Evidence for the Use of Reward Based Training

    • Gold Top Dog

    Evidence for the Use of Reward Based Training

     http://www.antrozoologisenteret.no/artikler/art_training_methods.pdf

    And, for those of you who wonder why I am so vehement in my defense of reward based training, I think Donna said it better here:

    http://www.thepryorfoundation.org/dufford.html

    There's always this assumption on the part of so called "balanced" trainers, that we positive trainers haven't a clue about how they train, but the truth is that we used to be them, and have now moved on and cannot come back over that particular bridge.  All we can do is beckon them from the other side to trust the bridge and come on over.

    Smile 

     

    • Gold Top Dog

    I wish in the Dufford article, there was actual data, and not just an anecdote. What was the number of trials to criterion for each dog? What were the "stress" behaviors? What was the rate of stress behaviors in each dog? I think I've read this article before, and still, without the data, we really know nothing. It's no more than any other anecdote any of us could tell. I don't disagree with it, I just wish it were more useful.

    • Gold Top Dog
    spiritdogs

    I find at least 4 different "mistakes" in this study:

    1.- The study was based on questionnaires, these were given to most likely to dog owners inexperienced on how successfully train a dog to do or (more important) stop doing something i.e. tapping a dog's nose, smacking, shouting, etc. Those are not effective techniques in any philosophy

    2.-Sit, come, leave it or heel are techniques that can not be taught other than with reward-based only exercises, giving this way a clear advantage to them.

    3.-"Many owners used punishment based training to teach dogs not to steal/chew objects. However this survey suggests that for everyday training, punishment is not the most effective method. Furthermore, for certain tasks reward-based methods are significantly more successful".

    This is another example of how unfair this study is. Dogs do not chew/steal objects every day vs teaching your dog how to sit and come. Once again using techniques like tapping the dog's nose as a "viable" punishment for this study.

    4.-"Although the survey found no support for the use of punishment-based methods, we are not suggesting that they can not be used effectively. Historically, punishment has been used successfully for many types of dog training. However, within the pet-owning community, training is very often performed by inexperienced people who are unfamiliar with the behavioral principles, hence the timing of the delivery of punishment may be inappropriate and it's use inconsistent" (emphasis added)

    Another example that back up the fact that this study is not accurate

     

     

    • Gold Top Dog

    And I will counter-point some of your points, Espencer.

    !) The whole idea of the survey was to understand how the common person is training their dogs and and what the study noted was that for dogs commonly trained with only punishment or some punishment, there were problematic behaviors. In both humans and dogs. It was never defined as a study of just certified dog trainers.

    2) The report did not say that sit, recall, or leave it could not be taught by punishment but that they rarely taught by punishment these behaviors. It could also mean that those who train mainly by punishment, aversive, or -R may train those specific behaviors with rewards, as a general. That means that, among the sample population, there are some predominately +P trainers that use rewards when nothing better will do.

     

    3) "
    This is another example of how unfair this study is. Dogs do not chew/steal objects every day vs teaching your dog how to sit and come. Once again using techniques like tapping the dog's nose as a "viable" punishment for this study."

    No, Espencer, your dog doesn't chew/steal objects every day because you are a caring and concerned owner who has been involved with him the day he came into your house and you have never allowed such behaviors to reinforce. You are confusing yourself with the average owner. And let's use nose-tapping as a generic phrase for any hands-on aversive, regardless of strength of applied force. One of our members that I haven't seen here in a while had a Dog de Argentino and she would poke fingers in his neck that he never thought of, in my opinion, as a punishment, but as a cue to break off whatever he was doing. So, people do nose-tap, finger-poke, rump-swatt, place themselves as physical barrier all the time. What the study is trying to say is that those are not always effective and sometimes cause other problems. As opposed to rewards-based training, which reinforces desired behavior more often, and also results in a better relationship, though that could be a combination of the dog having less fear and the human having a different paradigm for human-dog relationships. It is more common for dogs to live inside the house now than in the past. It is more common to feed dogs high-quality foods with better balanced diets than their owners eat, than in the past. So, I, with you, perhaps, would like to see a delineation between those two factors. How much is due to the dog being less afraid of the owner and how much is due to the owners' weltanschauung (world view) about dogs, in general.

    4) "Although the survey found no support for the use of punishment-based methods, we are not suggesting that they can not be used effectively. Historically, punishment has been used successfully for many types of dog training. However, within the pet-owning community, training is very often performed by inexperienced people who are unfamiliar with the behavioral principles, hence the timing of the delivery of punishment may be inappropriate and it's use inconsistent" (emphasis added)"

    I agree that the improper use of punishment is creating a monkey wrench in the works of trying to establish accurately whether or not punishment is effective, or not. And methinks thou dost protest too much. Some have said that CM is successful because he applies +P correctly. You will find, as you know, a great number of people, professional and amateur, who disagree that he does. Environmentally, without human intervention, +P doesn't work every time.

