Whoo, do I get to play devil's advocate here?
I actually agree in part with both of you, 4IC and Ron, if that's possible. So perhaps I'm playing devil's advocate for both....muaha.
I've seen a lot about the "punishment must be severe enough the first time, and no more than 3 times, or else it's not effective/is abuse. I've likely even said it a few times, but I'm beginning to question it myself as I grow and begin to think about it. Not that I'm any more apt to use it, but because I'm a critical thinker and like to question everything I read *G*
Firstly, the overall basis of the behavioural definition, as I've learned it in university in learning theory, is pretty much as 4IC states. There is nothing in the definition that states that behaviour will ever actually *disappear* (that's just the hope of dog trainers and dog parents!), or that it has to do so quickly in order to be considered punishment. Technically, any decrease in behaviour is considered punishment.
Then I look at the realism of using either reinforcement or punishment. Both are required throughout life in order to maintain behaviour. Behaviours that are not reinforced somehow, and that are not self-reinforcing, will eventually become extinct. In other words, even intermittent reinforcement is required through the life of the dog, in order to maintain behaviour.
I think the same goes for punishment, to be honest. Dogs who are trained using mostly punishments (I'm not pointing anyone out, but we all know somebody who has taught this way), or even just some punishments, tend to periodically need punishments throughout their lives as well to maintain the behaviours that they know (which includes "not doing something";).
For instance, the dog taught to heel, by being collar popped with a choke chain every time it is out of position (I'm using this example because I happen to know someone who teaches it this very way, so it comes to mind quickly), and the dog learns to avoid punishment, it stays in the heel position. Anyhow, that dog, too, will periodically require punishment throughout its life, in order to maintain the behaviour of staying in heel. It may be rare, but it's there, just as reinforcement of some sort is required to maintain behaviour. Without some contingency, the behaviour will not continue and the dog's heel would become more sloppy.
I also think that just like with reinforcement, once the context of a situation changes, the dog's understanding of the situation changes as well. So a dog who has been punished for jumping on people in the home, might not bring that out doors to say, a park, and thus will need to experience the punishment there as well. When we say dogs tend to be poor generalizers, I think it applies as much to punishment as it does to reinforcement. Dogs need to be "proofed" with punishments just as they do with reinforcements, so that in effect would require more than 2 punishments, as the context shifts.
I do think that the author of this article is correct in saying, though, that there is an inherent downside to using escalating punishments, and I think that is of more importance that whether it stops it in 2 or 3 tries. In other words, if you are going to do something to lessen behaviour, make sure the dog actually cares about what you are doing, rather than starting out small and going harder, and harder, and harder. And I do agree that via a cost/benefit analysis, if you are to decide to use a punisher, it does have to outweight the value of the reward. That might seem obvious to some, but in order for a punishment to be effective, it has to have a higher cost/benefit ratio than the reward does, because if not then the punishment certainly won't be effective long-term.
I have a sneaky suspicion that along with a variation in natural pain sensitivities, a lot of these dogs you see lunging in prongs or pulling in chokes, are dogs who have experienced either of the following:
1) Slowly escalating punishments such that the dog becomes habituated to even some of the most forceful punishment attempts. In fact, I believe at one point this is precisely how they trained protection/police dogs, so that they would not be bothered when people beat the dogs or hit the dogs, or try to hurt the dogs by force - the dogs had been habituated to force, and as such didn't take it as a punishment. It's also how a lot of dog fighters train, purposely, dog fighting dogs so that they will continue on in the most painful situations.
2) A dog for whom the cost/benefit ratio is higher for the reward than it is for the punishing effects of the prong/choke (these are just examples).
Even with my example with Shimmer and pawing, for her the cost/benefit ratio of pawing was greater than for the temporary removal of attention, so it was very slow to change. But in ignoring the pawing, and then heavily reinforcing the "sitting without pawing", the cost/benefit analysis of "sitting without pawing" become higher than the pawing, so it became the overall behaviour.
It's like I've seen with one lady working with her dog.....the punishments she was actually using was rather benign, such that it's not the intention of the woman that resulted in the ending of a behaviour. What she thought she was doing, was not seen as a punisher to the dog. Rather, it was the constant nagging, the sheer annoyance of the woman, that actually ended up acting as the punisher....lol. The woman didn't do anything inherently aversive, but the fact that she continually nagged over a certain behaviour, caused the dog to lessen the behaviour overall, and the cost of doing the behaviour became higher than the benefit from doing it.
It's like a child learning to stop leaving things around *G*. The nagging mother tends to win out in the end, not by being inherently aversive, or doing anything painful/fearful to the child, but simply by becoming an annoyance to the child such that the child learns to stop leaving things around to avoid the nagging of the mother. Some children (and some dogs) handle nagging better/worse than others, which is why it is not usually an effective strategy to recommend as a whole!
I think the issue with "2 or 3 punishments" is that if you have to give more than 2 or 3 decent punishments (decent, to mean effective) in a row, for a given behaviour, you have a problem with the type of punishment that is being used. Or if you have to use the punishment every day, in and out, there's a problem with the punishment. I'd even wager a guess, to stretch it, to say that if you don't notice some sort of meaningful decrease in behaviour after 2 or 3 tries, then it is not a very effective punishment. That might be a better way to put it (or it might not, I'm still deciding.....hah)
It's the same way with if I had to give a cookie every time my dog sat, and the dog wouldn't sit without a cookie, I'd have a problem with my level of reinforcement and the teaching. And that I can understand. There is no reason that anybody should be giving 5, 10, 20 punishments for one behaviour during a given situation, no more than there is no reason that once a behaviour is learned, I should be giving treats every time my dog performs a behaviour that I ask (although I rarely ask for a behaviour without it having some contingency, usually a life reward, - it's a way I live my life with dogs, so that they always have the option to make choices). Perhaps in 3 hours if the behaviour crops up again, punishment may be required then, but again at that given time, it should be effective enough within a few uses, to have an effect without becoming simply an unnecessary aversive.