ClickerSolutions Training Articles


This article was originally posted on the ClickTrain mailing list. It is reprinted here with the author's permission.

We trained “wrong” signals purposefully.

We presented a two-object discrimination. On a variable ratio and early in the discrimination training, before the animal has mastered the discrimination, we would suddenly change criteria, and what was HOT was then COLD, and vice versa.

When the animal approached hot, we sounded the wrong signal. At first the animal would go through with the now-incorrect response and NOT be reinforced. Since the animal had not yet firmly learned the discrimination / object association, it would soon go to the other object, and be reinforced. This scenario would be repeated until the animal would clearly hesitate at the wrong signal, and most of the time, go directly to the new hot object.

At that point, we would add a third object. The same pattern would be followed, that is, the correct answer would be changed without notice, and the first indication of the change would be the wrong just before the response. This time, of course, there are two more choices to be made. At first, the very first choice the animal would make after avoiding the now cold object would arbitrarily be considered hot and be reinforced. A little later, we would give a second wrong on the same run and wait for the animal to respond the one remaining object.

Later, we would add a 4th, 5th, and 6th object. If we had wanted to, we could have developed the discrimination to the hot object by using ONLY the wrong signal -- and then we could have really been in trouble, and that is what people have a tendency to do with the wrong signal. It seems to work so well, that it becomes overused, and the trainer forgets about the advantages of HIGH RATE OF SUCCESS.

There were some other ways we used to teach wrong, but this was the most straightforward. Other approaches involved feeding the animal just when it hesitates before responding to the newly cold object. We usually want the animal to scan about for the right thing to do, so we seldom terminated a trial with wrong, even if we reinforce wrong directly. We give the animal the opportunity to respond to a hot object, and then terminate.

It was also an exercise we could give a new trainer. It was tough to screw up as long as the protocol was followed carefully. Developing this behavior is really neat with some animals -- primates, dolphins, dogs (the brighter ones anyway, not so much the dumber ones), and cats (and there is a real test for the trainer because the cats don’t take much to high frequencies of wrong, especially on extinction), the parrots, and -- well, I won’t list all we have tried. Chickens and rabbits are NOT much into wrong, though some of them sometimes get it after a long while, and in simple situations. What is so neat with the animals that DO is that you can almost see the light turn on when they GET THE IDEA. When the light does come on, the signal generalizes very well, and wrong can be a powerful tool in shaping and directing behavior. We certainly found it useful in open environment work, especially with dolphins.

Bob Bailey
copyright 2002 Robert Bailey


| Training Articles Contents || Site Home |

Copyright of all posts is the property of the original author. Please obtain permission from the original author before copying, quoting, or forwarding.

List and Site Owner: Melissa Alexander, mca @