    I knew of a lady who let her dog run loose all the time. The dog had been hit by a car 3 times. The third time was hard enough to create a permanent physical disability. Then, she finally started walking the dog on a leash because even that third time was not enough to deter the dog from wandering whereever he wanted to go. Other dogs may continue to exhibit a behavior regardless of corrections from their canine fellows because the behavior they are doing is either ideopathic or it is rewarding in some way and they will only change when doing something else instead of that behavior is more rewarding.

    Dogs seek equilibrium and security, which includes security of resources. They are Nature's ulitmate capitalists, moreso than man, it seems. Given a chance, they avoid confrontation, establish their own territory, doing whatever feels good or helps them surviving, all of which are reinforcing factors.

    Again, punishment does not train, it only stops something. Reinforcement trains. That is, you don't train out something, you train something in its place. So, perhaps the problem with the study, if there is one, is that punishment is not really a training method, even if it is a quadrant of OC.

    Which does make rewards training superior, by definition. Now, I know we'll get into "whole toolbox" analogy. I'm one of the few here that actually carries a set of tools in my profession. I have a couple of hundred pounds of tools. I carry only what I need and what is effective. 9 times out of 10, I don't need the 15 lb sledge hammer, so I don't carry it. I carry one bag with the precise tools I need most often. And if something needs a soft tap, I use the lineman pliers (which we also call the electrician's hammer.) And that's usually to tap a flat blade screwdriver into a concentric knock-out in a box or enclosure. So, I don't have to brag that I have the complete toolbox, including hammer.

    So, if I use rewards most of the time, then the once or twice I use punishment, it will be just enough hammer at the time to accomplish a specific task, and then set aside when a pair of needlenose pliers is perfect for wiggling the knock-out out of the box. And it will be precisely used and only for that moment. If it's not effective, then the use is no longer required or suitable.

    Timing is everything. Hence the use of a marker. It is easier to mark the good thing with a sound and then reinforce, linked to that sound, than to be able to punish for wrong behavior the instant it happens. The problem with punishment is not just whether a particular human is competent in its use, it's a function of human limitation in reaction time, proximity, and ability to match a dog's changing state of mind.

    So, point 4 does point to a weakness in the study but it also points to a weakness in any human's ability to apply correction to a dog as another dog would do. I also question, aside from this study (and perhaps this would lead to another study to define the parameters) how often a dog responds well to another dog's correction. Case in point, neighbor dogs, Gracie and Harley. Sometimes, Harley noses in on the barkfest between Shadow and Gracie. Gracie will sometimes nip Harley to keep him back. But this doesn't stop Harley from trying every once in a while. So much for positive punishment between dogs. Or is Gracie misapplying +P and what dvd's can she watch to get "better" at it?

    • Gold Top Dog

    espencer
    spiritdogs

    I find at least 4 different "mistakes" in this study:

    1.- The study was based on questionnaires, these were given to most likely to dog owners inexperienced on how successfully train a dog to do or (more important) stop doing something i.e. tapping a dog's nose, smacking, shouting, etc. Those are not effective techniques in any philosophy

    2.-Sit, come, leave it or heel are techniques that can not be taught other than with reward-based only exercises, giving this way a clear advantage to them.

    3.-"Many owners used punishment based training to teach dogs not to steal/chew objects. However this survey suggests that for everyday training, punishment is not the most effective method. Furthermore, for certain tasks reward-based methods are significantly more successful".

    This is another example of how unfair this study is. Dogs do not chew/steal objects every day vs teaching your dog how to sit and come. Once again using techniques like tapping the dog's nose as a "viable" punishment for this study.

    4.-"Although the survey found no support for the use of punishment-based methods, we are not suggesting that they can not be used effectively. Historically, punishment has been used successfully for many types of dog training. However, within the pet-owning community, training is very often performed by inexperienced people who are unfamiliar with the behavioral principles, hence the timing of the delivery of punishment may be inappropriate and it's use inconsistent" (emphasis added)

    Another example that back up the fact that this study is not accurate

     

     

     

    On point #1, I don't see anything where we are told about the prior training experience of those interviewed. Perhaps experience correlates with something here, but we don't know. We don't know if this is their first dog or tenth dog. It's a flawed study, but the fact that it was based on questionnaires is the least of the issues in that regard. 

    #2, plenty of people successfully trained dogs to perform the behaviors listed using negative reinforcement in the past. By reward based I will assume you mean positive reinforcement (and this is why we should use the correct language, because it helps avoid this confusion). If I am wrong, let me know. 

    #3 We can't be sure it was really punishment, because we don't know for sure that the future likelihood of the behaviors was decreased. Just because the behavior stopped for the time being doesn't mean it was a punishment. Yet another flaw of this study. 

    #4 Despite all its flaws, I think this is the best thing we can take from this study. The average person doesn't implement punishment properly. You're much less likely to create problems with poorly timed and inconsistent reinforcers than you are with poorly timed and implemented reinforcers.