RAISING THE BAR
Originally published August 2013
As dog trainers, we have both hopes and expectations for our dogs when we train them. The problem I am seeing is that people are satisfied if Fido is willing to sit after several prompts, a treat on the nose and the cue repeated fifteen times in a row. This is not training, this is nagging. Just because training is done with treats involved doesn’t mean that the training is pleasant or humane; it just means that there are treats involved. I recently had a client demonstrate to me his dog’s skills after a week of practicing what they had done in their previous classes at an another skill. Sit took six verbal cues, pulling up on the leash, stamping his foot and then finally pushing down on the dog’s hind end. I wish this were an isolated case, but I see at least one of these dogs each and every week. My client may want advanced standing in a class because they have already trained their dog to do things either at home or at a different school, but when the time comes to demonstrate what their dog will do it is a series of prompts, cues, commands, hand clapping, finger snapping, food waving and leash popping with a reluctant dog on the end of the leash.
So just what do I mean by a well trained dog? What would that look like from the perspective of a dog trainer? Dr. Ian Dunbar devised a sit test some years back. What it looked like was the dog on a leash and the handler standing in the vicinity of the dog and asking him one time to sit, with one cue. If the dog sat successfully, then the dog and handler could go to the next test. The next test involved the handler sitting on a chair, and asking the dog with one cue to sit. If the dog sat they could go on to a more difficult test, with the handler doing something else. The point of the exercise is to determine if the dog actually understands the cue to sit, without a whole lot of prompting and cuing and showing the treat and jumping up and down and hiring a brass band and elephants just to get the dog to sit. You can (and should!) try this with your own dog. Will your dog sit on one cue, without a treat being waved in his face, when you are: standing in front of him? Standing beside him? Sitting in a chair? Standing on a stool? Waving your arms like a chicken? Holding a bag? Holding a toy? Climbing a ladder? Lying on the couch? Sitting on the floor? Lying on the floor? Standing with your back to your dog? Once you have a dog who will sit through all of these machinations, can you repeat the performance out of doors? At training class? In a busy park? In the presence of loose dogs? Beside a fountain? On a ski hill? At a coffee shop? Beside a busy road? When a fire truck goes by? Is there anywhere where your dog won’t sit if you ask politely?
This test is an interesting measure of the trainer’s ability to help their dog to learn effectively. Continuous prompting is actually a form of negative reinforcement because you stop nagging your dog when he finally gets the right answer. Negative reinforcement followed by positive reinforcement is one of the most powerful methods of training going, but when you are doing this with care it should be an elegant dance not a blunt force instrument of annoyance to you, your dog and the people around you. It is not fun to have a person who repeats themselves in your ear over and over again in an attempt to get the attention of a child or a dog. It is frustrating!
What we need to do with our dogs is raise the bar in both our training and our expectations. Consider for a moment when a toddler learns the first letter of his or her name; mommy and daddy exclaim over their brilliant little scholar and make a big fuss. The first letter IS a big deal! When that toddler is 21 and graduating from college, then knowing the first letter of his name should not elicit the attention of his parents in the same way. Far too often this is what I see with dogs after training. The family comes to training to learn about behaviours from their dogs, and at the end of the class, the dog has learned to sit when a treat is presented or they may even follow a treat into the desired position. Two years later, if you are still treating each and every sit, you are doing just what the parents of the toddler who congratulate their college grad on knowing his letters did.
I am partnered with a service dog. Right now, I am in the middle of changing over from my old service dog D’fer, to my new service dog, Friday. D’fer has passed the sit test, every day, every where, for the past eight years. Friday isn’t as experienced as Deef, and thus doesn’t have the number of always behaviours that he does, and couldn’t pass the sit test in as many different places. Can he do it on an airplane? Yup. On a train? Yes. How about on a merry-go-round, moving slowly? Yes he can. How about if I am on a hospital bed? Yes again. And how often do I reward him for sitting? From time to time. I don’t keep track. If he wants to have something and we are out and about, he may sit out of context to get my attention and ask. Once we were walking through a show ground, and there was a bitch in heat. I stopped to look at something, and he sat and stared at her. As I started to move away, he held his sit to get my attention. When he got my attention, I looked down at his longing face and told him “sorry Dude, not this time.” Not only did he not get a reward for sitting, I in fact penalized him for offering the behaviour I had taught him to do to get my attention. Why then would he keep sitting when asked? This doesn’t seem like it should fit in with Learning Theory.
