Tuesday, December 7, 2010

Ian Dunbar Seminar: Punishment, the Dog-Con System, and the Concept Formerly Known as Instructive Reprimands

Punishment is a hot button issue in dog training, and like every other trainer, Ian Dunbar has ideas about how punishment should (or shouldn’t) be used. However, he has unique ideas regarding behavior and misbehavior, ones I’d never heard before, so I can’t wait to share them with you.

But first, a definition: “punishment” is any consequence that reduces behavior. Therefore, a consequence cannot be considered punishment unless the trainer actually uses fewer as training goes on. This means that, for the average pet owner, most “punishments” are simply consequences which are painful, fear-inducing, or annoying… and completely ineffective! (I imagine they’re also confusing because the dog doesn’t understand why those unpleasant things are happening.)

The use of punishment in training is a difficult skill to master. To be done well, punishment must be immediate so the dog needs to know why he’s being punished. It must be consistent and happen every time the misbehavior does. It must fit the crime- serious enough that the dog wants to avoid it, but not so intense that it’s abusive. There should be a warning first, so the dog can avoid the punishment by changing his behavior. The dog must know what is expected of him, and if he fails, Ian says that the punishment the dog receives should instruct him on what to do instead.

This last point is interesting, and Ian contends that it’s often overlooked. I think he’s right. After all, whether it’s a collar correction for leash pulling or a time-out for playing too rough, most punishments only tell the dog which behavior was wrong, but do nothing to help the dog figure out what the correct response was. They don’t instruct the dog, and Ian believes that punishment can and should be instructive. He accomplishes this by using his voice, which allows him to use his language, feelings, and creativity in training to help the dog understand both how important it is to respond as well as providing consequences if the dog doesn’t.

He starts by letting the dog know if and when he ought to respond to a command through his “Dog-con” system, which basically formalizes when the dog may disobey. He created this system since he believes that it is rare that a person needs absolute reliability from their dogs.

The Dog-con system has three stages, each using a different version of the dog’s name prior to issuing a command. At Dog-con 1, Ian uses a dog’s nickname or a term of endearment to indicate that the command is more of a suggestion: “Hey, Huey, you wanna sit? No? Eh, whatever.” At Dog-con 2, Ian uses the dog’s real name to indicate that he expects the dog to respond. “Hubert, sit.” Dog-con 3 uses the dog’s full, formal name, and indicates that it’s show time; the dog ought to pull out his flashiest performance. “Hubert Lewis, sit!”

This concept absolutely hurt my brain when I first heard it, because when I give a cue, it’s generally because I want Maisy to do it. It took me awhile to figure out that there are times when I don’t really mean it. As it turns out, I use different cue words instead of prefacing the same word with different versions of her name. For example, whereas “come!” means “get your butt over here right now,” the variant “come on” means “when you’re ready, please head this way.” I didn’t set out to teach it this way- if I had, I would have chosen more distinct phrases- but Maisy and I both seem to understand the difference. So, I guess Ian and I are doing similar things, just in different ways.

Ian thinks 100% reliability on the first cue is impossible. Whether your dog is distracted or simply has better things to do, there isn’t a dog alive who will be perfect, even at Dog-con 2 or 3. However, instead of applying a punishment that causes pain or fear- things that are unnecessary in training- Ian uses his voice to get the dog to do what he wants. He does this using a method called repetitive reinstruction as negative reinforcement (RRNR). (Scientifically-minded readers will undoubtedly notice that the method is technically not punishment, a fact which Ian both acknowledged and dismissed by saying we shouldn’t get hung up on terminology. He said that the distinction is too fine to really matter.)

RRNR is actually pretty simple to implement. If the dog fails to respond, Ian continuously repeats the command until the dog complies. If the dog is at a distance, he will move closer to the dog as he repeats himself. If the dog was distracted, this helps bring the command to the dog’s consciousness, and if the dog simply chose not to respond, this lets him know that Ian was serious. Either way, once the dog has finally performed the cue, Ian will repeat the exercise until the dog responds on the first request, and then reward the dog.

Although many trainers worry that repeated cues will result in learned irrelevance, Ian doesn’t. He argues that if the dog fails to respond, the cue is already irrelevant in that moment, and you can’t make something more irrelevant. Besides, Ian makes sure that the dog does respond eventually, which means the dog will learn that the cue must be followed.

