Bomb proofing mistakes

Isn’t it surprising how easy a clean conduct varies and yet how hard it is to modify flawed conducts? Proofing mistakes embedded within the trained conduct is a common problem among animal trainers of all levels.

Take a defective heeling behavior for instance, when dogs either walk crossing their handlers’ path, or behind or in front of them. A further example would involve those far from brilliant static shifts of position, when the latency before offering the behavior is really terrible. Why is it so difficult to get rid off the problems just described? It is as if it was only easy to modify a behavior that is satisfactorily shaped from the start!

So what is the explanation? As it is usually the case in animal training, the truth is that it is our fault. There are two reasons behind.

On the one hand, the technique applied to shape the behavior is of the utmost importance here. We should keep the final behavioral goal in mind and refuse any beginnings which are incompatible with it. Trainers may experience something similar to the “blank page syndrome” when they start working with a new animal. Just as writers do, it is normal that we are fraught with anxiety when contemplating the long road that lies ahead of us. There remains so much to do before being able to present our work that the task looks daunting!

When we face difficulties the temptation to take easy shortcuts is strong. We don’t want to get stuck with that particular exercise and hence we settle for a downgraded version of the original behavioral goal. In a previous post (“Less is more”) I wrote on this subject. What I want to emphasize now is the importance of reflecting on the influence that today’s training session will bear not on the training session tomorrow but in those that will be carried in two or three years time, once most of our work plan will be in place. In other words, will then the lessons taught today help us or will they hinder further learning so that it will be necessary to clean up the dog’s head before progressing?

Let me illustrate this point with an example: imagine that I want my dog to lie down by moving backwards in order to make the final behavior look faster in execution as well as to prevent the dog from gaining ground when he is supposed to keep a perfectly static position. Having set that objective, suppose that I start rewarding my dog for offering a two-stoke lie down. In other words, I reward the dog for sitting down before lying down. My laziness will increase the amount of work for both of us at a later stage. Thus it would be wiser to be a bit more patient and wait for the particular lie down behavior I am after. In fact it would be a far better policy to reward an approximation to the correct behavior than settling for a complete but flawed behavior.

On the other hand, there is a further and more fundamental problem for even after teaching a behavior with mistakes to a dog, there must be the possibility to get rid off the flaws. Theoretically, it suffices to reinforce those instances where the dog gets closer to the set behavioral goal. Why is it then that the mistakes become sort of bomb proofed?

“Frailty, thy name is dog trainer.” May be we just attended a competition (either as part of the audience or as competitors) and with the boost in motivation we adopt the firm decision to become stricter in order to correct the problematic heeling. We start training without rewarding any flawed repetitions. However, after a busy week, we lower down the criterion once again and settle for the behavior as imperfect as it used to be. Subsequently, perhaps we attend a training workshop where we learn a wonderful technique than can help us in sorting out the problem. Again we experience a boost in motivation and start training with a stricter criterion in mind. But progress may be made at a slower pace than expected, so we relax the criterion again and go back to square one. To feel better we may think that John Smith had a flawed heeling behavior too and he won the world championship anyway. Few persons ask themselves if the rest of the pack is as good as John’s…

Anyway, the point is that this sort of vicious circle goes on and on. At the end of the day the result is that the incorrect forms of the conduct are subject to a variable reinforcement schedule whereas correct behaviors are typically kept under a continuous reinforcement schedule. Remember that behaviors on an intermittent schedule are less susceptible to extinction. That is why we cannot get rid off the mistakes no matter how much effort we think we are putting.

The conclusion is that you should never train behaviors under intermittent reinforcement schedules until you have reached their definitive form, at least if you work solely on conditioning processes. Otherwise the behavior will be harder to modify. This is an extremely important piece of advice for new trainers, but one that is often overseen. In fact, in Spain I have heard just two instructors regularly underlining this fact in their classes, Carlos Bueren and Javier Moral. I think it is an extremely important point that we all should emphasize and clarify to novel trainers.