The Great Dr. Ian Dunbar, part 1
I had the great opportunity to attend Dr. Ian Dunbar's seminar the past two days along with my coworkers! He is a very good speaker and had many great points. I am sure my friend Crystal will post a very thorough analysis of it later (hint!) but I just want to record and talk through some of my impressions on a few points I had strong views on or disagreed with.
One of the big points of the weekend was that feedback needs to be binary. The dog really needs to know when it is doing something right AND when it is doing something wrong. He doesn't believe that aversives need to be used, but he does want constant feedback given to the dog. If just given negative feed back dogs and people are clearly stressed and can shut down. If just given positive feedback, Ian says that just shaping can cause a dog to be frustrated and get easily stuck in a rut of offering the same behaviors over and over again. But playing hot and cold, with calm negative feedback and not yelling can result in very fast learning.
My argument with this is not the logic that giving both neg and pos feedback can be effective. I do think that a lot of dogs can do quite well with it and definitely believe that punishment doe not have to be aversive. However, I think that when a dog is first learning something that shaping a behavior is best done without any no-reward marker. I know that Lance especially can very easily shut down on me and that he would not want to be told when guessing wrong. Some dogs might not care, but I think that free shaping without the negative feedback can really promote creativity and without inhibiting a dog's behavior. And with a dog experienced with shaping and with a great trainer, there is no reason a dog needs to be frustrated and getting stuck in offering wrong behaviors. A good trainer would have a high rate of positive feedback (click lots) and even if the dog tends to stall out there is always something little one could reward (head turns) or treat placements to get the dog back on the right track. So personally when teaching a new behavior to my dogs I am going to stick with the positive only feedback and see no real point in his argument against with a skilled trainer.
Once a dog has learned a behavior I see nothing wrong in using punishment. Ian tends to use mainly positive* punishment* (correction, he uses negative reinforcement) while I still prefer negative punishment. I agree with him that time outs really aren't all that effective and tend to be used for too long, but I think turning your back on a dog, setting dinner back down on the table, or walking away from the door can all be very effective. But I also use verbal reprimands with commands I am sure are well taught at the current level of difficulty.
One of my bigger beefs was with Ian's method of negative reinforcement. He terms his technique Repetitive Reinstruction in Negative Reinforcement (RRNR), at least I think I got that right. Basically if dog doesn't adhere to a command, let's say a sit, Ian repeats the cue over and over again while walking closer to the dog until the dog finally does it. It is negative reinforcement because the annoyance ends once the dog does the behavior, but relentlessly continues as long as long as the dog doesn't perform. He previously termed it Instructive Reprimands as it also tells the dog what he wants them to do. While it may seem like nagging, Ian states that it is not. Unlike nagging, he will persist until the dog actually does the behavior, and then he will insist the dog to do it a 2nd time for any reward. Clearly not nagging.
I cannot ever see myself doing that with my dogs, let alone recommend it to average Joe whose dog is already used to tuning out his drones. I have seen lots and lots of "sit sit sit sit sit SIT SIT"s in my lifetime and it bugs the crap out of me. I don't think that repeating a command is evil, but I don't see any benefit at all to repeating it a thousand times until your dog does it. If my personal dogs don't sit the first time, I usually ask again, possibly in a more strict tone. If the dog does it, I tell him thank you but don't reward it since it wasn't on the first cue. If the dog doesn't do it, I silently walk in to help the dog (if at a distance) and then use gentle guidance with either body pressure, maybe collar pressure, or usually just a look. Once I get compliance I would then ask for it in the exact same situation (or maybe a tinier bit easier). If the original blow off was a behavior at a distance, I go back to that distance, or if the original behavior was a stand to a down, well then I make sure that position change is what I practice, not a down from a sit. I was disappointed that in all the examples, Ian would practice the behavior again so he could reward the dog for complying, but it was always right up close and the initial misbehavior was at a distance. (*note, in my personal choices of what I do, I'm talking about pet dog behaviors. In training for dog sports, my dogs DO always have a choice on whether to comply or not. My goal is to make training so fun that the dog doesn't even consider not playing with me. I also don't do the same corrections when the dog makes a "mistake" in dog sport training" that I personally use in everyday life.*)
On the other hand, with a behavior the dog does not know very well or in a situation they haven't been it, I wouldn't even bother with those corrections but would instead give him the benefit of the doubt and immediately help him out. Or in a training session where the dog is paying attention but simply not responding, I interrupt the opportunity by telling my dog "too slow," move a few steps before he completes the behavior if it was started, and ask for it again.
