R.E.A.L. Variable Reinforcement - Part 2
Last article we worked on the first and most critical part of using variable reinforcement - you. This month we will move to the next step in the process and work on your partner - your dog. First, however, I need to clarify something for those of you who have some experience with clicker training. The topic is variable reinforcement, not variable ratio schedules of positive reinforcement. (VSR) That comes later, after the behavior is firmly established. For the moment, we are still trying to get the dog's behavior to "wiggle", so we aren't really asking for two versions of exactly the same behavior - that comes later.
Now that we have run through our list of 30 arbitrary and variable reinforcements, your dog may be pretty confused. If you are a good trainer, you may be a little confused, too. First, we started shaping a behavior to the point where is was getting pretty consistent. Then I asked you to use a whole string of unpredictable reinforcements that caused the behavior to get pretty flaky. Your goal during the random reinforcement project was to watch how your dog's behavior changed as the consequences changed. Now that you have seen what variability can do, it is time to "get real" and start applying what you have learned. I want you to go back to the same behavior you used for your random reinforcement project and apply these four rules. This time, we are going to add some method to our madness and use these rules to make the reinforcement match the performance. Once again, your goal is to apply the rules and pay very close attention to how your dog's behavior changes...
R) Raise your standards: Make it harder to earn a click and treat. Ask for two or more repetitions of the behavior, faster performance or more enthusiasm before you reinforce the behavior. Be opportunistic and hit any of several criteria with bonuses. A bonus might be as much as ten times a normal reinforcement. Check your random reinforcement list for ideas of how to contrast average reinforcements and bonuses. As the behavior starts to take shape, try to focus on one criterion at a time. Example: If you were working on "faster" downs, stop worrying about "straightness" for a while and stick to speed.
E) Extras for excellence: Start giving relatively large jackpots for exceptional performance. It helps to start the shift to variable reinforcement with a pretty big "positive kick in the pants." While you might assume that the jackpot only acts to identify which version of the behavior pays off most, it also acts to sustain the animal through the dry spells that are caused by raising your standards. After each repetition, ask yourself the question, "Was it better or worse than the most recent repetitions?" If it was better, give extras, if it was worse, give nothing, and if it was a flat out mistake, say "wrong" and try it again.
A) Anticipate errors: As you vary the reinforcement, the dog will start to vary its behavior. This experimentation is necessary in order to start moving away from constant reinforcement. The dog will adapt to this new set of rules by trying to figure out why the old level of performance is not working, and why some repetitions cause fantastic payoffs. The dog does not immediately know that you want a faster "down", or a quicker "sit". The animal will make mistakes as it tries to figure out the new rules. These "errors" are similar to the behavioral "wiggling" that you saw last month with your random reinforcement project.
L) Lots of repetitions: Allow the animal a chance to practice the behavior enough to get comfortable with the new standards. Your immediate goal in varying the reinforcement is to get lots of repetitions, not to get great performance. You are really trying to communicate "if at first your don't succeed, try, try again." Over a series of many repetitions, you will see the wiggle start to straighten out. The dog's performance will start to take on the new higher standards, while less perfect behaviors will start to fade.
NOTE: If you do not see garbage, you will not see greatness. Varying the reinforcement will cause erratic behavior. If you trust the four rules of variable reinforcement, you will soon see great performance. You must allow the animal to experiment and adapt to the new standards. Do not be disappointed with the appearance of failure, as long as the animal keeps trying. If you lose the behavior entirely, drop your standards, reinforce more often until the behavior comes back. Once you are getting consistent behavior again, go back to varying the reinforcement and raising your standards --- more gradually, this time.
After you have run through "getting REAL" for a couple of training sessions, you should see some obvious changes in your dog's behavior. One of the most noticeable changes is the willingness of the dog to work for longer periods in the absence of reinforcement. This persistence is a critical link in having a dog who is willing to learn and perform difficult tasks. Trainers who are afraid to broaden the spectrum of reinforcement create animals who are reluctant to experiment and unwilling to persist without constant reassurance.
NOTE: If you start this exercise with a dog who is used to very consistent reinforcement, don't be surprised if the behavior momentarily disappears. Don't worry about it. The behavior hasn't been erased, it's momentarily submerged under your dog's understandable frustration. If the dog gets frustrated and stops offering the behavior, drop your standards and reinforce more consistently for a few repetitions. Make the transition as gradual as necessary to keep the animal trying. If you absolutely can't get anything going, play a simple game like fetch, go for a walk or simply end the session. When you come back later, the behavior and the dog will be refreshed.
Go to the next article | Go to Prior Article