FIXING BEHAVIOUR PROBLEMS
by Ed Bailey
Printed with permission
First printed Gundog magazine July/Aug 2006
Question: Your articles on training were very interesting. But, there was one thing I didn’t see that I have heard so much about, “clicker training”. I would like to hear your views on it and if it really works as some people say. It would be very interesting to see an article on it.
Reply: It is called “clicker training” only because the clicker became the popular sound maker, not because a clicker is absolutely essential for training, just a popular, useful tool. Clickers used to come free in a box of Cracker Jack. Now that dates me. We could also buy them at the local 5 and 10 cent store for a penny. The large sized ones cost a nickel. When they became a popular dog training tool the price jumped to 4.95 and up. The click-clack sound they make is from depressing and releasing a spring steel insert. As kids we called them crickets because they were always shaped and painted like bugs of some sort. My second grade teacher had a drawer full of confiscated clickers long before they were popular as an animal training tool. Today’s high tech clickers are a rectangular plastic box with the anchored at one end spring steel making the click-clack sound.
The clicker training technique reached its popular status first with the people who trained dolphins and whales to do their entertainment routines. They used whistles rather than clickers because the sound carried better, including under water. Dog trainers use whistles and/or a word or two in exactly the same way the clicker is used. A very good dog trainer friend always uses ‘good boy’ or ‘good girl’, as the case may be, instead of a clicker and it works perfectly well for him. The advantage of the clicker is that it is consistent in pitch and intensity. A whistle has the advantage of carrying farther and is a hands free in the mouth tool, but pitch and intensity is variable depending on how hard it is blown. The spoken word is quiet, hands free and even more readily available than a whistle or a clicker so can be used quicker than either of the other sound makers.
The learning (training) in clicker training techniques is a two step process. The first and most necessary step is to establish whatever sound you use as a ‘conditioned reinforcer’. This parallels Pavlovian Classical Conditioning in that an arbitrary sound (click, whistle, words) that is originally meaningless to the dog is paired with a meaningful primary reinforcement such as food treat, praise, petting, so the sound becomes associated with the primary reinforcer to the point where anticipation of the primary reinforcer acts as a reinforcement in its own right and when it has reached that status, the sound is called a conditioned reinforcer. Though sounding complicated, the whole association process takes only a few minutes. The more often the conditioned reinforcer and the primary reinforcer are paired, the stronger the association that is formed. How often is enough? It depends on the dog and on the consistency and timing of the pairing. Twenty or thirty times is plenty usually. Fifty would be overkill. Thereafter, when using the conditioned reinforcer in a training session, four or five repetitions of the pairing will bring the previously conditioned dog up to speed and training can proceed. The important thing to remember is the sound must be presented immediately before or coincidentally with the primary reinforcer for association learning to occur. Never have the sound follow the primary reinforcer. And, the renewal of the association through periodic reinforcement with the primary reinforcer will go on every time the dog is subjected to training, including refreshers.
To reiterate, a conditioned reinforcer is an arbitrary signal, initially without meaning to the dog, that is deliberately presented just prior to or coincidentally with the delivery of a primary reinforcer so it becomes so intimately associated with the primary reinforcement it acts in the same way as the primary reinforcer. The precision in the timing of the pairing and the number of pairings maximizes the effectiveness of the conditioned reinforcer. The conditioned reinforcer can be anything, but a sound of some sort is used most often.
Once established a conditioned reinforcer can serve several different functions useful in training the dog to perform complex behaviors: (1) It can serve as a terminating stimulus telling the dog the job is completed correctly. Dropping a bird into your hand after a retrieve is an example of when the conditioned reinforcer can act as a terminating stimulus. (2) It can act as a bridging stimulus to tell the dog he has earned the reward and will get it as soon as he totally completes the task. An example of this use would be when training the dog to come to heel on command. He is given the conditioned reinforcer when he starts toward the handler. It acts to sustain the dog until he gets to heel and can receive the reward of a treat or praise or whatever. (3) It can act as a keep going stimulus. For example on a long fall when the dog stops short, he is given the conditioned reinforcer if he looks or moves slightly in the correct direction. Each move in the correct direction is reinforced. Even if the dog is a long way out and you can’t give him a treat or praise or a pat on the head when it is needed, you can give him the conditioned reinforcer and it is just as good for rewarding him as the primary one. But all this is just to establish the conditioned reinforcer. To use it effectively you must now go the second step of the training – shaping.
Shaping is defined as taking some very small tendency in the direction the trainer wants and shifting it, one small step at a time toward some predetermined ultimate goal. This has been referred to by learning psychologists as ‘successive approximation’. Shaping is accomplished by reinforcing correct steps, no matter how small they are, and ignoring -not rewarding- incorrect steps. The conditioned reinforcers are used to instantly and effectively reinforce each correct response. Reinforced behavior tends to be repeated, non-reinforced behavior tends to diminish and disappear; it extinguishes. Shaping, then, is a learning process based on instant reinforcement of every correct step to enhance performance while ignoring mistakes so they tend to drop out.