While RRNR helps instruct the dog when he fails to comply, it doesn’t work when the dog is trying but gets it wrong. When this happens, Ian will use a “specific redirection,” which tells the dog what he ought to be doing. For example, if the dog is jumping up, you tell him to sit. If the dog is lagging or forging at heel, you can tell him to “hurry” or “slow down.”

It sounds like this works well for him, but I (and others at the seminar) are concerned about doing this since cues can act as a reinforcer, which would mean you’re actually rewarding the very behavior you’re trying to punish. When someone asked him about this, Ian seemed confused, so clearly he hasn’t had this happen. Either we misunderstood how and when he uses specific redirections, or the cues Ian uses don’t have a strong enough reinforcement history for them to act as reinforcers.

At any rate, while I agreed with Ian’s feelings on punishment- it doesn’t need to hurt or scare a dog- and I felt confused by the rest. But maybe that’s just me. What do you guys think? Have you tried RRNR or specific redirection? What’s your experience been? Do you do something else entirely? I’ll be curious to hear what you think!


Ninso said...

As I mentioned in a previous post, what I know about his methods is what I've read from blogs, but I find a lot of confusing. It also doesn't seem to me like it's anything new.

My dogs don't have 3 names--Lok is Lok, that's about it. Elo has about 12 different names, but only one formal and I doubt he knows the difference. Jun is deaf and doesn't know any name. I do use "more casual" commands like you do, but I certainly don't have a "system." I think my dogs are pretty good at figuring out when I mean it and don't mean it.

Maybe it's just me because I have a deaf dog, but when somebody says "it should be done this way" and that way can't be adapted to fit all dogs, I have a problem with it. The thing I like about more "conventional" (or maybe I should say scientific?) training methods (e.g., the concepts of reinforcement and punishment in general) are that they can be adapted to fit the individual dog, and if you understand them, you can train any dog. This doesn't seem to be the case with Ian's methods. Pretty much all the methods in this post would be completely useless with a deaf dog, since they seem to rely on wording and voice changes.

Also, the concept of "specific redirection" with the example of "heel" seems really strange to me. If the dog knows what "heel" means, I don't understand why it's necessary to have additional cues to fix lagging or forging. If the dog is lagging or forging because it doesn't understand heeling, why not just work on refining the meaning of heel? If the dog knows heel and is lagging or forging due to not paying attention or being over-enthusiastic, why wouldn't Dunbar then use RRNR--repeat the command and enforce until the dog does it properly? If "Slow" and "hurry" work right away, the dog is probably responding to the voice (think PMcC's work on universal voice patterns for certain cues) and you could use any words. But if it takes time, that means the dog is having to learn additional cues for the same behavior (seems messy to me), AND that you've tainted your heel cue--it no longer means "heel" it means, walk somewhere near me and lagging and forging are ok unless I tell you otherwise. If that's what you want, then I guess you're good, but I can't imagine why anybody would bother to teach a heel in that case.

Like, I mentioned before, the dogs, God bless them, somehow seem to figure out what we want, no matter how we convey it to them. But Ian's ideas just seem to overly complicate things.

This is all super interesting and I could go on, but I'm guessing I'm just echoing thoughts you've already had. Thanks for posting these synopses, very interesting to read!

Crystal said...

Wow, Ninso. I never considered that a lot of this wouldn't work with deaf dogs. I feel kind of silly for overlooking such a huge and obvious issue. I wish I'd thought of it- I would have asked him how he alters his approach for working with deaf dogs, especially since it is so voice-dependent. I suppose you could repeat the hand signal repetitively, but it's easier for a dog to ignore visual stimuli (by looking away) than it is to ignore auditory stimuli.

I have no issue with his Dog-con system, but I have so many nicknames for my dog that I'd NEVER remember when I'm supposed to use which name. I do like that he's thought about when he he "means it" and when he doesn't. I hadn't really thought about it, and it was good for me to consider how I let my dog know that a command is actually just a suggestion.

