Originally Published November 2013
Lately a few of my students have been struggling with very intelligent, high drive dogs who enjoy pushing their person’s buttons. If you say sit, they lie down. If you ask for got to mat, they offer a trick. If you send them to fetch up a dumbbell, they go out, turn and face you and sit. If you leave them on a down stay they bark. When I work with these dogs, I can often turn them around quickly and get the desired behaviours, often to the frustration of their people. What am I doing that their handlers are not, and why can I turn this around when other trainers cannot? I will start with the caveat that I am NOT magic. I just have a few pieces to the puzzle that my students often don’t have.
The key factor that I have is a better understanding of how learning works. When we read training books, it often sounds like all learning is linear and builds upon itself. Training plans are often presented like knitting patterns, where if you follow all the steps and you use the right wool and needles, you will get a predictable outcome every single time. While this is somewhat true, there are a few things that we have to take into account when working on learning.
|When you are knitting a scarf from a pattern you can put the project down and nothing will likely change. There are not knitting gnomes who come along and unravel your work or add extra bits when you are not looking. This is not true of training. When you are training your dog, a lot of things can change from session to session or even within a training session that can effect the outcome of your training session. Image credit: dole / 123RF Stock Photo|
The first thing to take into account is that learning is only linear when the skill set upon which learning is built is solid to begin with. If the dog has a fuzzy idea of the foundational skills that are needed for an advanced version of the same behaviour, then he is not going to be successful at the advanced version because he doesn’t have the background to understand the more difficult work. Think back to learning how to read; if you gave this blog to a first grader they would be hopelessly muddled in multisyllabic words, clauses within sentences and my sometimes non linear route to an idea. They don’t have the foundational skills necessary to follow along on the sentences that build upon “see Spot, see Spot run!”
Next I also understand that success is built on success. When you are working with a dog, starting new work cold without doing a few things he is good at means that he is not starting from a point where he has gotten a few answers right. When I am teaching something new, I usually start by getting the dog to do something he knows well for a few repetitions, and then do the foundational work with him and then start stretching his brain with new work. In order for learning to be a fluid process, you have to start from where the dog is at when you start the training session, not from where he was the last time you worked on the behaviour. Learning is not like knitting that can be put down and picked up and carried on from one session to another. Learning is a much more fluid and dynamic process and in between training sessions, learning is still happening. Sometimes that learning may impede the next session, and other times it may help that session to move forward.
Yesterday one of my staff came over to learn more about horses and riding. He has had a couple of ground lessons with Kiki, John’s mare (whose real name is MacKenzie), and has learned to groom her, to lead her and to mount and dismount and ride on the lunge line fairly successfully. Yesterday I asked him to lead her unsupervised through a field to where we were going to work. He got about fifty metres into the 200 metre walk and Kiki put her head down to graze. Kiki is very committed to grazing. She loves grazing. She also knows how to walk politely for a stranger. I coached him through the final 150 metres and he eventually got her to the riding area. Then I took the lead rope and asked her to step forward. After twenty minutes of starting and stopping and grazing and getting reinforced for fiddling around instead of walking forward, she was baulky and difficult to handle. Keep in mind that for me, I can walk her through a hay mow with feed pans of grain on the ground and she normally would not baulk. Never the less she gave me a very hard time about moving forward. Why would she do this? Because between the last time I walked her somewhere and this time, she had learned that stopping to graze might work and that by planting her feet she could get more opportunities to graze. The twenty minutes that she spent with a novice handler unravelled some of the training we have already done and I had to go back and review some of the “walk beside me without pressure on the lead rope” lessons we have already done! Often in class, the work that a student is doing may be interrupted by what they have learned between lessons.
Another thing that can happen is that behaviours that we don’t want get inadvertently reinforced. I have a number of students who click after the behaviour has occurred. When they are working on simple foundation behaviours this can actually help them to have early success because they don’t click until the dog is firmly in the sit or firmly in the down. When you are trying to train very precise behaviours though, late clicking means that you are clicking for things you don’t want and the dog develops the habit of fidgetting or fiddling with the behaviours until the handler clicks. In an extreme situation what you end up with is a dog who thinks that your target behaviour is not what you asked for, but a whole string of behaviours that terminate with the asked for behaviour. I watched a dog this week when asked to sit go through the following string of behaviours; he sat, lay down, crawled backwards, sat again, stood, barked, and then sat and held the sit for a micro second. Then the handler clicked. This dog is very, very smart and he is pretty sure that the behaviour his person wants when she asks for sit is this long chain, because he is reliably being clicked for it. Not only that, but he seems to be reinforced by his handler’s frustration. The more frustrated she gets, the more fidgety he gets when she asks for simple things. He is a brilliant dog, but very frustrating to work with because when he doesn’t get feedback in the right place, he throws in all sorts of embellishments. While embellishments might make for a nice sweater, they can make for a very frustrating teaching experience when the trainer has an idea in mind for what behaviour they need the dog to learn.