So how do you get from the point of using positive reinforcement to teach a particular behaviour, to the point that a service dog is at where they will do the desired behaviour anytime, anywhere and will in fact offer the behaviour you want even when it costs them to do so? Quite simply, you raise the bar. There are several stages in learning and the key to success is to neither stay in one stage too long nor to expect the dog to be able to do behaviours that are above his skill level, but to keep raising that bar!
I like the model of awareness, acquisition, application and always to describe where a dog is when they are learning a behaviour. If the dog has no idea what will result in a reinforcer, then you have to start by making them aware of what behaviour specifically will earn a reinforcer. It doesn’t matter if the trainer is using shaping or luring or prompting or modelling, if the dog doesn’t know what gets the reinforcer then you will not be able to be successful until your learner is aware of what the target behaviour is. In some respects we could say this is the second stage in that the trainer has to also have an awareness of what you are trying to teach to begin with. Once the learner and the trainer are both aware of what the target behaviour is, then the learner can move towards acquisition. There is no clear line between awareness and acquisition; they are part of a learning continuum. The learner has passed through awareness and into acquisition when the behaviour is being repeated, but it may not be an accurate iteration of the behaviour and it may not be a fast or flashy or finished iteration, but you can see the structure underneath the behaviour when the learner is in the acquisition phase, which you probably couldn’t while the learner was in the awareness stage.
This dog and trainer are in the awareness phase of training the sit. The trainer knows what she wants, and the dog is getting the idea. Image Credit: Sue Alexander
It is during acquisition that I see people start to abuse the trust of their learner. Let’s go back to sit as an example. If the dog has figured out that sitting gets the reinforcer, and then the trainer and learner have practiced this in two places with no other dogs, then if the trainer asks for the sit in the dog park with thirty other dogs around, it doesn’t matter how good the treat is, the likelihood is that the dog won’t be able to give you the behaviour you want, quickly and efficiently. At about this point, the novice trainer comes to me and says “with our last dog, we would have used her prong collar to make her sit, because she obviously knows what the command is, and she is being disobedient.” I would argue that this is putting the dog in a completely unfair situation. Think back to when you were learning to do fractions in grade school. Can you imagine what it might have been like to be asked to do fractions on the playground, at recess when all your friends were playing on the swings? THAT is what it is like for a dog who has just learned sit in the living room to be asked to sit in the middle of the dog park. Add a painful consequence to that mix and pretty soon you are going to have a dog who just doesn’t want to participate any more. At this point the trainer gets frustrated because the dog stops sitting altogether and the trainer thinks that the dog knew the behaviour.
The other end of this spectrum is the trainer who over does the “proofing” level of the training. They teach the dog that treats come for sitting at home, at the training hall, on the driveway, on the front lawn, on the back deck, in the lobby of the apartment building, in the presence of a known dog, in the presence of an unknown dog, at the park, at the vet’s, at the kennel and so on and on and on and on. These trainers pay for each and every sit, each and every time and wait for the behaviour forever and ever and ever. Eventually for these dogs, the sit decays too, but for a different reason. In this case it is a bit like ordering in a restaurant. The dog surveys what the reinforcer of the day is, and if it isn’t something that he wants, then he doesn’t sit. Sit for kibble? Thanks, no. How about sit for a bit of dried liver? Not today. Sit for dried salmon? Nope. Sit for garlic toast? Oh…okay.
The key to moving from acquisition to the next stage of application is to take into account a number of factors, and then train for those factors. The first thing is to make sure that the dog has actually acquired the behaviour in the context of a number of different scenarios. Different people in different locations for different reinforcers should be part of the program. Once the dog is offering the target behaviour reliably for a variety of people in a variety of places for a variety of reinforcers, then I would start putting the behaviour into chains. Your dog has to have more than one behaviour on cue for this to work and at the beginning this means that there will be several behaviours in the acquisition phase in the chain but this will still work nicely. Let’s say that you have sit, down and touch all on cue for three different people in four different places using kibble, a Frisbee, a ball and liver bits. Then I would start on chains of two behaviours. Since I use a marker to train, it might look like this: Sit, touch, click, liver bit. Touch, down, click, throw a ball. Down, sit, click, kibble. I would practice at two behaviours for a while until the dog was reliably following my cues, in a variety of places. I have a plan for when we get stuck; let’s say that the dog stopped following the sit cue for instance. I would manipulate the chains so that the sit became the terminal behaviour, meaning that is always followed by a click and a treat of some sort. Then it might look like this: Touch, sit, click, liver. Down, sit, click, Frisbee. With only three behaviours it becomes a little bit tricky for the trainer to not become predictable, but at first this is how I start to make a weak behaviour stronger.
