Clicker training
From Wikipedia, the free encyclopedia
Clicker training is the process of training an animal using a conditioned reinforcer, which indicates to the animal ("marks") the precise behavior that was correct. It was originally used in training animals such as dolphins and pigeons, for which traditional methods of obedience training aren't useful. Although it is a form of operant conditioning, the more popular name "clicker training" comes from the fact that the conditioned reinforcer used by many dog trainers is a small mechanical noisemaker called a clicker. Clicker training has also been successfully used in horse training.
Contents |
[edit] Co-Founders
B. F. Skinner first identified and described the principles of operant conditioning. But it was Marian and Keller Breland, two of Skinner’s first students, who saw the possibilities for animal training, and who developed reinforcement and shaping as we know it today.
Starting with pigeon projects in World War II, the Brelands formed a company, Animal Behavior Enterprises. They developed the first marine mammal training programs for public exhibits and for the Navy. They created the first free-flying bird shows, and a host of commercial animal exhibits, from piglet races to chickens playing tic-tac-toe, to an entire “IQ Zoo.”
After Keller’s untimely death in the 1960s, Marian married Navy dolphin scientist Bob Bailey, who joined in the pioneering work. Radio-carrying cats were steered through cities and into buildings. Dolphins located targets many miles from their trainers, at sea. Ravens and other birds, carrying cameras and directed by lasers, could fly to a specific window of a skyscraper and photograph the people inside. Gulls, expert sea searchers by nature, could locate and report life rafts and swimmers far offshore.[1].
[edit] Advantages
One of the challenges in training an animal is communicating exactly when the animal has done the behavior that the trainer is attempting to reinforce. As a simple example, consider teaching a dog to back up. At the instant that the dog moves backwards, the trainer must let the dog know that it has done the correct thing. However, the traditional "good dog!" takes so long to say that the dog might already have moved on to some other behavior. By the time the dog realizes it is being praised, it might be moving forward again, or even sitting and scratching.
Besides the imprecision in timing, using the trainer's voice for information means that the actual signal will vary. The trainer's voice, pronunciation, tone, loudness, and emphasis can change even during the same training session, and training is slowed when the animal has to realize that some variations (loudness or enthusiasm) aren't important while others ("good dog" vs. "bad dog") are.
A conditioned reinforcer solves these problems. The conditioned reinforcer can be any signal that the animal can perceive, so long as the signal is brief (to prevent the problem of imprecise timing) and consistent (to prevent the problem of variations that may confuse the animal). Dogs and horses are often trained with a clicker or cricket, a small metal tab that makes a "click" sound when pressed and released. For dolphins, the conditioned reinforcer is usually a whistle. However, not all conditioned reinforcers are sounds. Goldfish can be trained using a quick flash of a flashlight as their "clicker" [1] , and deaf dogs can be trained with a vibrating collar [2].
[edit] Methodology
The first step in clicker training is to teach the animal that the clicker sound means that they will get a reward. To do this, the trainer does what is called "charging" or "loading" the clicker. The trainer clicks the clicker and simultaneously or immediately thereafter gives the animal a reward, usually an unaccustomed, tasty treat, one small enough to be consumed almost instantly. (Some trainers substitute play with a favorite toy. However, this practice can interfere with the goal of maintaining a high rate of reinforcement.) The trainer performs up to 20 repetitions per session.
Some animals tend to learn the association much more quickly than others. Dogs, for example, often learn the association in one session, with as few as five to 10 repetitions. Progress may be tested by waiting until the dog's attention is elsewhere and then clicking. If the dog immediately looks toward the trainer as though expecting a reward, it is likely that the dog has made the association.
After that, the trainer can use the clicker to mark desired behaviors. At the exact instant the animal performs the desired behavior, the trainer clicks and promptly rewards. One key to clicker training is the trainer's timing; clicking slightly too early or too late rewards and therefore may reinforce whatever behavior is occurring at that instant. Another is to create opportunities for the animal to earn rewards very frequently. A reinforcement rate of one click/treat (C/T) every two to three seconds is common among professional dog trainers. Finally it is often necessary to break down even simple tasks into smaller sub-tasks (see chaining) or to start with easy-to-meet criteria which are gradually tightened.
[edit] Examples
Many desired behaviors start with the nose-touch, where the dog learns to touch an identified target, such as a small piece of plastic, with its nose; that behavior can then be transported to perform useful tasks or interesting tricks such as flipping a lightswitch or ringing a bell to go outside.
Training the nose touch begins with getting the dog to touch a target with its nose; trainers sometimes use a guided method, such as placing a dab of peanut butter on a small plate or plastic target; others prefer shaping, where the target is placed in easy reach, such as in the trainer's hand between the trainer and the dog, and the dog is rewarded each time he moves in the target's direction or actually touches it.
When the dog is consistently touching the target, the trainer progresses to a target with and without food and in different positions. Eventually, the trainer can transfer the behavior to a bell, for example by holding the target behind the bell so that the dog has to touch the bell to get at the target, and then rewarding the touching of the bell. When the dog is reliably touching the bell, the trainer now adds the act of opening the door to the reward each time the dog strikes the bell.
Targeting for Horses: For horses, loading or charging the clicker is usually not done. It's best for horses that a clear marker is used so that the horse does not expect "unearned" treats.
[edit] External links
[edit] References
- ^ The Don’t Shoot the Dog! News, (1996, November), p. 1.
