This morning is another part of the conference I have really been looking forward to -- teaching clients how to use the clicker with their dogs. I have worked with clients here and there over the years who wanted to use the clicker to help their dogs target things, usually chairs or bus stop poles or push buttons for street crossings. Not that I ever knew what I was doing; we just sort of improvised as we went along. Even though in my head I had no idea why it was working, since my clients couldn't click the instant the dog located the target object but instead had to wait until they (the clients) touched the target object physically, nevertheless it did work, even with two people (me and the client) not really knowing what we were doing. I remember one time we used it to get the dog to locate a mid-block bus stop that was marked only by one skinny little pole in the grass four feet from the sidewalk, and it worked like magic. That should have been enough to convince me on its own, but instead I worried too much about the intellectual reasons why it shouldn't work (the timing of the click is really important and the blind handler can't see to click at the exact right second) rather than paying attention to what I actually saw happen in front of me, which was the dog not only indicating the pole perfectly but also walking up to it in such a way as to put the handler's hand right at the pole, like the dog had understood the only way to get a click was to position the handler's hand next to the target object. I have also heard of several guide dog handlers over the years doing other cool things with the clicker. I think I sort of mentally categorized those things as cool but not something that would ever appeal to more than a few guide dog handlers, because the clicker is just one more thing to carry, because clicker training is a whole new thing you have to learn, blah blah blah. Always plenty of reasons not to try something new, right?
At GDB they teach the clients how to use a hand target when teaching their dogs to locate a physical object. Let's say a chair for purposes of this discussion, although it could also be anything else like a mailbox, a trash can, a push button, or any other stationary object. They have a pre-workshop lecture, then a workshop where the actual skill is taught, then a post-workshop lecture. (Is this the right place to say that I wish ALL client skills could be taught this way? For example, a morning explanation of curbs and turns. Then a working session in town on curbs and turns. Then a post-session discussion and Q and A. Maybe not for grads who have already had dogs, but for people who are new? What do you all think?)
From my understanding, the way they teach the practical skill is this: the instructor first teaches the client how to use a hand target, standing just one pace away from the chair. The hand target they use is a closed fist. (Leader puppy raisers teach "touch" with an open hand, and get the dog to nose-bump the palm.) Clients are taught to bring their hand down in an arching movement from about shoulder-height down to target-height, because if they just brought the hand over from the right side, the dog could begin curling around to reach the hand before it ever got to the target. They touch the chair with their closed fist, the hand target, and when they feel the dog's nose touch their hand, they click and deliver the treat right on the chair. The client practices the motion without the dog several times until it is smooth. Then the dog comes out. The client clicks and treats a couple times just to get the dog excited and ready to play the clicker game before starting with the hand target. If the dog doesn't touch the hand, the client is not supposed to move the hand to the dog; rather, he or she is supposed to remove the hand and wait a few seconds to let the dog realize it has missed a chance for reinforcement, and then try again. They do the initial exercise with the dog on leash, and only when the dog is moving toward the chair enthusiastically on leash do they pick up the harness.
I ask why so many reps on leash before harness? It seems like it would be easier to get the dog to pull to the chair when the handler has the handle in hand, since leash = not pulling and harness = pulling. The answer I get is that the reps on leash have to happen so that the dog realizes the chair is not an obstacle that should be avoided. Good answer! Someone else says, though, that if you have a dog who is very reluctant to pull on leash at all (which I do in my string), you can go straight to the harness if you are having difficulty getting drive on leash. I guess that, as with all other training best practices, there are always dogs for whom a different approach will work better. Best practices are great as guidelines, though. I think of them as something that should be followed unless you have a reason for not following them, and can explain and defend that reason.
This is an exercise GDB dogs learn in class, with their new partners. At Leader the puppy raisers teach "Touch", which is a very useful cue that I love for so many different reasons. We teach chair targeting about halfway through training at Leader, but we do it by shaping, waiting for the dog to develop an interest in the chair and shaping that into a nose target. It takes a long time with some dogs, and I wonder what would happen if I were to just put my hand, palm out, on the chair as a visual "Touch" cue, and skip the whole shaping part? Would it speed the process along, or would the Leader Dogs, who have practiced "Touch" their whole lives and generally love doing it, be so focused on my hand that they would not notice the chair? Who knows. Maybe it doesn't matter? Some of my dogs took a while to learn "Chair" but they are now all doing very well with it. In fact, I was really pleased when I took them to Oakland Mall the other day and tried out using the cue "Chair" in a new place and the majority of them started looking from side to side as soon as they heard that word, and then zoomed in to the first chair they saw. The word means something -- I trained them something! (This should not be all that exciting to a professional dog trainer. But the truth is, it never gets any less exciting. This stuff is just neat.)
The next thing we discuss is country work -- working the dog in an area where there is no sidewalk. Guide dogs are generally trained to follow the left edge, or "shoreline" in mobility terms. This is because it is easier for the dog to see oncoming traffic and take action to avoid it if necessary when the traffic is coming towards the dog (as opposed to coming from behind the dog). Also, it is much easier conceptually for a dog to learn to follow a left shoreline than a right one. From the dog's perspective, if it is following a right shoreline, the dog has to both keep the handler from twisting an ankle or bumping anything on the right side, and also stay close-but-not-too-close to the right edge, an abstract concept to the dog when compared to the black-and-white job of the dog keeping itself next to the left edge. We always recommend that clients follow the left edge whenever possible, but sometimes there are clients who have to follow the right edge for some reason (aggressive dogs behind a fence on the left edge, or a very busy road outside the client's house with no controlled intersections). Some of the most difficult working situations I have ever encountered have involved trying to convince dogs to follow a right shoreline when they don't want to. I have come to believe that the best way to do this is to ask really thorough questions in the pre-class phone calls, find out if anyone needs to follow a right shoreline, and make part of the matching process choosing a dog that has no natural objection to walking on the right side of a sidewalkless road. Most dogs, if asked to walk on the "wrong" side, will naturally drift over to the left back where they think they belong.