In the application stage of learning, the dog learns that she can sit anywhere, anytime for a variety of reinforcers. This dog is on the crest of being in the always stage of the sit because she will sit out of doors, on a ledge, beside a fountain, with traffic moving around, for a variety of handlers and a variety of reinforcers. Image Credit: Sue Alexander
As the dog is able to integrate a behaviour into a chain of two behaviours, with it not necessarily being the terminal behaviour, I move to three behaviours for one click and treat. Once we can do three I start playing a game called three hundred chickens where I ask for increasing numbers of behaviours, increasing by one behaviour for each iteration, for each click and treat. It might look like this: Sit, click, liver bit. Down, touch, click, Frisbee. Touch, down, sit, click, throw ball. Sit, down, sit, touch, click, liver bit. Down, touch, touch, down, touch, click, kibble. The goal of course is to get to three hundred behaviours per click/treat. If we get stuck, we go back to a significantly lower number of behaviours. If for instance we were at 32 behaviours and the dog refuses a cue, I would prevent the dog from leaving and doing his own thing until he did that final behaviour, and then start at one behaviour, then 5, 10 and so on until I approached where I was when we hit failure. In doing this, I create a variable schedule of reinforcement that is in line with the learner’s experience. This stage of application where the learner is building repetitions into his repertoire, in a variety of places, in chains of behaviours, with different handlers and for different reinforcers is what gets you from the training stage of learning to the stage that a service dog works at. This is the neglected stage and this is where many trainers just stop training and either settle for behaviours that only occur when a threat is imminent or when a valuable enough reward is available.
Once the dog has learned that he can integrate a new target behaviour into his repertoire in a variety of places, he is ready for the always stage. In the always stage of a behaviour, it just always happens when it should. It probably gets reinforced from time to time, but often that is incidental; it often doesn’t get reinforced with any sort of a plan. Very few behaviours get to the always stage for most dogs. Taking Eco, my German Shepherd, as an example, he has relatively few always behaviours. He is extensively trained, and mostly through the use of R+, however he has a degenerative back disease of unknown origin that prevents him from travelling by car or doing training often. His “always” behaviours are the ones that we use all the time. He stops at open doors and won’t go through them unless told. He waits to take his food at mealtimes. He goes into his crate when he is asked. He reliably comes when called. He accepts any handling I have to do. He takes his pills every day. Apart from that, he is pretty rusty on things like heeling, sit stays (probably hard on his back, so we don’t really ask for a whole lot of sits from him unless related to medical treatments), long down stays (he probably gets uncomfortable in any event!), fetching, going around things, pushing a door or a ball and other tricks or obedience trial behaviours.
Eco used to do a lot of training and he knows a lot of behaviours, but they are not always behaviours because of his back injury. We cannot practice often enough to get past the stage of application because he would get hurt. Image Credit: Melanie Wooley.
D’fer, my 10 year old Chesapeake is a little different. After eight years making his living as a service dog he has rock solid leash manners, a rock solid sit, a rock solid down stay, and he is always willing to alert to an oncoming medical event. He will retrieve pretty much anything I ask, anytime, anywhere and he will take things to other people for me. He has a lot of always behaviours. So how do you get from application, where the dog learns to apply the behaviours across a wide variety of conditions on a variable schedule or reinforcement, to always, where you can ask for an expect the behaviour 95% of the time or more often than that?
Picking up his leash and handing it to me is an “always” behaviour for D’fer. I don’t have to reinforce it very often because we have taught it carefully and systematically. Often, D’fer will pass my leash right before we do things he likes doing, so co operating with me will result in activities he likes, reinforcing D’fer’s co operation with my request.
Once again, the boundary between application, where the dog will give the behaviour in a wide variety of environments, for a wide variety of handlers, for a very thin and variable schedule of reinforcements, then you are ready to transition into the always stage. In the always stage, I will insist. You don’t have to be heavy handed to insist. It is a little bit like asking a guest to do something your way. You have to be polite, but some things are not negotiable. If you visit my barn, I insist that you don’t smoke in the barn. I am polite, but I insist that if you are going to smoke, you take that outside. If you have the nerve to break that rule, I will ask you once and then I will take your cigarette out of your hand and put it outside. Smoking in my barn is dangerous to me and my horses and I will insist on you doing that out of doors only.