As for the specific redirections... Ian is unapologetically a pet dog trainer, so I doubt he cares if the dog forges or lags a bit at heel. In fact, I got the feeling that there are very few times where he wants absolute reliability. It sounded like the only time he needs that perfection is for an emergency recall and/or sit at a distance for safety reasons. I'd imagine he uses specific redirections for the rest. On the whole, a competition person will want MORE reliability from their dog than he does.

I do want to make it clear that I have a lot of respect for Ian. I don't necessarily understand how/why he's doing what he's doing, nor how/why it works, but I believe him when he says it does. That said, I haven't drastically altered my training approach. Assuming I understood him correctly, some of what he recommended just doesn't work with my dog.

Dawn said...

I guess you could say I probably do use a variant of this. I am constantly telling Magic to "get up here!" during heeling. Its not working.

My dogs have nicknames "Dammit(insert name here)" being the most common, but I guess I cant imagine using that in the show ring! I will say that this name is the one they respond most quickly too, so would that be my con3?

Laura, Lance, and Vito said...

I have not really tried RRNR. I did it once or twice and it just annoyed the crap out of me. I didn't like sounding like a broken record player. And since I have good success with my normal method I felt no need to change. I can see how he has good success with teaching ave pet owners this method though as they are already going to nag their dogs, so here is a way to nag effectively.

And you know that I was not happy with his answer to my question on cues as reinforcers. I think that specific redirection doesn't make since at all for competition trainers and I don't see a reason to teach average pet owners that either. Dogs learn behavior chains very quickly. So I think even if the cues don't have a ton of rewards behind them to make them a reinforcer in an of itself, giving a cue while the dog is doing the "naughty" behavior can easily lead to a way to get attention. The old I bark first, then I sit. Or jump up first, or paw at a leg, or any other behavior to get the owners attention and feedback.

Raegan said...

The more I think on RRNR, the more I think I like it. To be used as intended - effective and humane correction for pet dogs.

I firmly believe that pet owners should not be pitched the idea that they need to be 100% R+ all the time. It's too much pressure for someone who wants a trained dog as opposed to someone who wants to train their dog. Pet owners NEED to be able to say "No" without feeling guilty because they need a way to stop behavior at that moment. They are unlikely to be interested in developing the mechanical skills (someone who is interested in developing the mechanical skills will almost always fall under "wants to train the dog" ie process vs. outcome orientated). Pet classes NEED to be teaching people how to apply effective and humane punishment because they're going to anyway. It's just a fact of priorities.

Ninso said...

Don't feel silly! I wouldn't have thought of it if I didn't have a deaf dog. A lot gets modified, but I do like to see methods that are at least capable of modification. Just the pressure of forward movement would probably work for the RRNR. But I'm not sure if you'd be able to adapt the rest.

And the average pet owner probably isn't going to have a deaf dog anyway.

On the point about Ian's stuff working but not with your dog . . . Of course all dogs are individuals and learn differently, but I also think our dogs learn to learn the way we teach them. So if we're all of a sudden drastically modifying our method of communication its probably super confusing for them. I bet this stuff works better on a dog trained that way from the start.

trillium said...

I'm late to the party, but I'm going to toss it in anyway. (With a slight side trip but I'll eventually get to my point.) As you may remember I'm a 911 dispatcher, there are many times I use RRNR to get compliance when the other person is too drunk or too upset to actually answer my question. It's *amazingly* effective. I would estimate 1 in 50 recognize what I'm doing, before giving me what I want. And even those that say Hey you've asked me that before, I can say well yes but you didn't answer it yet. Others eventually answer my question without ever recognizing that they have been manipulated.

In terms of the girls, I rarely have to use the same methods. (But I firmly believe that most dogs are significantly smarter than a drunk guy calling 911.) Even if its not as clearly defined as Ian does it there is (perhaps just subconsciously) a difference in the way I pronounce their names which they can interpret in the same way as using their pet/real/full names.

Actually I think what this conversation has made me think, is that I just talk to freaking much. It's a wonder they ever figure out what I want. I think next weekend I'm going to try a session where I don't say anything just use hand signals and body placement and see what happens. If its not raining. We can't get enough momentum in the house to try something like leaving out words.

Crystal said...

DAWN- If it's not working, it's time to try something else, don't you think? :) I had to laugh at your dog-con 3 name of "dammit." That's hilarious. On a more serious note, though, we should get together sometime for an informal training session. Too bad it's winter and cold, but maybe we could meet up at a pet store? It would be good for me and Maisy, and fun, too!

