Rewarding Mistakes
I often reward a dog that has made an error in training. This is seriously counter intuitive, since most of us know that the way to increase a behavior is to reward it, and the way to decrease a behavior is to ignore or punish it. So why on earth would I reward a dog that has performed in a manner that I do not wish to see repeated?
To preserve attitude.
There are several instances when rewarding a mistake is a good decision. Here are a few examples:
The handler has created the problem behavior. For example, the handler routinely drifts into their dog while heeling, causing the dog to develop a habit of heeling wide. Because the dog has been trained to heel wide, it is likely that he believes heeling wide is correct. If you suddenly withhold rewards before the dog has learned the new heeling position then they are effectively being punished – which threatens attitude. Instead, retrain the handler and allow the dog to adopt the new position naturally by helping the dog, incrementally raising criteria, and rewarding in position.
The dog believes she has performed correctly. Lyra sometimes returns around the high jump with her dumbbell. She comes back around the jump at full speed, clearly believing that she has done a wonderful thing. She does not yet understand that she must always return over the jump - even when I throw poorly or she is moving at speed. Lyra has not learned the criteria - she cannot meet criteria if she does not know what they are. Under these circumstances, I reward her for the speed and enthusiasm she shows, and simultaneously I make a mental note to train the return over high jump so that she will be correct in the future.
The dog lacks confidence or shuts down easily. I am focused on keeping the dog “in the game” – if a fragile dog senses any disapproval then they tend to shut down or go into avoidance. Dogs that have shut down are impossible to train so I avoid it at all costs. That means I may reward an error to keep the dog working for me. Dogs that are lacking in confidence need to be trained in a manner that allows them to be right almost all of the time – learning problems must be broken down into tiny pieces. Of course, this is the best way to train all dogs, but with fragile dogs it is essential.
The dog misreads the handler's cue. Here's an example: in the process of giving a “sit” hand signal, the trainer brings their hand too high in the air, effectively giving the hand signal for “down” as their hand returns to their side. (this is quite common; watch yourself in the mirror if your dog has a habit of downing immediately after sitting). If the dog goes down, I will reward the behavior – I’d rather have the dog remain confident in their abilities (and fix the handler) than have the dog become a slow or passive workers in order to avoid mistakes.
My rule of thumb is to prioritize attitude over accuracy; it’s pretty easy to get accuracy in a dog that wants to be in the game with you. Once your dog has plenty of attitude and is as excited about work as you are, then you can begin to raise the criteria for rewards. In obedience this approach is somewhat novel because speed and attitude are not judged in competition, but some of us choose to train for it anyway. You'll have a lot more fun and your working career will last longer.
16 comments
Another example, in agility, is continuing on in spite of a run by or off course, especially in young dogs, or even celebrating it! Much of the time it is handler error, and the rest of the time it is trainer error ;) so I just keep going, which is fun and rewarding to the dog.
Thank you for continuing to post thought provoking and helpful topics!