The Difference Between Chaining and Increasing Criteria

Hi Cricket,

Last week you asked a question about shaping. For the benefit of the list... You had lots of criteria, and you gradually added them, but as you progressed, some of the earlier behaviors fell apart. So you asked if at some point you are supposed to judge multiple criteria. I sent your question to Bob Bailey, and here is his response:

"Yes, you usually keep the old criteria. I say that, but it is really the trainer's choice. Sometimes, for a SHORT time, the trainer may ease up a bit on the old stuff while concentrating on the new stuff. If the overal behavior is fairly complex, with many fine points (that is, responses, remember?) to it, and you are ready to add another tough step to the chain, you might want to get the early part of the chain beyond 80 percent, say up to 90 percent and over a number of sessions. In general, when you use extinction to eliminate unwanted behavior you must be sure that wanted behavior (or responses) are strong enough to resist the extinction. I would certainly work on the individual pieces to get them strong and then work on the chain, one piece at a time and each piece being exactly right. Then, when the chain is ready for the next piece, add it on. Don't move ahead if the behavior is not ready. "

I'll need to add this as an addendum to my chicken notes. Here's my interpretation of when you'd use the 80%-and-move-on method versus the method described above.

Say I'm shaping the spin. I'm trying to shape a single behavior in gradual steps. Each time I change the criteria, the entire behavior changes. I may start with a glance, but next time I want head movement. So this behavior I do by increasing the criteria each time I get the desired response to 80%. After all, I don't want the animal to get incredibly comfortable with a glance to the side because that's not where we're headed.

Now for a chained behavior. This would be easy to see in something like a retrieve, which is a group of distinct responses (a take, a turn, a recall, a sit in front, and a give). But would this also apply to something like a competition sit? The sit must be tucked, straight, square, and fast. All of those individual criteria must be present in the final picture.

Bob is saying, I believe, that this behavior is essentially a chain. The first thing I do is work on a tucked sit. Because the tucked sit must not disappear when I work on square, I don't automatically progress at 80%. I go for 90%, and I do several sessions at that level. But even with the chained behavior, I don't do so many sessions that the dog is convinced that the tucked sit is the final behavior. When I add square, I'm only judging the quality of the "square" BUT because tucked is part of the chain, if the sit isn't tucked, I abort the exercise. I don't add straight until the tucked, square sit is strong.

Do you think that's right? The difference between the sit as a chain and the retrieve as a chain, is that all of the sit behaviors are done at the same time, so I can neither back chain, nor use Premack.

Melissa Alexander
Positive Paws Canine Resource Center
Seattle, WA
mcalex@clickersolutions.com
copyright 1999 Melissa C. Alexander

| Training Articles Contents || Site Home |

List and Site Owner: Melissa Alexander, mca @ clickersolutions.com