When I talk about behavior chains, I talk about the importance of completing the chain. Because in a chain each cue serves as a reinforcer for a previous behavior, dropping cues is actually failing to reinforce — and we know that’s a bad thing. Unreliable reinforcement leads to unreliable behavior. Variable reinforcement leads to variable behavior. (That’s great when we’re shaping, not so great when we’re maintaining.)
Today I broke a chain.
I was getting ready to leave the house, so I opened the door and called the dogs in, sending them to their kennels in my bedroom at the far end of the house. They ran past me, and as they hit the hall I remembered that Undómiel’s crate wasn’t in my bedroom, but was outside for cleaning.
I followed the dogs down the hall and found Penny in her crate, waiting patiently. Undómiel, however, was circling the place where her crate should have been, looking displeased. How was she going to get her treat for crating promptly if she didn’t have a crate? So I handed her a treat and then led both dogs out again to retrieve her crate from outdoors.
Now, this is where a lot of people would say, She got a treat for going to the right place, so she was reinforced, so everything’s okay.
It wasn’t okay.
Remember, a treat is just a treat until it affects behavior. Intention means nothing. You cannot identify reinforcement or punishment until you see the next repetition. Even though Undómiel had scored a tasty gourmet bison biscuit (something new I picked up which they love), she was about to tell me that it was yummy and all, but it was not a reinforcer, at least not in that moment.
So I placed the kennel in its usual place, both dogs watching, and I went back down the hall to close up and to get another bison biscuit to split between the dogs. Then I cued, “Kennel.”
On a normal day in my house, this cue tells the dogs to go to their crates and any nearby humans to take shelter. My dogs love their kennel cue and they will go in as direct a line as possible, frequently clearing the couch en route. Undómiel is particularly enthusiastic, or as our petsitter described her, “I’ve never seen anyone so eager to trade freedom for a tortilla chip.” Penny usually ducks to one side to avoid Undómiel’s rush and then runs after her because there’s a cookie for second place, too. This is a very simple game we play, running to a crate on cue, but it always results in treats and comfort, and they have very reliable and very enthusiastic behavior in response to this cue.
But this time, I cued “kennel!” and Penny turned and started off. Undómiel hesitated, and then she started off at a trot. A trot, for crying out loud. Around the couch. Looking back at me every few steps. She was responding to the cue, but there was a distinct air of distrust and dissatisfaction — until she saw her crate, right where we’d left it, at which point she dove in and waited happily for her treat.
I tested the cue again just now and both dogs went directly to their kennels, did not pass go and did not collect $200. They got the same tiny bison treat.
This was the exact same half inch of a biscuit she’d gotten a few minutes before. The same treat. What happened?
The answer is that going to the kennel is not a single behavior, though we often think of it that way, but a very short behavior chain. Navigate the living room, go down the hall, take the corner, see the crate, enter the crate and wait for delivery and the door closing. And it’s a behavior chain which never varies. We may start from slightly different places in the living room, but we never enter the crate and then go down the hall. There’s an order to this.
When she was denied the final part of her chain, she could not complete the chain to earn reinforcement, and that final behavior and all the previous behaviors were not reinforced. Even though I gave her a treat, I gave it for circling and looking for her crate, not for waiting in the crate, and more importantly it was not enough to counter the disruption of the chain and its anticipated reinforcement. The disappointment and frustration of not being able to complete her chain was stronger than the tiny (and unrelated) cookie.
Now yes, I’ll be the first to admit this was hyperbolic. Undómiel is rarely shy about her emotions. “You broke the rules!” With most dogs, the difference in behavior would have been much less… dramatic. But it was because this chain was so well-worn, so solid, so utterly predictable that the disruption was an issue. If this had been a chain which often varied in order, or ended at various points, it would have been much less of an ordeal when it was disrupted. Dogs don’t care which obstacle an agility course ends on, for example, because there’s no one obstacle which consistently predicts the end and treats or toys. The chain can end anywhere and still lead to reinforcement.
“But wait, Laura! Isn’t it bad that her chain fell apart so fast?”
First of all, it didn’t fall apart. It lost enthusiasm and speed. She still went directly to her kennel without pausing or exploring other options. And the recovery was nearly immediate, a single reliable repetition to reassure her that the rules were back on. So no, I don’t think it’s bad. I think it’s a more visible than average view of a process which happens frequently and is often overlooked in a less-dramatic dog. If that single error had disrupted the behavior for a long while, yes, I’d be worried about our previous training! But one repetition of asking, “Are you sure? Is this still what you’re looking for?” actually shows a very responsive attention to feedback.
What’s the take-home message? If you have a behavior chain which must be completed in order — say, search/locate/report, or a formal retrieve — make sure you complete the chain. If the dog doesn’t have a drug cache to report, he can report an all-clear, for example. If the retrieve goes wrong, interrupt with another cue and redirect, don’t just leave the chain dangling.
And remember, a treat or an aversive isn’t a reinforcer or a punisher until you see the next repetition of behavior. The learner makes that call, not the trainer.
Here’s to reliable chains!