The quest for greatness
One characteristic of a good shaper is flexibility—a willingness to change course based on new information coming in. Good science shares that characteristic. As Susan Friedman once said, “A fact is only a fact until it’s replaced by a better one.” I love this pithy reminder that although something may seem obvious or indisputable, rather than becoming complacent, we continue searching for something even better.
This is as true in the training community as it is in the scientific community. Every year we grow and evolve. We discover faster, better, more creative ways of training. For example, at ClickerExpo, the epicenter of groundbreaking positive reinforcement training, faculty members themselves never fail to learn something new from each other every single year.
As good shapers, we must be willing to revisit the principles by which we train—no matter how comfortable and familiar they have become. In that spirit, the “10 Laws of Shaping” have evolved since Don’t Shoot the Dog! was first published. We have learned more and more about the science that powers the clicker training engine.
Variable schedule or natural variety?
A few years back, Karen Pryor and I were working together on updating some of the information within Karen Pryor Academy’s (KPA) Dog Trainer Professional program. During that update process it was hard to overlook the fact that the third original “law” of the 10 Laws of Shaping was no longer consistent with what we actually teach in KPA.
The original 10 Laws of Shaping were published in 1984 and, as Karen said, “We’ve learned a lot since 1984! We’ve gotten much better at shaping over the years.” The original third law states:
During shaping, put the current level of response on a variable schedule of reinforcement before adding or raising the criteria.
That suggestion conflicts with what we now firmly believe and advise people to do in KPA: when training a new behavior, use a continuous schedule of reinforcement (CRF). In shaping, we advise using a CRF schedule for the approximation we are strengthening. So, a dilemma. We certainly didn’t want to remove the still very valuable 10 Laws of Shaping from the course. But we didn’t want to confuse people with conflicting information either. So we decided it was time to update the beloved 10 Laws of Shaping, beginning with law number three.
“As soon as the animal learns that if it doesn’t hear a click, try again, it’s already entering the world of natural variety,” said Karen. That natural variety Karen refers to is built into the shaping process inherently. Animals are not robots. They don’t exhibit the exact same behavior every time. They can’t. Their speed, position, and muscle movements are inherently variable. We humans are not robots either. We don’t perceive the animal’s movements the same way every time. We click a fraction of a second (or more!) sooner or later than we did last time. We withhold a click for a brief moment and see if the animal will give us something a little different, a little closer to our goal. If they do, we click that instead.
Consider the following example of a shaping session:
- Animal dips its head and gets clicked for it
- Animal dips its head in a slightly different way and gets clicked for it
- Animal dips its head, but his trainer is slightly too slow, so animal dips it again. Click!
- Animal tries something different—a sit—and hears no click.
- Animal dips its head again. Click!
The natural variety is built in, and a variable schedule isn’t necessary. “If you actually put each criterion on a variable schedule before moving on, it would slow you down and interrupt the flow of the shaping process,” says Karen.
Original intent and early uses of variable schedules
The original thought behind advising trainers to sometimes withhold a reinforcer until an animal performs two iterations of a given behavior (which Karen called “asking for two-fers”) was to get a naive animal used to the idea that it doesn’t always get reinforced. The idea would prevent the animal from potentially suffering the frustration of an extinction burst when it suddenly fails to earn a reinforcer for a previously reinforced behavior.
Decades ago, some trainers would use these frustration-filled attempts at gaining reinforcement to select for more intense behavior. While “surfing the extinction bursts,” the trainer would purposely withhold reinforcement, inducing brief frustration in the animal. That frustration often resulted in a bigger/higher/faster/harder behavioral response that could then be clicked and treated.
But, based on what we know about extinction (it is very frustrating for the animal, slows learning, and can lead to aggression or quitting altogether), using extinction bursts deliberately to produce behavior directly conflicts with our desire as clicker trainers to keep animals confident, stress-free, and happy throughout the shaping process. The focus on the animal’s comfort with the training process results in stronger, more reliable behaviors and a much better overall training relationship.
The whole shaping process is a dance between trainer and learner. The more skilled the trainer, the more fluid the dance. If the trainer has poor observational or mechanical skills, it probably leaves the animal feeling like his partner has two left feet. But, even a bad dancer who is trying hard and fun to be with is better than a dancer who suddenly leaves the floor without warning.
Additions to the 10 Laws
While the deletion of the third law was the most significant update to the 10 Laws of Shaping, we made some other important changes as well, including adding two new principles.
