Shaping: Behaviors vs. Results (dog agility edition)

I have been thinking a lot about behaviors lately, because behaviors that are rewarded are reinforced. The mantra of trainers everywhere is “what gets rewarded gets repeated.” So if you’re trying to create new behaviors in order to get different results, you have to figure out what behaviors to reward and how exactly to reward them. You can lure new behaviors, but it turns out that rewarding after the fact is way more effective than luring, because in luring, the lure is the focus; in shaping a behavior without luring, it becomes a puzzle: what did I do that resulted in that reward, that reward that I wanted so much? I have found, with Rush, that shaping–helping him puzzle out what he did that got him the reward he wanted–has resulted in solid, enthusiastic behaviors. He loves agility; I’m very proud of just how much he loves it!

In shaping a behavior, the hardest part for me is to figure out where to start and what steps to add. Sometimes I have to work three or four different pieces separately. To train Rush to put up with having his face clipped, for example, I worked on shaping “hold still while I hold your chin in one hand and look at you” as a completely separate piece from “hold still while I run this clipper,” which was, in turn, a separate piece from “hold still while I touch this clipper to your face.” It was quite a while before I put two pieces together–touching his face with the (turned-off) clipper while I held his chin–and even longer before I put all three pieces together. I didn’t get the result I wanted–“hold still while I clip your whiskers”–until I’d done a lot of work on little pieces, individual behaviors.

Back when Rush was a puppy, I tried to train all of his agility obstacles using shaping, rather than luring. There were times that was quite a challenge, but as Rush learned more obstacles and I learned more about shaping, it got easier. The lightbulb moment, though, occurred when Debbie told me Rush was ready to learn the tire jump. The tire jump is a somewhat notorious obstacle. In competition, dogs often avoid it, or jump between the frame and the tire, or duck under the tire. Typically people train it after they train jumping, and they put something on the sides of the tire so that the dog doesn’t jump between the tire frame and the tire. Alternatively, the handler lures the dog through the tire–and the dog is looking at the lure, not the tire.

Debbie’s method of training the tire was different. She had a tire–and no frame. Unobtrusive stakes had been duct-taped to the side of the tire and the tire was staked to the ground. Again, there was no frame. It was the essence of the tire. By the time we started the tire, Rush had a good startline stay, and he’d learned a lot about shaping; I’d shaped the tunnel, the startline stay itself, a “go round” with a big traffic cone and a jump wing, and a little tiny teeter that was 18 inches wide and six feet long with a fulcrum that was only three inches high. He knew what it meant that I was looking at something and had a handful of treats. So I stood and looked at the essence-of-tire and eventually Rush came toward it, I said “yes!”, and I threw the treat where he’d see it through the tire–and he went through it. We did that a few times, with me standing still, and as Rush got more confident that going through the tire was the right thing to do, I started running alongside him and he started running back and forth through the tire. We repeated that from scratch for a few training sessions. Rush’s tire jump is a solid, safe performance and he does it without hesitation, even with a slice. You can see how relaxed he is about the tire in this photo (taken when he was two-and-a-half years old):

Photo by Joe Camp

Photo by Joe Camp

I used a similar method to shape the broad jump. Rush’s first “ordinary” jump was a single bar jump, he was fourteen months old, and we rapidly raised the bar from eight inches to 20. He jumped 20 inches in competition until he was two years old, when I raised his jump height to 24″.

I shaped Rush’s weaves, starting with 2x2s. I shaped the end behavior of the dog walk, teeter, and a-frame long before I let him do the complete obstacles. I knew exactly what I wanted: right paw on a target, dog stopped dead. (It occurs to me that this may be why Rush so often stops at the end of the obstacle with one paw up. I only cared about that right paw, and I let him do what he wanted with the left–so he keeps it up and ready to rock ‘n’ roll.) I trained his target behavior starting when he was about three months old, using what I call the “big blanket” method, which I first saw demonstrated by Susan Garrett.

The “big blanket” method was my first exposure to the successive approximations needed for shaping behavior. If the goal is to have the dog put his foot on a target, it’s possible to start with the target and the dog in a small space, and reward the dog for getting closer to the target until finally the dog is touching the target. That works. But the “big blanket” method is so much easier. You start with a big blanket and you spread it out on the ground. The dog is over there somewhere and you are over here somewhere, next to the blanket, holding that oh-so-tempting handful of treats. The dog wanders toward you, and steps onto the blanket–and you say “yes!” (or click, but you always have your voice, so “yes!” can be a lot easier–at least, it was for me) and throw the treat off the blanket. After a few repeats of this, just fold the blanket in half, and the target is now half the size, and the dog is still hurrying to step onto that blanket because there’s something about the blanket that makes you say “yes!” and throw a treat–and every so often you fold that blanket in half, and five minutes later the dog is hurrying to step onto a target that’s about 6 inches square and fits in the back pocket of your jeans. I switch to a washcloth when the blanket is about a foot square. When the dog is reliably hitting the washcloth with it folded in quarters, I switch to a white bathroom tile. Easy for the dog to see, can be half-buried in the dirt to make it smaller, and cheap.

So my end behavior for the teeter, the dogwalk, and the a-frame was exactly the same: I wanted Rush’s right front paw on that tile. I put the table next to the obstacle, and I had Rush jump up on the table, and go right down to the target. He knew what to do, just as soon as he saw the target, and it wasn’t long before he did the whole obstacle, and correctly, every time. That was when I started burying the tile, a little bit more every time, until I was finally able to take it away. You can also use a big plastic lid, and a pair of scissors, and just scissor off a bit of the lid at regular intervals.

(A note on the teeter: because Dancer was so worried about the teeter, I really didn’t want Rush to be worried about it. I had a tiny training teeter at home and that was the first contact obstacle he learned. He was never surprised to discover that something moved under his feet–because he’d already experienced that.)

I have shaped a lot of non-agility behaviors with both dogs, as well. They sit and wait for me to put their food bowls down; they wait on their mats while we eat because that’s their best chance of getting something off our plates. (Dancer is especially fond of roast chicken.) Both of them have a very nice “gimme five.” A very mundane behavior that saves me a ton of time is nail Dremeling. When I set up the grooming table and get out the Dremel and get a pocket full of treats, both dogs come running. That took months to train, and I reinforce it with fresh chicken liver treats every time, but I can do all 34 nails (16 on each dog, plus Rush’s dew claws) in less than five minutes, including the time it takes to unfold the grooming table and put it away when I’m done.

(Notice that I haven’t discussed reward quality here. Reward quality really matters. That’s why I have a batch of chicken liver treats in the oven right now.)