CAPÍTULO 4. METODOLOGÍA DE LA INVESTIGACIÓN
4.4 Análisis de los datos
Ian Dunbar, July/August 2006 1. Tautological—Punishment should be punishing
Punishments must effectively reduce the likelihood and future frequency of the immediately preceding behavior.
Repeated “punishment” is the best evidence that the “punishments” were inef-fective and therefore, not punishments at all. Instead the dog perceived the so-called
“punishment” as inconsequential, annoying, or abusive. On the other hand, when punishment is effective, the dog will no longer misbehave. Indeed, one definition of a dog trained to criterion could be, that punishment is no longer necessary. Check-ing the effectiveness of punishment (or any trainCheck-ing technique) should be ongoCheck-ing process, via a Test-Train-Test format.
2. Immediacy—Punishment must immediately follow the unwanted behavior
Ideally, both “crime and punishment” should occur in less than a second. The relevance of punishment depends very much on its immediacy.
3. Instruction—Punishment should also indicate the desired re-sponse
In addition to indicating “What is Wrong” (i.e., the immediately preceding be-havior), a punishment should also inform the dog “What is Right.”
Often, the desired behavior is evident from the specific environmental context.
For example, a punishment in the vicinity of a closed door, or another on-leash dog, usually indicates that we would like the dog to sit. Even better, the desired response may be re-emphasized by the wording of the reprimand. For example, “outside!”
would be a suitable instructive reprimand for a puppy caught in the act of eliminating indoors
4. Strength—Punishment need not be unpleasant, painful, or scary, it just needs to be effective
It is a huge misconception that punishments are necessarily unpleasant, painful, or scary and need be administered using some training tool or gizmo. On the contrary, a dog may be effectively punished using soft and sweet instructive reprimands. Other-wise, it remains an exacting art to precisely estimate the exact optimum strength of a punishment in any given training scenario; trainers constantly walk a knife edge—ei-ther they are too lenient and have to up the ante, or they are too severe and the dog
ished using soft and sweet instructive reprimands. Occasionally, when proofing for life-threatening situations (e.g., honoring open doorways, curbside boundaries, and an emergency sit/down response), stronger instructive reprimands are used.
5. Duration—Punishment should be of extremely short-duration
Extremely short-duration punishments are most effective, wherein a dog starts to misbehave, is immediately punished, and stops misbehaving (starts behaving as desired), all in less than second.
Continued misbehavior, however, usually requires continued punishment. Long-duration punishment is less commonly used, and comprises two separate psychologi-cal processes: positive punishment and negative reinforcement. The instructive aspects of a long-duration punishment are the onset and the offset: commencing punishment informs the dog he is making a mistake and terminating punishment informs the dog he is back on track again.
6. Warning—Punishment should be preceded by a warning
If not warned beforehand, the dog cannot learn the meaning of the warning, which would be inane. And if not warned beforehand, the dog cannot avoid the pun-ishment, which would be inhumane.
A dog must have adequate opportunity to avoid punishment by acting appropri-ately as previously trained. For punishments to be really effective, the dog must first be taught acceptable alternatives, and, in all instances, the appropriate instruction (request/command) must be given before the dog is punished for not responding as desired. In a sense, the request (command/stimulus/antecedent) now becomes a warn-ing, as we now enforce (without force) a previously learned response.
7. Prior Training—Punishment should be preceded by adequate prior training
Before being punished, the dog must know what the owner expects him to do, i.e., if this is wrong, what is right? Therefore, before punishing the dog for making mistakes, the owner must first teach the dog the appropriate response.
Because there is only one right response, teaching a novice dog what we want him to do takes only a very short (finite) amount of time. On the other hand, there are an infinite number of “wrong” responses and so, attempting to teach a novice dog by punishing undesired responses takes an infinite amount of time—the Myth of Sisy-phus! Hence, with novice dogs, punishment-training is much less time-efficient than reward-training. And time is of the essence in pet dog training.
8. Consistency—Punishment must be consistent
In order for punishment-training to be effective, the dog must be punished each and every time she misbehaves. A dog only has to get away with misbehaving once to learn that there are times when she is not punished. Henceforth, the dog will reserve
her misbehavior for times when the owner is absent—either physically, mentally, or functionally.
The eighth criterion illustrates why punishment-training is less effective than reward-training. Most learning theory experiments on the use of punishment were performed by computers training laboratory rats and pigeons. However, no owner is 100% consistent 100% of time. Whereas some experienced competition and working-dog trainers may achieve close to 100% consistency during an intense short-duration training session, most dog owners are highly inconsistent living with their dogs at home.
Because of the difficulty in satisfying the eighth criterion, punishment-training methods are seldom expedient for pet dog training. Punishment-training is reserved for highly specific and short exercises (during which we hope the owner/trainer can at least approximate 100% consistency) for enforcing (without force) previously learned commands, and for absolute proofing in life-threatening situations.
But…the most wonderful aspect of reward-training is that inconsistency does little harm, and often does a lot of good. Occasionally, a dog may be rewarded for incorrect responses (temporarily increasing time and trials to criterion), but the mis-timed rewards still serve to cement the bond (classical conditioning). And of course, inconsistently rewarding a dog for correct responses cements the reliability of the de-sired response. Variable reinforcement ratios, and even utterly random reinforcement, are excruciatingly effective. v