I have a few more points of his points I want to write out and think over and will likely post a bit more tomorrow. And really I did enjoy the 2 days of seminar I attended, I think it is great when a presenter really makes you think, whether you agree with them or not! The 3rd day of his seminar is tomorrow but I will be off at agility :)
Anyone have any comments on the above? What do you see as the appropriate usage of punishment with a known behavior? Do you agree with Ian in that a no reward marker, or even a a bit of punishment (non aversive) is a good tool in the learning stage?
Click here for part 2
"If my personal dogs don't sit the first time, I usually ask again, possibly in a more strict tone. If the dog does it, I tell him thank you but don't reward it since it wasn't on the first cue."
That's a pretty good description of how I do it, too. Not for things I'm teaching her for the first time, but for situations that she knows. A sit, for example, she knows how to do.
When I went to the Terri Arnold seminar in July, she used a similar theory. First, she said that you give the command. If the dog doesn't listen the first time, you give a more stern command in kind of a growl-y/serious voice. If they don't respond the third time, you correct them.
Her theory was that when dogs correct each other, they don't just attack savagely for no reason; there's a build up of warnings which humans should follow too. Otherwise the dogs will think you're just a psycho and shouldn't be listened to at all.
I personally don't believe in repeating commands. I make the mistake sometimes, but I try to make an effort to only give one command. I've seen firsthand how easily the dogs (Layla) ignore someone who just blathers on (Pat) because it means nothing to them. Similarly, when I run agility with Layla, I try to keep my talking to a minimum. I won't say "jump" at every jump. I never give commands to jump unless it's a discrimination. But I try to be quiet so that when I do say something, she'll be (hopefully) more inclined to listen.
As far as your last question -
I use a mixture of both positive and negative reinforcement as well as positive and negative punishment. I don't really understand what a No Reward Marker is (I've read about it before but can't remember) but I do use verbal markers when something is done incorrectly. If Layla pops out of a weave poles, I mark it verbally at that pole, not when she completes them; if she breaks early on the recall, I mark it when she breaks (or as close to it as I can), I don't wait until she arrives.
I haven't had a lot of experience in shaping, but I do agree that a no-reward marker (or any negativity) shouldn't be used. Since the dog isn't really sure what you want, they're not intentionally doing something wrong. I was told that in shaping, basically just see what the dog offers up and it'll make them more confident their ability to try new things and also make them confident that if they do something wrong, it isn't the end of the world.
Good luck at agility!!
I will indeed have much to say. Perhaps too much, but honestly, I learn best by writing about things.
I just wanted to comment on this: "I was extremely annoyed in that all the examples, Ian would practice the behavior again so he could reward the dog for complying, but it was always right up close and the initial misbehavior was at a distance."
He talked a bit more about this on Sunday. Basically, he said that the first couple of times, you'll need to get all the way up to the dog, but after that, they'll begin to understand that you mean business. As a result, they'll respond sooner- when you're further away.
I'd love to see your other thoughts on the seminar.
I would NEVER use the RRNR. To me, if there is a punishment there, it is in the stern tone and pressure of moving closer to the dog. I don't think "sit sit sit sit" means a THING to the dog at that point. You might as well be saying "no no no no no" as you approach, or even saying nothing at all. Plus, as you pointed out, the dog learns to comply up close after repeated commands, but that does nothing to help the dog learn the sit at a distance.
Different things work for different dogs, but I think for most dogs other methods would be more effective. I don't think Dunbar's method would be BAD necessarily (Though it could be, depending on the dog. I don't think Lok would respond well. Elo might shut down from the pressure also.), just not as effective as other things.
As for punishing known behaviors, I try to stay away from it. In my experience, there is usually a reason for the misbehavior, e.g., not properly trained to respond with that level of distraction. The one I will "punish" for is a blown off recall, and "walking down" is my punishment of choice. Usually when I use punishment though, it is to deter behavior that I don't want, rather than correcting wrong responses to my cues.
I DO occasionally use the "negative feedback" in my shaping. Sometimes if the dog gets stuck offering something I don't want, an "oops" will correct it, or a quick turn away. But I don't use this much, and the initial problem is probably my fault to begin with.
You're much better than me, Laura - ha! - I attended Ian's seminars back in Feb and still haven't got round to doing a post about them - and now probably never will! He covered the same things with us and I had similar thoughts to you regarding what you said.
