Shaping is commonly defined as:
The differential reinforcement of successive approximations toward a target behavior.
Now, that’s a mouthful!
In plain English, this basically means that you start by reinforcing (rewarding) little pieces of the behavior, and gradually build up the behavior until you reach your final goal.
However, let’s break this down and look specifically at each of the major elements in the definition of shaping. Understanding each of the three core elements will help you when training your dog and when designing your own shaping plans.
Target behavior. Your target behavior is your final goal. It’s what you want to achieve as a result of the training process. You’ll want to make sure you have a clearly defined target behavior before you start training. This means thinking through exactly what you want the dog to be doing, what situation(s) the dog will do the behavior in, what cues you’ll use, and any other particulars.
For example, perhaps you are worried about your dog running out the door when visitors come over. So, you decide you will teach him a really solid stay. However, saying “I want my dog to sit and stay and not run out the door” isn’t nearly precise enough! Exactly when and where do you want the dog to stay? In order to effectively solve this problem, you’ll want to be specific regarding when and how the dog should do the behavior.
So instead, your target behavior might be to have your dog go to his dog bed in the living room when the doorbell rings and then stay lying on his bed while you open the front door to collect a package or let a visitor into the house.
Successive Approximations. The successive approximations are the baby steps you will use to train your target behavior. Rather than starting with the final behavior, you will start by rewarding approximations to the final behavior. Each step, or approximation, will take you closer to your goal.
For instance, did the previous example of teaching the dog to go to his bed when the doorbell rings seem impossible to you? This is a behavior that would be pretty difficult or even impossible for a lot of dogs! However, by breaking the behavior into a series of smaller, achievable steps, it becomes much easier to teach.
You could start by teaching your dog to voluntarily go to his bed. Then, you could teach him to do this when the doorbell rings. Next, you could teach him to stay on the bed for longer and longer amounts of time, while you move around the room. Then, you could have him stay in his bed while you open the door and finally, while you let a person come in. (Note: These are all still very BIG steps. Each of these steps would likely be broken down into a dozen or smaller steps.)
Some people are skeptical about shaping. They think that it will take a lot of time to train a behavior if they have to break the behavior down into lots of little steps. However, the reverse is usually true. Small steps actually make the training go faster because the communication is clearer and the animal understands exactly what to do.
Differential Reinforcement. The basic process that keeps shaping going is reinforcement. Every time the animal does the right behavior, the trainer follows this with something that the animal likes, such as a food treat, a belly rub, the chance to chase a ball or anything else that the animal finds rewarding. The trainer knows that reinforcement is occurring if the animal continues doing the behavior and actually does the behavior more frequently.
The interesting thing about shaping is that there are always some behaviors being rewarded, and other behaviors that don’t earn rewards. This is the process that behavioral scientists and psychologists call differential reinforcement.
Back to the example with the dog bed. If your dog has not been previously trained to go to his bed on cue, you might start by giving him a treat anytime he puts a foot on the bed. Since you have his favorite treats and you start by standing right next to the bed, he figures this out fairly quickly. Pretty soon, he understands this step and is repeatedly offering to step onto the bed with his front feet. Every time he does this, you reinforce the behavior by giving him a treat.
So, you go on to your next step (your next approximation) and now only give him a treat if he puts all four feet on the bed. If he only steps on the bed with his front feet, he no longer gets a treat. Even though this behavior previously earned a treat, it doesn’t anymore. Your pup starts experimenting a bit, and it doesn’t take long for him to catch on to the new requirement. Now he is going all the way onto the bed each time. Once he is doing this reliably, your next step might be for him to go all the way to the bed and then lie down. For this step, standing on the bed would no longer earn a reward, but lying down would.
At each step, the approximation you are currently working on will earn reinforcement. If the dog offers behaviors from previous steps or completely unrelated behaviors, these won’t earn reinforcement. As a result, the behavior that is earning reinforcement increases in frequency, until the dog is spending most of his time engaged in this behavior. The other behaviors that aren’t earning reinforcement don’t happen as often or disappear entirely. This is the process of differential reinforcement.
(That said, trainers do sometimes go back and reinforce behaviors from previous steps, depending on the dog and on the behavior being trained. This can be useful sometimes if the dog is getting frustrated or confused with the current step.)
The trick with shaping is to divide the behavior into the right steps so that every time you move to the next step, it is very easy for the animal to figure out the new requirement. If the steps are too big, it will be too difficult for the animal to figure out what you want when you move to the next step. This can lead to frustration and confusion, because the animal will be attempting different behaviors, but not getting reinforced often enough.
Designing the right shaping steps is the topic for another blog post (or even a whole book….), but for now I want you to just remember that with a good shaping plan, it should be very easy for the dog to succeed and it should not be difficult for the dog to figure out what you want at each step. If the dog is offering a lot of extra behaviors or unwanted behaviors or can’t figure out what you want, you’ll need to add in some intermediate steps or even rethink your whole plan.