensive to meat eaters.

I am a meat eater.  I believe it is perfectly fine to kill and eat animals.  I enjoy eating meat and do so with a clear conscience.  I was able to state that to my vegetarian friend honestly, and there was never a need to try to make it sound like the difference between us did not exist.

I never had a problem with the fact that she believed that it was wrong to kill and eat animals.  I knew that she disagreed with my position as a meat eater, and that she considered my eating choices to be wrong.  And it was perfectly fine with me.

I never presumed that she was accusing me of being an "animal murder" because she did not believe in eating the meat of killed animals.  I never took her choice as some kind of expression of moral superiority.  And I was certainly never insulted when she talked about her own beliefs on the matter.

Why are dog trainers continually insulted these days?  Why do so many presume that any statement of difference in training approach is a personal attack on themselves?  Why the cry of "moral superiority" in the face of different beliefs and choices?

• If one says that he or she chooses not to use training tools that operate through application of an aversive stimulus (some measure of pain or discomfort) to the dog, those who believe that those tools are acceptable get insulted.

• If one says that one trains using the tools and techniques that he or she considers most effective and humane, those who do not use those tools and techniques get insulted.

• If one says that he or she does not incorporate correction into training, those who choose to use correction get insulted.

• If one says that one solves behavior problems through positive reinforcement based training techniques, those who use other means to approach behavior problems get insulted.

• If one points out that he or she does not agree with the techniques and approaches used by a particular well known trainer, those who agree with that trainer get insulted.

I could go on . . .

When did straightforward expression of a different point of view become so offensive?  Why take everything so personally?

I would like to challenge all dog trainers out there to stop taking offense when differences are expressed! 

Certainly, there is a time and place to take offense when one truly is the victim of a personal attack.  By way of example:  "Sally Jo teaches her clients to abuse their dogs" is a personal attack.  "I don't consider the use of shock collars to be humane" is NOT a personal attack!

I would love to see more objective discussion of training philosophies, techniques, approaches, and results.  "I disagree" could so easily replace "that's insulting" or "that's offensive", and I believe that our debates and discussions would start to bring about greater understanding among trainers who hold different points of view.