Like you, I don't believe in using correction or punishment when teaching - I think it's unfair on the dog as you don't know if he is refusing because he is confused or because he is being "disobedient" - therefore I always give the benefit of the doubt and lower the criteria, etc.
But once a dog has learnt something, I am not afraid to use correction or punishment, which I know is controversial in today's PC dog training world. But I believe that if done fairly and consistently in a way the dog understands - and you build a very strong bond with the dog otherwise through other activities together - than an occasional bit of correction does not damage the bond between you. And as I always say - if it does, then I don't think you've got much of a bond in the first place.
In my experience, as long as an aversive is simply that (something unpleasant to avoid) and not emotional bullying (which is what correction often ends up being) - then it doesn't necessarily have to traumatise the dog. It's no different to them learning not to go too near a fire because it might burn or learning not to harass the family cat because they might get scratched - those are all unpleasant but non-emotional consequences of that course of action and they simply learn to avoid doing those actions. I don't think there's anything wrong with teaching dogs that certain actions have unpleasant consequences - that's what they would learn in the wild too.
To me, having an aversive makes it easier for the dog to make the right choice - if I do this, something unpleasant happens - if I do that, I get a reward - obvious choice...I find this makes things a lot easier when you're faced with a situation where the inappropriate behaviour is self-rewarding (eg. lunging and barking) and you're trying to get the dog to choose the alternative only by the temptation of rewards. Well, sometimes the 'bad' behaviour is so rewarding that it trumps the alternative that comes with your rewards (OK, so if I lunge, I don't get treats & praise but so what? Lunging is much better fun!) and so its a hard choice for the dog. So I find coupling an aversive to the inappropriate action to interrupt the behaviour makes it a less tempting course of action and helps to encourage the dog to make the 'right' choice of the alternative appropriate action. Then you can reward & praise the hell out of the dog for choosing the alternative and gradually, the dog will always choose the alternative in more and more situations, to the point where you can dispense altogether with the corrections/aversives. Basically, they learn faster with the 2 types of feedback and this is what Ian believes too. This is how I trained Honey and now that she is older, I hardly ever have to use corrections for anything - I simply say "No" (which I know is still correction in some books) - but "No" in our case is simply a signal for "this action will result in unpleasant consequences" - and Honey understands that and doesn't find that stressful.
(...continued)
(...continued)
However, this only works if you follow the binary feedback system that Ian was talking about. I think the reason correction usually fails is that people only correct without also rewarding/praising when the dog stops doing the wrong thing or gets it right - so then the dog gets confused, stressed and shuts down. Naturally, if you only get told off for what you shouldn't do but never given any indication of what you SHOULD do, it would really stress you out. So praise coupled with verbal reprimands - that is the way I've raised Honey, especially where things like household rules are concerned - there is a lot of praise & verbal reprimand given together until she understands the boundaries - sort of like that game we used to play as kids "Hot - Hotter - no, Cold - Cold, Freezing - Warm - Hot!" etc.
The only place I don't use correction at all is for things like dog sports, eg. our dancing, coz I feel that isn't 'important' like everyday manners and should be completely voluntary therefore it is unfair for me to use correction to get a behaviour there. For me, doing a Spin is optional but doing a Recall (in the park) is not. I have a lot of tolerance for things like tricks but I have zero tolerance for things like aggressive behaviour. This may be exaggerated because I have such a large, powerful dog and just cannot afford to have any margin for error, especially as she weighs 20kg more than me and can easily overpower me.
I'm not sure I agree with the negative reinforcement technique Ian uses either - for one thing, it's not very practical in many situaitons! And it has the dangers of teaching the dog to ignore you, like you say. I prefer a very swift, sharp but unemotional positive punishment to interrupt the behaviour, followed immediately by praise for stopping the behaviour or doing the 'right' thing - the key with this is the timing though and again, most people really struggle with the timing for corrections - which is why it is such dangerous thing to promote to the average Joe 'coz it can do so much more damage if done wrong, compared to positive methods. I think correction is a very effective & powerful dog training technique if done right but it is so difficult to get it right, and you also have to know what is the appropriate type & level of correction for each dog's temperament...just too much for the average Joe.
Hsin-Yi
Thanks for the your well thought out response guys!
Tania- I agree with you that the majority/all of "misbehaviors" have a good reason from the dog's perspective at least! That is really why I a very careful of using anything more than a no-reward marker on trained behaviors. And like you I think a recall is one thing I will always walk down. But with other basic obedience stuff I think that once you get to a certain point in your training, I do expect my dogs to respond to sits, downs, stays and a loose leash regardless of the distraction. I probabally never get to that level with other behaviors like tricks or even their formal obedience stuff so I see it more as my failure than theirs.