LAURA- I like your comment that RRNR annoyed the crap out of me. I do think any given method needs to work for both the dog and the trainer.

You and RAEGAN make a good point- there needs to be an effective way for the average pet owner to "punish" their dogs. I love R+ training, but it's impossible to use all the time. Instead of getting hung up on "you must be positive!" I think it would be far better if trainers started talking about dog-friendly ways of stopping unfriendly behaviors.

NINSO- in addition to the fact that my dog has learned to learn a certain way, I have learned to handle her a certain way. When something is new and uncomfortable for me to teach, Maisy's going to pick up on that. My teaching won't be as effective.

TRILLIUM- I'd love to hear about your experimental training session. I think I want to get more stuff on verbal only, and use less body language, so I think it's interesting that you're going the other way. :)

trillium said...

Well since I can't sleep I decided on an impromptu training session. I used only behaviours that we already know. (sit, heel, down, stay (sit/downs), stand tall, touch, leave it, take it, and over.) Figuring 10 would be enough for variety and small enough I could track it easily. So I made a chart, and planned it out in 4 phases. In phase 1 we went through all of the behaviours using both voice and hand to make sure that peanut understood exactly what I was expecting in the way she was accustomed to the request coming to her. Behaviours were performed at 100%. (which I expected, they were things she should know, we were in the living room with no other distractions.) Phase 2 I gave the hand signal waited a count of 3 and if no response gave the verbal command. I got 40% response without the verbal queue, and 95% after I gave the verbal queue. Phase 3 I gave the physical cue, waited to a count of 7 and gave the verbal queue. I got 20% response from physical and 55% from the verbal.This indicated to me that she sees the physical and verbal as one queue, and needed both parts to know what to do consistently. The longer the delay between part 1 (physical) and part 2 (verbal) the less likely I was to get the response that I wanted. I didn't do phase 4 which was going to be physical cue only, because a) im getting tired, and b)I wanted to end on something positive, so I went on to some new behaviours we have been working on (through a tunnel, and walking on a elevated board). I was going to work on a training plan but I think I'm going to sleep on it tonight. I'm tempted to start a brand new behaviour that I only teach with hand signals so that its not at all influenced by voice, and reserve my voice for encouragement, marking, and things that are really really important (come here now! Leave it (that skunk). And phase out my voice for the things that we already know by slowly increasing the length of time between the physical and verbal cue. What do you think of that plan.

Crystal said...

Wow. Interesting! I get the opposite- Maisy very clearly understands physical signals, but is inconsistent with verbal-only cues. How did you initially train the behaviors? We did a lot of lure-reward for her early behaviors, which naturally creates hand signals.

I would definitely encourage you to strengthen your physical cues. Not necessarily because I think your voice needs to be for marking, encouragement, etc., but mostly because I have a friend whose older dogs have become deaf/mostly deaf as they've aged. Having hand signals in place has made things much easier for them.

trillium said...

I initially taught the commands verbal/physical simultaneously. I'm a chatterbox by nature and talk to them all the time, it's most natural for me to do them at the same time. I think she has paired the word/gesture with the right action, but not word/action gesture/action. It will give us something to work on while its too muddy to play outside though.

Wei-Han Lai said...

I like Dunbar's classification and explanation of behavior mod techniques. I think he does a great job of going into detail the intricacies beyond basic operant conditioning.

Where I disagree with him and most trainers and common owners, is that the ultimate goal seems to be to try to communicate with the dog through words. I question the importance of verbal cues in every day life. It may be important for obedience competitions, I don't know, but I rarely ever find it necessary to use verbal cues as a normal pet owner. I don't support Cesar Millan's methods, but he is an example of someone who does not communicate with dogs with words, but with more basic and understandable communication. Not all dogs are border collies, so relying on words for communication seems like an unreliable method. In short, it's like Dunbar is basing his methods on things that haven't been very well researched and documented - how well does a dog actually understand words? how well do dogs understand tone? etc.

Las Vegas Dog Trainers said...

This text is worth everyone’s attention. How can I find out more? Really great news!!! this information is well worth looking everyone. Good tips.
Las Vegas Dog Training