Sharing ClickerExpo’s charm
I'm a ClickerExpo junky. The atmosphere at ClickerExpo is wonderful, and the presentations are informative and entertaining. ClickerExpo 2010 certainly lived up to my expectations. It's difficult to pick a favorite from among the three days of great sessions, but Alexandra Kurland's session on "loopy training" was definitely one of the highlights.
|Loopy Training: "Sit"|
Although I spend most of my time training dogs, I was eager to attend a session offered by a horse trainer. One of the things I've learned over the years is that clicker training concepts work across species. What's more, trainers of different species often look at training through a different lens, and listening to those trainers provides an opportunity to see through that lens. I’d like to share some of the things I picked up looking through Alexandra Kurland's loopy lens at ClickerExpo this year.
The secret to great training
From the first moments of the presentation, Loopy Training: Efficiently Teach Complex Behaviors, the audience was riveted. An engaging ClickerExpo faculty member, Alexandra Kurland hit the ground running, emphasizing and then reiterating the secret to great clicker training—being a splitter rather than a lumper. The best trainers break behaviors down, splitting them into small steps rather than lumping them together. This system helps prevent frustration and makes training much more fun and effective.
What is loopy training?
It’s important for trainers to learn to recognize the "ven" (happy) and "punir" (poisoned cue) looks in their animals. Rebecca Lynch’s recent article, Poisoned Cues: The Case of the Stubborn Dog, explains poisoned cues and describes re-training as the “cure.” In the DVD The Click That Teaches 14: The Poisoned Cue, Dr. Jesús Rosales-Ruiz, in conversation with Alexandra Kurland, discusses poisoned cues, both in general terms and as it applies to horse training.
One of the best ways to ensure a "ven" animal is by “going in circles" during training—in other words, through loopy training.
Loopy training features built-in "loops" that help training flow smoothly. Loopy training is not revolutionary in terms of how it actually works; experienced clicker trainers will recognize it as the principle operating during their most effective training sessions. Nonetheless, thinking about clicker training in terms of loopy training is a great way to boost proficiency and get better and more consistent results.
It's easiest to explain loopy training through an example. Here’s a loop that can be used to teach an animal to target.
The last element of the loop is also the first element of the loop. Once training starts, you can continue looping for as long as you want (although short sessions are generally a wise choice). A loop like the one above helps the trainer and animal move quickly and without pause between trials (repeats) of a behavior.
Each element in the loop has a purpose. For example, removing the target before delivering the reinforcer gives the trainer the opportunity to present the target again. What's more, an ideal loop contains emotional balancers that ensure that the animal remains calm but attentive during training. In a clean loop, the trainee performs all the elements within the loop smoothly and without hesitation, and no unwanted behaviors creep in (as might happen if the emotional balance is off). Once the loop is clean, it’s time to move on to the next criterion. In fact, when a loop is clean, the trainer should move on!
Both sides of the click count
Each element in the loop is important. Trainers sometimes focus too much on what the animal is doing, and lose sight of their own critical role—appropriate criteria-setting, good click timing, and prompt delivery of the reinforcer (treat, toy, etc.). If a loop isn't working, a trainer should consider all of the elements of the loop—including both sides of the click—to determine the source of the trouble. Here are a few of the questions to ask when troubleshooting a problematic loop:
- Are the criteria appropriate?
- Are the clicks timely?
- Is the reinforcer being delivered promptly?
- Is the placement of the reinforcer suitable?
- Is the intrinsic design of the loop sound?
As loops are examined and refined, patterns emerge. The search for patterns can expose skills that need improvement. Perhaps a trainer is not very good at tossing food, for example, and the animal has to spend time finding it.
The Loopy Training Session at ClickerExpo emphasized that the key to great clicker training is ensuring that, 1) criteria are well selected and well defined, 2) timing is impeccable, and 3) the rate of reinforcement is high. The more skilled the trainer, the more likely it is that the learner will remain happy and engaged.
“Looping” back to ven (happy) vs. punir (poisoned cue) learning, Alexandra Kurland asserts that poisoned cues are a product of bad training—specifically of lumping. As mentioned above, loopy training is a splitter's tool. With loopy training, the trainer can create tight loops that exclude unwanted behaviors. Once a loop is sound, it can be "enlarged" slightly by raising the criteria, and then worked until there is success in that new loop. The loops spiral out gradually until the goal behavior is achieved. Loopy training encourages raising criteria in small enough increments to ensure success, and therefore a ven state, in the learner.
"Touch-Sit-Down" Behavior Chain
Loopy training for behavior chains
Loopy training can also be used in behavior chains. Let's say there is a dog that already knows how to sit, lie down, and target on cue. Here are sample loops for each of those behaviors:
Cue: Sit -> Dog sits -> Click -> Tossed treat (moving dog out of sit) -> Cue: Sit
Cue: Down -> Dog lies down -> Click -> Tossed treat (moving dog out of down) -> Cue: Down
Cue: Presentation of target (with accompanying verbal cue, if appropriate) -> Dog touches target -> Click -> Target lowered and treat delivered -> Cue: Presentation of target
Once the above loops are clean, the trainer can begin using them in short behavior chains. For example:
Cue: Sit -> Dog sits -> Cue: Down -> Dog lies down -> Click -> Tossed treat -> Cue: Sit
Stringing behaviors together to form more complex loops works because cues create effects in two directions. In the example above, the dog has learned that the cue "down" leads to clicks and treats. The cue to lie down has two functions in this loop—it reinforces the dog for sitting, and at the same time tells the dog what the next "hot" behavior is going to be. The entire chain remains strong because each link in the chain was built out of clean loops. It’s possible to add more elements to a chain as long as each of those elements is clean as well.
Another option is to string all three behaviors together, as in this example:
Cue: Presentation of target -> Dog touches target -> Cue: Sit -> Dog sits -> Cue: Down -> Dog lies down -> Click -> Treat -> Cue: Presentation of target
Behaviors can also be put together with intermediate clicks, as in the following example:
Cue: Presentation of target -> Dog touches target -> Cue: Sit -> Dog sits -> Click -> Treat -> Cue: Down -> Dog lies down -> Click -> Treat -> Cue: Presentation of target
It’s also possible to click and treat between every single behavior, as long as the loop remains sound.
Experimenting with new techniques is a great way to stay interested in training. Whether you are just getting started in clicker training or have been doing it for years, try loopy training. No matter what species you work with, designing training in terms of loops will help you determine which of your skills need polish—and it will help you get the best possible results.