Non-Linear Learning: Teaching a good "Finish."
For most of our lives we are taught to think and behave in a linear fashion. For example, if I say "A, B, C..." the most likely thought that pops into your head is "D." If I say "1,2,3" your first thought will most likely be "4." While some fundamental behaviors can be acquired through a linear process, complex behaviors are often beyond this limited perspective. Though learning the ABC's is a required part of any decent education, it is only a stepping stone to a more complex behavior - written language. It is obvious that though reciting the alphabet is linear, language is not. To compose sentences, paragraphs and essays, you must leave linear thinking in the dust and develop a larger repertoire that includes combining the basic units of language (vowel and consonant sounds) in many different ways. With clicker training, the metaphor rings true - those trainers who continue to think in a linear fashion are like children who know their ABC's perfectly but can't make heads or tails of D-O-G.
To see how this applies to training, consider teaching a dog to do a finish using a linear approach. According to psychology texts, you would wait until the dog offers the tiniest indication of a move around your right leg and reinforce it. You are then supposed to reinforce any tiny improvements until the dog is doing a perfect finish. While this process is touted as the mainstay of operant conditioning, called "successive approximation", it is often impractical. The trainer who uses this method exclusively will spend long unproductive periods waiting for something to reinforce and will ultimately take hundreds of unnecessary tiny steps to reach the goal.
In contrast to linear learning, there are other approaches that more fully utilize the dog's mental abilities. One such form is called "component/composite" learning. In this methodology, individual simple behaviors are used as components to build composite behaviors, the way alphabet letters are used to build words. The ability to quickly combine behaviors into unique patterns offers several advantages. First, dogs that are skillful at combining component behaviors can fluently recombine components to create new behaviors quickly. Second, by learning component behaviors, the dog can carry knowledge from one setting to another with ease. Third, combining simple components allows for rapid repair of behaviors that have started to decay.
To see how this approach works, consider a dog who is familiar with two component behaviors - following a target stick and "sit" at "front." From the front position,(See illustration, Position A) the handler holds the tip of the stick out to the right of her body. The dog can see the target and, on cue, immediately follows it (Position B) around the handler's body to a position that is roughly at heel. (Position C) Over about 30 repetitions, the target stick can be faded so that the dog is initiating the behavior on his own. This can be done by gradually sliding the stick back in your hand so that less and less of it extends down as a target.
With this type of training, the handler has combined the initial components of a "finish" into a rough example of the behavior - in about 15 minutes. Because the dog is already competent and experienced at targeting, (like your ABC's) the dog does not need to go through many successive steps in order to quickly perceive the goal of the exercise.
(NOTE: If the dog falters and doesn't follow the stick correctly and therefore doesn't go all the way around to heel, you would, of course, have to shape the behavior in linear increases over a series of repetitions. All the more reason to make sure the dog really knows targeting, so that you don't have to resort to wasting time. )
If the idea of skipping all those tiny steps seems unnerving, don't worry about it. Rather than causing problems, getting to the goal of the behavior quickly bypasses a very serious training hurdle. When trainers take unnecessary baby steps, a behaviorally sophisticated dog gets bored silly. This would be like expecting a speed-reader to tick off the letters of the alphabet every time you spell a word, like thinking ABC-D ABCDEFGHIJKLMN-O ABCDEF-G to spell "dog."
The next step in our non-linear "finish" requires that we use something called "behavioral adduction." Simply stated, adduction is a three stage process that includes 1) gaining control over two separate and distinct behaviors 2) creating a slightly ambiguous situation where both behaviors are likely to be triggered at the same time which then 3) causes the spontaneous blending of the original behaviors into a third, unique, behavior.
