Dave and I have been taking turns blogging about Tajji over on Book View Cafe. For those folks following the saga of rehabilitating a retired seeing eye dog here, here's the latest entry.
I suppose everyone has had the experience of looking into an animal’s
eyes and wondering what thoughts were passing through its mind, or how
it was experiencing the moment. When learning to train an animal, the
wonder becomes tinged with frustration, as one realizes how poorly one
hears the animal even when it’s frantically semaphoring its thoughts and
emotions. What chance of sensing its internal life, then, when the
animal is quiet?
Looking back from the other end of the leash, such uncertainty disappears. As the result of hundreds of centuries of co-evolution with mankind, Tajji’s mirror neurons give her a pretty good model of my internal life, as far as her purposes are concerned. And what she lacks in empathy, she makes up for with the startling speed with which she learns the rituals of play or training or everyday life.
Paradoxically, the combination of two highly desirable canine traits, empathy and trainability, yields a dog that can actually be more difficult to train. For instance, two of the drills we do with her daily if possible are the “name game” and “puppy in the middle.” In the name game, you throw a piece of food on the ground in front of the dog. As soon as the dog is finished eating, but before it lifts its head, you call its name, then mark and reward any turn of the head towards you. In “puppy in the middle,” two people take turns calling the dog away from the other partner, rewarding in the same fashion for a head turn or, hopefully soon, the dog’s return.
In both games, Tajji rapidly began anticipating the next move, so that it became difficult to mark the exact moment for reward. However, in yesterday’s class, we found out that doesn’t matter. Just keep marking and rewarding the turning as best you can, and you will still accomplish the purpose of both games: inscribing deeply in muscle and emotions a strong value to the dog’s name so that you can get it looking back at you even when a trigger—another dog, a person, a noise—pushes it over its threshold into the unthinking depths of limbic behavior.
Other foundational games include:
Touch: touch and eventually follow a hand not holding a reward. A good relaxation move (if the dog is reaching to touch your hand, it’s not looking at the trigger) and a wonderful way to move the dog around when teaching more complex behaviors. Interestingly, Tajji quickly learned or relearned Touch to one side or the other, but it is still, weeks later, difficult to get her to touch a hand held over her head.
Eye contact: Show a treat in each hand to the dog and then hold your hands out to the side, shoulder height. Wait for the dog to look at you, mark and reward with the treats. At first the look will be brief, and probably not include eye contact. Eventually you’ll be able to demand a long gaze, even when you’re jiggling your arms and hands around.
Puppy zen: get the dog’s attention with a treat held over its head, then lower the treat slowly, stopping or even removing the treat if the dog reaches or lunges for it. Eventually, you’ll be able to ask for and get total stillness until the treat is at its mouth; at first, you’ll need to reward in a long swoop to prevent failure.
Counter-conditioning is an important tool for desensitizing a reactive dog. In yesterday’s class we started counter-conditioning against Tajji’s reactivity to other dogs. We brought her out to about 100 feet from a pair of blinds from behind which a helper could select and show various size fake dogs. We wondered if Tajji would just ignore the fakes, but we needn’t have worried. She was already amped up for some reason, and caught a glimpse of one of the dog dummies on the way across the field (poor situational awareness on my part), which triggered her just fine. We calmed her down with some T-touch (basically, mindful massage along the lines of “don’t pet the kitty the wrong way,” which, it turns out, is excellent advice for all animals) as well as most of the exercises above.
During this phase, Sandy pointed out how some of my training moves were working against me. I wasn’t noticing when she liked an exercise so much that it raised her excitement. In such exercises, I was told to move more slowly, and allow more time between trials to let Tajji “cool down.” Deborah pointed out later than I tended to get wound up, and raise the pitch of my voice, which wound Tajji up as well.
Finally we began the counter-conditioning. Using radios and hand signals for communication, we introduced Tajji to 3 different dog dummies of varying realism. All were quite triggering at first—once we got her to look in the right direction. That was only a problem the first time; after that she knew where the evil dogs were coming from. Then we used her alertness to train the “look-at-that” command, another foundational and somewhat more advanced game. Fortunately, dogs are one of the very, very few non-human animals who understand pointing, and I believe they are the only animals in which this ability is easily evoked. That seemed to work well.
The counter-conditioning was simple. As soon as I could tell Tajji was looking at the other dog, I started shoving high-value treats into her mouth. It was immediately apparent that she wasn’t too aroused to work, for she ate each time. (Not eating is almost always a sign of being over threshold and therefore incapable of learning.) As soon as I got into the game myself, re-familiarizing myself with her extremely subtle signals—sometimes merely something about the set of her head seen from above), I was able to reward quickly enough to make some real progress.
Next week we’ll be working with a real dog: one of Sandy’s. For Tajji, this will be a real challenge, for Piper is a small terrier of some sort, and Tajji apparently loathes small dogs. (I suppose there’s nothing like being ankle bitten while in your seeing eye harness and unable to defend yourself to condition a deep and abiding hatred of tiny yappers.) In the meantime we’ll work hard on the foundational behaviors, and see if we can figure out a way to continue the counter-conditioning. Like as not we’ll work with the sounds she triggers on. That happens a bit too often, and we want to terminate any rehearsals as quickly as possible, for practice perfects bad behaviors as much as it does the good ones, perhaps a bit more.
