The following is information on clicker training taken largely from Karen Pryor's beginners manual, "A Dog and A Dolphin." Those interested in learning more about clicker training can order this book, as well as "Don't Shoot The Dog," from Sunshine Books or Direct Book Catalog (1-800-472-5425 or 1-800-776-2665). Videos, and training materials, by Gary Wilkes as well as other books by Wilkes and Pryor are also available. Clicker Training - The Basics
By Linda Michaels and her Clickin' Boxers (Shug, Tashi, and Sailor)A misconception of those not familiar with clicker training is that positive reinforcement just means "food." This is incorrect. The crucial element in getting wonderful behavior out of a dog is NOT the food reward. The dog is not working for the treat; the dog is working for the click. The sound of the click is the magic signal that brings about that wonderful performance. The first step in training a dog using this method is to teach it that every time it hears a click, it's going to get a treat. Once the animal knows that the click means "treat is coming," the trainer can use the click to mark a behavior she likes, and then, gradually, to shape or develop something more complex, such as a response to a cue. Lets take an example from dolphin training. Suppose, on several occasions, a dolphin heard the whistle (click), and later got a fish, when it happened to be jumping. Soon it would start offering jumps every time the trainer showed up. Then it might be allowed to discover that jumping only "works" when the trainer's arm is raised. So a raised arm becomes the green light, as it were, for jumping.
The trainer could gradually impose other conditions--jumping only works when the direction of the jump is away from the trainer; when the jump is higher than four feet; when the jump occurs within three seconds after the arm is raised. At the end of a few training sessions the trainer has trained the dolphin to jump on command and with precision; and the dolphin has trained the trainer too; "All I have to do is make a certain kind of jump when she sticks her hand up, and she immediately gives me a whistle (click) and a fish (treat) every time!"
Note that the click is NOT used as a command. It does not tell the dolphin or dog to start doing something--the hand signal does that. The click tells the dolphin or dog, during or at the end of a behavior, that the trainer likes that behavior and the dolphin or dog deserves a treat for it. You don't have to stick with food, either; you can also associate a conditioned reinforcer with a pat, or a toy, or maybe just another chance to work.
The click has now become a conditioned reinforcer. In the language of psychologists, food, petting, or any other pleasure is an unconditioned reinforcer--something the animal would want, even without training; the click, a conditioned reinforcer, is something the animal has learned to want.
WHY THE CONDITIONED REINFORCER IS CRUCIAL:
Because of the split-second timing that the conditioned reinforcer makes possible, the click also communicates just exactly what it is that the trainer is looking for. This allows you to teach the animal what you want, in a very clear way, one detail at a time. For example, let's say that a dolphin has assimilated one rule (jump facing this way), and you know that because the animal almost always jumps with the proper orientation when you signal it to jump. Now you can add another detail or rule. You can decide, "I'll only reinforce the higher jumps." Pretty soon the dolphin has learned one more detail (I have to jump facing this way, and I have to jump high).
TO BEGIN:
Get yourself a clicker and a few treats. Make the treats small enough so that you can give the dog fifteen or twenty treats without filling him up. Some dogs will work for kibble, especially just before dinnertime, but you might have to go to something more tempting. Teach the dog the meaning of the click by clicking the clicker and giving a treat, four or five times, in different parts of the room or yard (so the dog doesn't get any funny ideas that this only works in one place.)
Then click the clicker and delay the treat a few seconds; if you see the dog startle and actively look for the treat, you will know the signal has become a conditioned reinforcer. Now you can establish a behavior -- this is called shaping.
Let's say you want to shape "chase your tail." The dog is now excited about earning clicks and treats and is likely to move around, and maybe even to whine and bark. The instant the dog happens to move or turn to the right, click and give the treat.
Wait again. Ignore everything the dog does except moving to the right (don't hold out for miracles; one turn of the head or one sideways step with the right front paw is all you need.) If you catch the behavior in three or four reinforcements you will see your dog turning to the right further and more often.
Now you will find you don't need to reinforce just a single step to the right, but can reinforce right turns that go several steps, perhaps through a quarter of a circle; from a quarter turn, a full circle may come very quickly.
Stop and put the clicker away with lots of hugs and praise, and try again the next day, starting with a single step, then a quarter circle, and then more; it will come much faster the second time.
From one circle, the next step is to get two circles, and the next step is to go for a variety, rewarding a half circle sometimes, then two circles, or one, or three, or one and a quarter. This keeps the dog guessing.
There are other behaviors you could use for practice, such as targeting, in which you shape the animal to touch some object with its nose. The purpose of this experiment is not to teach the dog the trick, but to show you how to use a conditioned reinforcer to shape behavior, and how effective this kind of reinforcement can be.
The reason the clicker is used instead of a voice command is because you can't say a word with the split second precision that you can achieve with a click. With the clicker and a little practice, you can reinforce very tiny movements--one paw stepping to the right--in the instant that it occurs; a praise word is inevitably fuzzy because it takes longer.
Also, we talk to and around our dogs so much it is difficult for them to sort out the meaningful words from the stream of noise we make. You will in fact see the difference very clearly in the way the conditioned dog responds to the clicker.
SOME ADDITIONAL RULES:
* Only click once. Resist the temptation to make multiple clicks for especially good behavior. Use a jackpot of several snacks to mark the behavior that earned the big reward.
* Don't use your clicker to get the dog's attention, or to get the dog to come.
* Don't use your clicker for encouragement, or as a start signal. You should always be thinking, "What am I actually reinforcing?" Click for success.
* Use the right size and type of food treat for your dog. Don't use things that take too long to eat--crunchy biscuits for example.
* Keep training fun. Toss in a few easy tasks when you are working on the hard ones.
* Once your dog has learned a new skill, or even just the beginnings of a new behavior, don't click for every correct response.
* Expect to progress in fits and starts. If a behavior your pet knows well suddenly falls apart, "go back to kindergarten."
* Initially, if you have more than one dog, separate them and train separately.
I highly recommend attending a Pryor-Wilkes seminar if possible, or finding a clicker trainer in your area to work with. There is a clicker list and a www page of trainers around the world (www.phys.unm.edu/~helix/trainers.html). This is the best method I have found for training dogs (or any animals) that are considered difficult to train.
Copyright 1997 Linda Michaels