There are too many risks of fire to allow smoking in my barn. If you come to my farm, and you pull out a smoke in my barn, I will ask you politely to take it outside. If you don’t I will insist. If you still don’t then I will eventually be quite insistent and I might even escort you out of the barn. In general, if you know how to get out of my barn, and you understand that smoking in a barn could result in a fire, you will leave if reminded. If you don’t understand the reasons that I ask you to take it out of the barn, you might protest, but when I lay it out, most people are willing to leave the barn. The same is true of our dogs who really understand the behaviours you are training. Once the dog understands what you want and has practiced the behaviour in a variety of contexts, then you can insist on the behaviours when you need them. Image credit: iofoto / 123RF Stock Photo
Let’s consider the sit behaviour again. If I have been working with a dog daily for 8 weeks, as sometimes happens when we board and train dogs, they will have passed through the awareness, acquisition and application stages of sitting. And then one day, out of the blue, I will ask for a sit that is perhaps inconvenient or out of context for the dog. The dog may look at me as though I am out of my mind. He may just stand there. If he tries to wait me out, after about a minute, I will take a step away and call him in and ask him a second time. If he sits, then on we go with whatever we were working on. At the point of always, I only reinforce the perfect sits, and only occasionally, but never ever when I have had to ask twice. What I do at this point is switch from positive reinforcement to negative reinforcement. If you don’t do as I ask the first time, I am going to change the picture, and ask again, so you may as well get on with it. Like asking a guest to take their smoke out of my barn, the first time I ask, and the second time I insist, and I keep insisting until the outcome is what I want. Dogs who have a solid foundation on the previous three stages understand the deal. They get reinforcers through the day, for things that are difficult or things that are new or things that are stellar. They get positively reinforced for things they know well once in a while, but for things that my dogs really understand, they don’t get positively reinforced, because they know it well.
Negatively reinforcing behaviours yields very strong and resilient results because learners know that they can avoid annoyance by doing what is asked of them. Also, when “always” behaviours have strong positive reinforcement histories, the behaviours themselves become reinforcing. And finally, always behaviours pop up in places that lead to other behaviours in chains that lead to reinforcement. If I am waiting in line in a coffee shop and I ask D’fer to sit, then after I order my coffee, we will often go for a walk so that D’fer can toilet. Sitting leads to doing something he wants to do for its own sake. This is how I get most of my always behaviours.
The key to being successful with “always” behaviours is to ensure that you don’t ask for them before the dog is ready for them, but that once the dog knows the behaviour, you don’t stay in the acquisition or application stage beyond the time when you should. A word about punishment here. Punishment is anything that is done to DECREASE the frequency of a behaviour. Many of my clients want to reach for punishment when the dog doesn’t comply to a behaviour that they consider to be an “always” behaviour, as in “if he doesn’t do as I ask, there is an unpleasant consequence to that choice”. The problem with this is that the target behaviour is still the target behaviour, and if you use an aversive in this case, then you aren’t punishing non compliance. You are punishing whatever the dog was doing in the moment that you used the aversive. Let’s say that you have asked the dog to sit, and the dog doesn’t comply. What he DOES do though is to look longingly and adoringly in your eyes. You pull out a water pistol and you spray him in the face, because he didn’t sit when you asked. What behaviour are you decreasing in this case? If you said “looking longingly and adoringly in your eyes” you would be correct. If you said “not sitting” you would not be correct because you didn’t decrease the not sitting behaviour; you decreased something other than not sitting. When you are talking about an always behaviour, resetting the picture and asking again until you get the behaviour you want is going to yield a much stronger result.
The true value of building “always” behaviours is a deep partnership between you and your dog. You can depend on your dog’s behaviour and your dog can depend on you. Image credit: chalabala / 123RF Stock Photo
At the end of the day, when you have “always” behaviours you have a dog who is your partner. You can rely on your dog to do his always behaviours when you ask them, and your dog can rely on you to not ask for those behaviours just to show off or to prove that you can make your dog do what you ask. Always behaviours are behaviours that you have built together with your dog and that deepen your relationship with one another. Always behaviours are the result of careful raising of the bar.
