Reprinted from the Collie Connection, Winter 1999-2000
Babe is now learning the "stay" and the same issues of positive vs. negative reinforcement arise. So far, I have used the clicker as only one tool in my training bag, thinking that teaching some behaviors cries out for the clicker and it's foolish to train those behaviors any other way, but for other behaviors, using the clicker is contrived. One of latter behaviors, I thought, was the stay. At what point in the "stay" do you need the precise timing that makes the clicker such a brilliant tool? Rather, you need a precisely timed correction when the dog breaks the stay. So, I concluded, the stay is better taught with negative reinforcement. Enter Babe, forget corrections. What a joke! The idea of training Babe to stay by correcting her is ludicrous--I can't think of a faster way to get her to run away than by correcting her for not staying. So I began the long, laborious process of teaching Babe to stay with a purely positive, operant method--clicking her for progressively longer times staying put.
Rather than a contrived use of the clicker, teaching the stay with the clicker I found to be operant conditioning at its most rarified. It became obvious that I had to put this behavior (staying) on variable reinforcement immediately to make any reasonable progress at all. "Variable reinforcement" means that I make Babe stay put for varying amounts of time before rewarding her for it--specifically, unpredictable intervals of staying. Then I try to increase the average amount of time that Babe has to stay put before she is rewarded. If I just increased my expectations gradually and consistently, Babe would learn how long she has to stay before she gets a reward, and then she could anticipate the reward, perhaps breaking the stay before I clicked her, in which case she would get no reward and we have to start over. But if I am unpredictable, Babe would have many successful stays, which were shorter than she expected, and if I decide to wait a long time, Babe would think "when is it, could it be any time now?" Of course, she would eventually also break her stay if I were pushing for a long stay under either method, but the unpredictability of the reward turns out to be a really powerful way to reinforce a behavior in general.
To train Babe to stay, I am using a place board, which is a little wooden platform that is just as long and wide as one Collie and is about 3" high. A place board makes it black-and-white that what I want is a stationary location, because the dog has to step off the platform to leave that location (an idea I learned from Jim Dobbs). I start out by not requiring a particular position on the place board--she could stand, sit or lie down. When you think about it, staying in one location is a separate concept from holding a certain position. So I let Babe choose her position and reward her for staying on the place board. Quickly, however, she adopted a sit for her staying, which was great, because that will make it easier to phase out the place board--if she is sitting or lying down, staying in one location is a much easier concept than if she were standing.
Remarkably, this method is working! Because Babe is so frantic to train, the "stay" seemed to be our new eons-away goal. At first, getting her to stay was really challenging, because in addition to being frantic to work, Babe is a classical clicker-trained dog, who realizes that her responsibility is to offer behaviors, which she does with neurotic intensity. But suddenly she caught on, and we are progressing more rapidly (with variable reinforcement) than I ever imagined. I even have some fine points to shape during training the stay, such as clicking only when Babe's two front feet are still (not an easy task) and when Babe is looking me in the eye.
Who can say where Babe's training will end up? She is not a dog to be underestimated! Perhaps one day you will see her in the competition ring.
Babe Top|Rescue Top|Site Map