To see how adduction can be used in a constructive manner, let's return to "finish", which we left hanging with the dog standing at heel. (Position C) The next obvious step is to teach the dog that a "sit" is added to the end of the behavior and that the dog's backbone must be aligned with the handler's direction of travel. In order to keep a good eye for position, I recommend making a capital "T" on the ground with chalk or light-tack masking tap. Stand facing the dog just behind the upper right hand cross bar of the "T". (see illustration)
In order to use adduction, we need two separate behaviors - 1) sit and 2) walk a small, clockwise circle. We have used the target stick to teach the dog to move in a loose circle around the handler but the behavior is hardly strong enough to be considered a stand-alone component. If you see that the dog is confused by the next stage of training, return to the turning behavior and practice it until it is solid. You may include your body position as part of the behavior, but the following paragraph assumes that you are standing facing the dog and the dog is doing a circle in front of you. For the purpose of this article, we'll call this behavior, "Circle."
Start with the dog at front and take a step backwards. This leaves you some room for the dog to circle in front of you. Practice having the dog do some simple following in circles, figure 8's and "s" curves. Gradually slip into exclusively circles. After you have a fluid session going, pick any repetition and say "sit" as the dog is following the stick. If the dog immediately sits, click and treat. Repeat until the dog is listening intently for the "sit" cue while following the stick. What you will see after a few repetitions is a circling/sit. This is not meant to be the final form of the behavior, so don't panic if the sit is a little crooked or the dog's butt slews around as the dog turns and sits at the same time. You can control either of these reactions by changing the timing of when you ask for the "sit" and by using "wrong" to tell the dog what doesn't work. Now attach a new cue to this new, composite behavior - any word will do. Present the cue just before you say "sit" about 20 repetitions and the dog will start to connect the new cue with the new behavior. Next, fade out the word "sit".
Now that you have a new composite behavior (circle and sit, on cue) position yourself on the "T" mark to put the behavior in context. Regardless of your particular goals for "finish," you will now be able to carefully watch the dog's behavior and manipulate the timing of the command "sit" to affect your dog's alignment. Have a friend stand immediately behind you and click the clicker for sits that are prompt and straight. Sometimes it helps to have the assistant trigger the cue for the sit. If the dog sits too quickly or too slowly, say "wrong" and try again. If the dog twists his back excessively, say "wrong" and do it again. If the dog sits slightly behind you, place your target stick in a spot that will cause the dog to move a tiny bit farther forward as you trigger the turn-sit. This is also an example of adduction, as you are triggering an immediate "sit" while telling the dog to touch the target - two "commands" at the same time that cause the dog to blend both behaviors into a turning/sit, farther forward.
If you are already happy with your method of teaching "finish", this process has another application that may interest you. Take your "turn/sit" and connect it with a simple go-out to a standing target. Try these steps...
1) Place a remote target about 10 feet away from you.
2) Say "touch" or whatever signal you've selected for targeting.
3) As the dog goes to the target and bumps it, give your cue for the turn/sit.
4) Build some momentum so that there is a very short gap between repetitions.
5) Move the target farther away from the dog.
6) Send the dog out and give the "turn/sit" cue BEFORE the dog gets to the target.
7) Remove the target. (Have an assistant pull the target in between repetitions, unnoticed, while the dog is at "front," looking at you. ) Give the cue to "go" and arbitrarily give the signal for "turn/sit" at some point in the go-out.
8) Practice longer and shorter distances so that the dog pays attention to the cue and will effect the turn/sit at any distance up to about 30 yards.
Now you can send your dog out to a distant target, give a cue and the dog will sit and turn to face you. NOTE: If you teach this behavior after using the "turn/sit" for "finish" you may have to go back and tweak the "circle" behavior to make it a complete 180 degrees, before you try to use it to teach a distant turn/sit.
Teaching your dog a broad selection of components allows for greater flexibility in training. As your dog becomes comfortable with combining and recombining components into more complex, composite behaviors, the learning process speeds up considerably. While a simple ABC type knowledge is a necessary ingredient to learning, teaching our dogs to utilize non-linear combinations of behaviors can lead to patterns that would otherwise be indecipherable -- such as CD, UD, and OTCH.
Go to the first article | Go to Prior Article | Go to Next Article | Back to Clicker Project Index