Many reward-based dog trainers are familiar with Murray Sidman’s writings about punishment and its fallout. One impact, which I have been teaching people about for many years, is the fact that punishment (or bad experiences as a whole) generalise more easily than new behaviour learned by means of positive reinforcement. In particular, punishment generalises to the location where it occurred. I think I must have read this in a Psychology text book over 20 years ago, and I don’t have the reference. Think of the implications for dog training. If a dog training school uses correction or aversive methods, the corrections can generalise to the environment – namely, the training grounds. What impact do you think this would have on the dog’s motivation and learning ability?
Now there is recent scientific research that explains this effect, which has been known from experience for a long time.
Clearly in our evolutionary past, generalising a bad experience to the location where it occurred would have had survival value. You might need many experiences to learn where to find food, and to generalise that knowledge. (We know that dogs need repeated experiences to generalise a behaviour learned from positive reinforcement – take Ian Dunbar’s “does your dog know what sit means?” exercise). But if a predator jumps out from behind a fruit tree where you are gathering food – on one occasion – you would be well advised to learn from that one experience, and avoid that place in future.
If it turns out that the predator does not normally lurk there, it may take many experiences of being safe before you “undo” the memory of danger. This is why we need to give our dogs repeated safe exposures to something that they are frightened of and want to avoid. There is a good reason to err on the side of safety.