* “Swamp collie is a joking pejorative sometimes applied to German Shepherd Dogs by fanciers of other breeds. Funny thing is, most GSDs detest water.
The intrepid "swamp collie"* |
Looking back from the other end of the leash, such uncertainty disappears. As the result of hundreds of centuries of co-evolution with mankind, Tajji’s mirror neurons give her a pretty good model of my internal life, as far as her purposes are concerned. And what she lacks in empathy, she makes up for with the startling speed with which she learns the rituals of play or training or everyday life.
Paradoxically, the combination of two highly desirable canine traits, empathy and trainability, yields a dog that can actually be more difficult to train. For instance, two of the drills we do with her daily if possible are the “name game” and “puppy in the middle.” In the name game, you throw a piece of food on the ground in front of the dog. As soon as the dog is finished eating, but before it lifts its head, you call its name, then mark and reward any turn of the head towards you. In “puppy in the middle,” two people take turns calling the dog away from the other partner, rewarding in the same fashion for a head turn or, hopefully soon, the dog’s return.
In both games, Tajji rapidly began anticipating the next move, so that it became difficult to mark the exact moment for reward. However, in yesterday’s class, we found out that doesn’t matter. Just keep marking and rewarding the turning as best you can, and you will still accomplish the purpose of both games: inscribing deeply in muscle and emotions a strong value to the dog’s name so that you can get it looking back at you even when a trigger—another dog, a person, a noise—pushes it over its threshold into the unthinking depths of limbic behavior.
Other foundational games include:
Touch: touch and eventually follow a hand not holding a reward. A good relaxation move (if the dog is reaching to touch your hand, it’s not looking at the trigger) and a wonderful way to move the dog around when teaching more complex behaviors. Interestingly, Tajji quickly learned or relearned Touch to one side or the other, but it is still, weeks later, difficult to get her to touch a hand held over her head.
Eye contact: Show a treat in each hand to the dog and then hold your hands out to the side, shoulder height. Wait for the dog to look at you, mark and reward with the treats. At first the look will be brief, and probably not include eye contact. Eventually you’ll be able to demand a long gaze, even when you’re jiggling your arms and hands around.
Puppy zen: get the dog’s attention with a treat held over its head, then lower the treat slowly, stopping or even removing the treat if the dog reaches or lunges for it. Eventually, you’ll be able to ask for and get total stillness until the treat is at its mouth; at first, you’ll need to reward in a long swoop to prevent failure.
Counter-conditioning is an important tool for desensitizing a reactive dog. In yesterday’s class we started counter-conditioning against Tajji’s reactivity to other dogs. We brought her out to about 100 feet from a pair of blinds from behind which a helper could select and show various size fake dogs. We wondered if Tajji would just ignore the fakes, but we needn’t have worried. She was already amped up for some reason, and caught a glimpse of one of the dog dummies on the way across the field (poor situational awareness on my part), which triggered her just fine. We calmed her down with some T-touch (basically, mindful massage along the lines of “don’t pet the kitty the wrong way,” which, it turns out, is excellent advice for all animals) as well as most of the exercises above.
During this phase, Sandy pointed out how some of my training moves were working against me. I wasn’t noticing when she liked an exercise so much that it raised her excitement. In such exercises, I was told to move more slowly, and allow more time between trials to let Tajji “cool down.” Deborah pointed out later than I tended to get wound up, and raise the pitch of my voice, which wound Tajji up as well.
Finally we began the counter-conditioning. Using radios and hand signals for communication, we introduced Tajji to 3 different dog dummies of varying realism. All were quite triggering at first—once we got her to look in the right direction. That was only a problem the first time; after that she knew where the evil dogs were coming from. Then we used her alertness to train the “look-at-that” command, another foundational and somewhat more advanced game. Fortunately, dogs are one of the very, very few non-human animals who understand pointing, and I believe they are the only animals in which this ability is easily evoked. That seemed to work well.
The counter-conditioning was simple. As soon as I could tell Tajji was looking at the other dog, I started shoving high-value treats into her mouth. It was immediately apparent that she wasn’t too aroused to work, for she ate each time. (Not eating is almost always a sign of being over threshold and therefore incapable of learning.) As soon as I got into the game myself, re-familiarizing myself with her extremely subtle signals—sometimes merely something about the set of her head seen from above), I was able to reward quickly enough to make some real progress.
Next week we’ll be working with a real dog: one of Sandy’s. For Tajji, this will be a real challenge, for Piper is a small terrier of some sort, and Tajji apparently loathes small dogs. (I suppose there’s nothing like being ankle bitten while in your seeing eye harness and unable to defend yourself to condition a deep and abiding hatred of tiny yappers.) In the meantime we’ll work hard on the foundational behaviors, and see if we can figure out a way to continue the counter-conditioning. Like as not we’ll work with the sounds she triggers on. That happens a bit too often, and we want to terminate any rehearsals as quickly as possible, for practice perfects bad behaviors as much as it does the good ones, perhaps a bit more.
* “Swamp collie is a joking pejorative sometimes applied to German Shepherd Dogs by fanciers of other breeds. Funny thing is, most GSDs detest water.
No comments:
Post a Comment