Very complex behavior sequences can be learned by a dog, but only through the process of shaping. It is always based ultimately on some presently existing behavior to use as a first step and then, no matter how elaborate the final behavior you want to shape, you can find the first step and establish a series of intermediate goals to reinforce until you can get the performance you want. So how does it wok in the real world?
I mentioned above how a conditioned reinforcer can be used as a bridging stimulus when training the young dog to come on command. Let’s say we are using a clicker because it as at home in the yard early on a Sunday morning and you don’t want to wake the neighbors by yelling or blowing a whistle. So you can use the clicker. First, jump start the dog by a series of clicks with instant treats to bring the dog up to speed. When your pup, at 3 or 4 months old, is wandering out about 20 feet or so from you, call his name and tell him ‘come’. Then watch very closely with your thumb poised on the trigger of the clicker. When he turns to face you, click. Wait, a step or two toward you, and click again. Each move in your direction, click, until he is right up to you then click and give him a treat. Each correct move is reinforced, each incorrect move is ignored. Do not say ‘no’ or make any correction as it will be a reinforcement for going the wrong way. After about 15 minutes or so let him get 30 or 40 feet from you then call his came and say ‘come’. Again, click on his first move toward you, then wait until he moves a yard or so toward you before clicking again. Make him work a little harder to get the click. After a few repetitions he will come running when he hears the word ‘come’. Again, any wrong move anytime along the way is ignored, totally. It won’t be long before the dog is doing things to make you click even when you don’t give the command. Unsolicited behavior must also be ignored so the pup learns to do things only when told to and then he will get the reinforcement. All other times count for nothing.
Or, lets say you want your dog to let you know when he needs to go outside to do his toilet chores. Hang a bell beside the door to the outside for him to ring to tell you he needs to go outside. Training him to touch the bell with his nose is called targeting. It is accomplished by simply clicking any movement toward the bell and continuing one step at a time until he touches it. Then click and reinforce with a chunk of cheese. Teaching walk at heel is done by having the dog target the end of a stick with his nose. Then lead him with the stick, reinforcing him with a click when his nose touches the stick. Or you can have the dog target on your hand when you walk if you don’t want to have the stick in his face..
Clicker training techniques are useful for teaching obedience or for teaching parlor games, true. But they can be very useful to teach hunting skills also. Like retrieving. They are also useful in training how to be still for a host of things like nail clipping and grooming. But the problems come when the dog realizes that if he does a certain behavior like sit, he can make you click. Training is not complete until the dog does the chosen behavior only on command, doesn’t do something else such as lie down when told to sit, or doesn’t sit when told to do something else. To avoid this, a cue must be learned and only that cue evokes the behavior and is therefore reinforced. All other things are extinguished by withholding the reinforcement.
The best feature of shaping with the use of conditioned reinforcement is the dog wants to do it, wants the reinforcement and so is happy in the performance of the trained behavior. There is no coercion, none of the sulking, depressed, automaton like demeanor seen in dogs trained by negative reinforcement. Clicker training techniques, shaping with positive reinforcers, have been used in a wide variety of species from polar bears and horses to dogs, cats, rabbits, and even fish and chickens, and of course dolphins and killer whales. The technique is also used for teaching children, especially those with learning disabilities, and know it or not, it has been used on us in one way or another all of our lives. All become more playful, more relaxed, more intelligent in that they can learn faster, are more curious and more bonded to the trainer than when any type of negative training was used . The dog learns to think. The absence of punishment in the training makes a more happy and less fearful dog. Long term recall of learned behaviors is much improved. A competent clicker trainer gets more complete training, faster than with any of the conventional methods.
So, does clicker training work? You can bet on it. Is it as good as people say? It is gaining in popularity so you can pretty well assume that it is a good and efficient technique. The clicker itself has some good points and some limiting problems. The quick, easily given sound is consistent in intensity and pitch are pluses. The distance the sound can travel is a limiter. I don’t hold out much hope for a clicker in milo stubble or standing corn with a forty mile an hour wind rattling each stalk. Even a whistle would be compromised, but not eliminated like a clicker would be. A clicker is most efficient in a small space like a yard or inside a house. A clever use of an e-collar would be to just use the beeper signal as a conditioned reinforcer, but never mash the shock button. It would be very effective at long range and in the presence of background noise. Frequency and intensity would be constant also. The beeping locator collar would always have to be left at home, however. It would totally confuse the whole issue.
An amazing thing about clicker training is that Pavlov set out the basis for conditioned reinforcers a hundred plus years ago, and B.F. skinner was shaping rats and pigeons in the 50s. In 1960, he even suggested that a clicker would make a good conditioned reinforcer for shaping dog behavior. Sometimes dog trainers are a bit slow in adapting to new suggestions.
For solutions to your dog’s behavior problems or behavior related training problems, you can contact Ed Bailey at email@example.com
Updated 2nd January 2007