“Be Prepared Before You Start”
So many times, the trainer misses that very first “clickable moment” in a shaping session. When that happens, it sets the whole session off to a bad start. This result is especially common with trainers using a prop or a target of some kind. It’s so important to be ready to catch that very first inclination the animal shows toward your goal behavior. “Grab whatever goes by that’s in your general direction,” says Karen.
When we click and treat an animal for looking, sniffing, or pawing at the mat or the platform on the ground, we are not only saying that the goal behavior has something to do with that object, we are also reinforcing his curiosity and his braveness. The more times curiosity and braveness are reinforced, the more confidence the animal gains. We end up with an animal that is creative, is excited about interacting with novel things, and readily offers new behavior. That kind of animal is much easier to train.
When we miss that very first clickable moment like the sniff, the glance, or the paw touch, we are telling the animal: “It’s not that object. Try something else.” Often the animal will do just that. The animal asks the trainer a question with the sniff, look, or touch: “Is it this?” The lack of a click tells them: “No, it’s not.” We owe it to our learners to be good dance partners and not miss a beat.
“Keep training sessions continuous.”
This principle was added to help reduce the amount of “dead-air time” in shaping sessions. Just like those times when the TV displays snow or the radio emits static, if only for a moment or two, dead-air time can be the kiss of death in shaping. If the animal is left hanging, it might be because the trainer waited too long to click, set up the environment poorly, or is holding out for too much too soon.
If your animal is, for whatever reason, left hanging, you’re doing something wrong. He will find a way to fill that dead-air time—often with behaviors that have nothing to do with what you were looking for, or with displacement behaviors brought about by the stress of lack of information.
As Karen puts it, “He should be working or thinking the entire time, except for the moment he’s consuming/enjoying his reinforcer.”
On changing trainers midstream
To make room for the two new laws, the other law we decided to remove was “Don’t Change Trainers Midstream.” While this is still sound advice, it can be difficult to adhere to in some cases. In addition, we felt that the principles relating to the process of training and how a training session takes place are more important than who is doing the training.
For maximum shaping efficiency, it is indeed ideal for one trainer to work on shaping a behavior throughout the acquisition phase. Once the behavior is on stimulus control, the cue can then be transferred to other trainers. This is fairly simple to adhere to in a household setting, but sometimes more difficult in a zoo, aquarium, shelter, or other environment. While the efficiency of shaping that behavior would suffer a bit, as long as the second trainer is a proficient shaper, progress toward the goal behavior should resume fairly quickly with either trainer.
Switching trainers has less to do with the principles of shaping and more to do with best practices in training. We removed that law to make room for the others and keep it to an even 10.
Breaking the “Laws”
The 10 Laws of Shaping aren’t actually laws at all. “Principles,” said Karen. “We should call them principles instead.” They don’t fit the definition of a scientific law, such as the law of gravity, which is a statement of fact based on repeated observations that describes some aspect of the universe. Nor are they laws in the criminal sense. There are no shaping police to come and lock up shaping scofflaws. “Principles” better describes Karen’s training guidelines, as it does not have the same connotation as the word “law.”
Also included in the new “10” title is the word modern. This is by design as well. Just as clicker training is the modern way of training, these principles, at Karen’s suggestion, are described that way as well. “We need modern in the title,” she said emphatically and with a smile.
Finally, we wanted the new Modern Principles of Shaping to fit with TAGteach-friendly phrasing. In TAGteach, instructions are phrased in the positive. So, rather than “Don’t interrupt a training session gratuitously” (old law), it’s “Keep your attention on your learner” (modern principle).
Each principle is succinct (usually five words or fewer) and phrased to tell the learner what to do, instead of what not to do. This is, of course, exactly what is done in clicker training.
The characteristics of a great trainer
Looking at the changes made to the original 10 Laws of Shaping, it is quite remarkable how much we didn’t need to change! Used by trainers all over the globe, Karen’s guidelines on how to conduct a good shaping session are still so applicable more than three decades later. Given that, it would have been easy for Karen to respond with “Not a chance” when I swallowed hard, approached her one day, and said, “I think we should rewrite the 10 Laws of Shaping.” Instead, she said unhesitatingly, “Great idea. Let’s do it.”
That’s Karen. Humble. Flexible. Always willing to learn and improve. Always displaying the characteristics of a great scientist and a great trainer.