When I watch this particular team, I am reminded of a skit I saw once of a couple at a restaurant. They are served their meal and when the waiter returns to ask what they think of the food, they hesitate, so he whips their plates away and replaces them with a “better” plate of food. When he returns and asks again, the diners are not fast enough to give him feed back so he takes their plates away again and brings them back with “improvements”. This happens six or seven times, the chef is brought out, the maitre d’ and finally the manager and the closing comment is “some people….you can never please them!”
|When trainers don’t have a clear target behaviour, I think this is how some dogs must feel; nothing they do will make the trainer happy. Like the wait staff at a restaurant where the customer gives little or no feedback, the dog keeps trying new things to make the client or in this case the trainer happy. Image credit: lisafx / 123RF Stock Photo|
I think that when the dog is not reinforced at the right time, he must be left feeling the way these diners did; like nothing he can do will meet the needs of his handler. This is a bright dog who needs immediate feedback for each and every iteration until he understands what is wanted. This particular dog also tends to go through a long and protracted extinction burst when he finally understands what you want. This can be endlessly frustrating for handlers. I watched him working on item discrimination recently. He really knew the item that they were working on. When put in a pile of four other items, he reliably chose the wrong item, looked at his handler, and then repeated the drill. Sometimes he would go to the other items ten, twelve or even fifteen times before he picked the right item. This is a dog who really likes to be sure! An extinction burst is a phase of learning that dogs go through that can drive trainers nearly screwy. What happens is once the learner understands the criteria, then they go through a period of not giving you what you ask for, over and over again. This is important because of what is happening in the brain. It implies that the brain is not only collecting data on what earns the click, but also that it is collecting data on what doesn’t earn the click and that information may be every bit as important to the learner as what does earn the click. While this may feel like your learner is unravelling the knitting, in reality, they are just adding some interesting pieces to the pattern.
The final issue I see regularly in class that impedes learning is when the trainer hasn’t determined a target behaviour for the dog. If the trainer has a “sort of, kind of” definition in mind, then the dog is going to give you “a sort of, kind of” performance. When the dog knows the behaviour and they are embellishing it, the fastest and easiest way to overcome that is to take a moment to clarify in your own mind what behaviour you want and then train THAT. Getting upset that the dog may have known the behaviour at home, or may have known it yesterday isn’t going to help. Likely there is a good reason for the dog fiddling around with the behaviour, but the reason doesn’t change the fact that you have to reknit when you the behaviour has decayed. When Kiki wouldn’t walk on, I could have been angry and frustrated, but she is a smart horse and my frustration is just an indication to her that she can keep her feet still and drop her head and sneak in a bit of grazing while I am busy losing my cool, which brings me to my final point.
Dogs are the masters of picking up cues. They know when you put your running shoes on to dash out to the car and retrieve your paperwork so that you can call your insurance agent that a walk is not imminent. They also know when you put on your running shoes in preparation for a walk. It can be very baffling to the trainer how the dog knows the difference when both cases involve getting the shoes out, sitting down on the bench in the hall, putting the left shoe on before the right, tying the laces, and then standing up and reaching for your keys and walking out the door. When you are frustrated with your dog because he has yet again undone your knitting, he can tell when you are just going to deal with the behaviours you have in front of you and when you are going to actually deal with those behaviours. He has determined the cues that precede actual training with the opportunity to play with you. Training is a mirror image activity. While you are training your dog to carry out specific skills and activities, he is training you. If you are unclear on your target behaviour for the session, don’t be surprised if he is really, really clear on his target behaviour. Every time you reinforce your dog for a correct answer, he is busy training you to give him reinforcements himself. If sitting makes you throw the frisbee, then throwing the frisbee makes him sit; it really is a closed loop where you and the dog have an agreement about two behaviours you are willing to exchange. The more advanced your training, the more elegant the interaction becomes. When your dog enjoys your frustration behaviours (and yes, some dogs do!) then you are exchanging frustration behaviours on your part with your dog’s fiddling around and unravelling the knitting. Having a rock solid picture of the target behaviour allows you to change the cycle and come to a different agreement, and ultimately to complete the knitting project you have taken on with your dog when you train.