But I digress. We were talking about how GDB trains dogs to follow the edge. They start with some on-leash patterning in an area where there is no sidewalk and the shoreline is a curb, first with the dog getting clicked multiple times for putting its front feet up on the curb (something it is already familiar with, because of up curbs) and progressing to clicking for walking along the edge and staying close to it. GDB (and every other school I have worked for besides Leader) teaches a turn to the shoreline every so often to make sure you are where you're supposed to be. So as the team is walking along next to the edge, they stop and the handler tells the dog to turn left. The dog makes a 90-degree left turn and puts its feet up on the shoreline edge. I like this method because it gives the client a way to know for sure that the dog is where it's supposed to be. At Seeing Eye we used to tell clients they should check the edge every time they hear a car coming, still a sound guideline in my opinion.
Another part of working in areas without sidewalks is navigating safely around cars parked on the roadside. Ideally, especially in a new guide dog team, the dog should bring the handler up to the car so that the handler can find it by putting out a hand. The handler should then listen for traffic and direct the dog around the car when it sounds safe. Historically, I have never taught this indicating-a-car-and-then-going-around-it as a separate skill from following the shoreline, which is ridiculous because, of course, it is a separate skill. GDB teaches it by backchaining. Probably most people reading this blog know what backchaining is, but in case someone doesn't, let me explain that backchaining is starting with the end point when teaching something new. So, for example, if you're memorizing a poem, you memorize the last line first. Then the second-to-last line, and so on. That way, when you're reciting the poem, it gets easier and easier the further into it you get, and your confidence is highest at the end. To return to the parked car, the dog targeting the shoreline with its front feet is the last line of the poem. The dog coming around the front of the parked car and approaching the shoreline is the second-to-last line. So you start with the feet-up-on-the-shoreline just past the car, and then go a little further back, and a little further back, et cetera. Backchaining is the most effective way to teach any type of targeting skill. Of course it was used in guide dog training long before clicker and treats, we just reinforced the end destination with praise rather than treats, but the clicker gives a turbo boost to the method.
One caution that was pointed out to us was that once the initial work of targeting the shoreline was done, we should switch to clicking while the dog was in motion and in good position right next to the shoreline, rather than expending too many clicks on targeting the shoreline. If we're not careful about that, we can create a dog that stops and turns to the curb too often, without being prompted. I believe this has been a concern over the years when the subject of teaching country work with the clicker was discussed. Trainers were always afraid they would create a dog that stopped without being asked to and turned to the edge where it had been most heavily reinforced. This conference and these demos have given me more and more confidence every day that what the behavior that you click is the behavior you get more of. I know, I know, all the clicker training books, of which I have read dozens, say the same thing. But it is so easy for us as GDMI's to say, "Yeah, for pet dogs, but what we do is different; we can't take the risk that we'll get some other behavior that interferes with safe guide work instead of the behavior we think we're clicking." The answer to that objection, of course, is that all of us need to seek out every opportunity we can to get better with the clicker, both in understanding it theoretically and in hard skills -- hands-on practice. Hey, what are our pet dogs for if not for experimentation with the clicker?
Anyway, at Leader we just now officially switched to teaching country work with the clicker. Just this past week I trained it to my dogs, and, wow. It is completely different than training it without the clicker. Historically, most instructors have hated teaching country work. It has always been tedious and there have always been some dogs to whom it just doesn't make sense. They don't understand why they get corrected when they drift away from the edge, they don't understand why they have to indent (turn down an intersecting street to indicate an intersection), they drop their tails and get a sullen look on their faces as soon as they get out of the truck and recognize the country work environment. It's like instructor and dog both go, "Oh, God, not this again." It is night and day training this traditionally and training it with a clicker. Even my one dog in this string that has a little bit of a right tendency was glued to the shoreline by his second trip out there. I am grateful for every single step in the process that culminated in me being able to use the clicker to train country work to my dogs at Leader, believe me!
After country work, we discussed back chaining in general, for targeting things other than chairs, like poles and trash cans. Of course we have really been talking about back chaining all day. Clients teaching chair targeting was back chaining (start right at the chair and gradually get further away), and so was teaching the dogs to work around parked cars when doing country work. This discussion was more of a review and wrap-up of back chaining principles.
When GDMI's are teaching something using back chaining, it is important for us to observe and click for intention. A lot of times, when the dog is first learning to target something, it makes very subtle movements, like a sideways glance at the target that lasts less than a second, or a slight head turn towards it, and if we are watching closely enough to click that, it will make the dog understand that much quicker. When clients are doing a back chaining exercise, it becomes a tactile exercise. The dog gets clicked when the client touches the target. Even though the timing isn't that precise, the dog has a better understanding of the game -- because it has played the game many times with instructors already -- so the timing isn't quite as important. The dog already has a mental category for the game the client is playing with it -- "Oh! That's the show-me-that-thing game again. Got it." GDB emphasized getting the behavior on leash first. When the dog is consistently anticipating and moving toward the target, then you can stop using the hand target and use the harness instead. When the behavior is strong in harness, then you can add the verbal cue -- "Chair" or "Trash" or whatever -- as the dog is in motion toward the target.
We as humans are so lucky to have dogs to work with. Dogs are so willing to interact with us and do what we want them to do that even when our timing is bad, they still understand what we want most of the time, and do it. I don't know if I will ever cease marveling at how lucky we are to have them.