Hsin Yi- I totally agree with you and Ian that punishment does not need to be aversvives. But I still can't see myself using it in the learning stage. I think just playing HOT is perfectly fine for the majority of behaviors out there. Preventing access to things is often enough as long as you teach them how to get that thing, and put it on cue. Even some self reinforcing behaviors like barking, can be put on cue and taught with quiet at the same time.
But, (thinking "out loud" here) I suppose I use pressure and the release of it to teach certain behaviors like stay which certainly is a type of punishment.
Blue, my GSD, says, "WTF!" and walks away if I use a no-reward marker. That tells me that my saying, "Ooops!" is aversive to her.
Hi there Laura, thanks for this very thorough review of the seminar. I don't always get to go and a lot of what Ian is presenting these days is so new that I've not heard the polished talk myself as of yet!
You make some very good points. I'd like to point out however, that Ian also recommend teaching a new behavior without any punishers, the format of using both negative and positive feedback (never scary or harmful) doesn't come into his program until you are proofing a dog for reliability, same as you.
You mention that Ian tends to use mainly positive punishment. I'm curious what he said that led you to that conclusion and to hear what kind of positive punishment he said he uses.
I see your point about repeating a failed distance exercise up close, though in practice it looks different, and the handler closes the distance to the point where the dog will reliably responds and then begins training anew from that point and beyond.
Again, thanks you for the thoughtful report. I look forward to hearing your response!
I got the impression that Ian doesn't introduce his RRNR until the proofing stage, but he did talk about using instructive feedback in the learning phase. Not meant as a punisher, but the use of a no-reward marker to speed up the learning process. He went though examples of teaching just through using "cold," using "hot and cold," and just using "hot" and explained why he thought giving positive and "negative" instructive feedback was best. Basically just shaping a behavior can lead dogs being stuck in offering something and can take longer to teach the behavior. I personally think that will work well with many dogs, but at least one of my personal dogs will shut down if he thinks he is guessing wrong. I prefer shaping without any inhibitions on the dog, and letting them completely think through the problem. But you are absolutely correct that he doesn't want punishment in the learning phase!
As for his use of positive punishment, you are right, I misstyped and will edit my post! I meant to say that Ian prefers negative reinforcement over negative punishment, while i still prefer negative punishment as a technique. He did say that he doesn't tend to use a lot of negative punishment and specifically hates timeouts.
Negative reinforcement is HIGHLY effective, but I just couldn't get over the technique he advised. But I suppose if the average pet owner is going to repeat commands over and and over again anyway, one could argue they might as well learn how to have it be effective :) As for my personal use, I think it would be too annoying!
And I know Ian did discuss a ton more on Sunday about using RRNR in action and how it ends up increasing the distance of response. I just wasn't able to go that day! It seems to me though that if the dog is able to do that behavior at the specific distance (not in the learning stage) then why wouldn't one reask for it in the original context (same distance, same position change)? Ian already does re-ask for the behavior (which I do like!) so why not take it one step further to really work on proofing? If the dog cannot do it at that distance, well then I wouldn't be adding punishment or negative reinforcement in anyway as he's not yet ready for proofing.
I really did enjoy his seminar a ton and wish I could have gone on sunday. I just tend to write about things I have questions over as it gets me to think about them more. Things I agree with aren't as useful to me to write about :)
Thanks again for such a thoughtful, clear post, I love hearing feedback and how his seminars are interpreted. And you explain things well and have a really nice blog here. :-)
I agree that not every dog would respond well to Ian's ideas as outlined, that's why it's so important to have many tricks up our sleeves as trainers. Cheers! ~ K
I use RRNR as the fourth stage of the process in off-leash play during Puppy Class and only after "Sit" is well proofed with adult dogs. No stern voice, no raised voice, no ugly tone or ugly face ... certainly not in any way aversive but ... amazingly and quickly, extremely effectively. Within just 20 trials the dog sits at a distance following a single request and so, improved quality of life for life.
Not nagging because the dog always has to sit, AND, if you have to repeat the request then the dog has to repeat the exercise until he sits following a single request. Trying to explain the routine is difficult, seeing it is easier to comprehend. Maybe check out a few videos on dogstardaily.com such as, http://www.dogstardaily.com/videos/distance-sits-action
The point is: it works! And it's easy for owners to do.
ultra woofs, ian
Thank you for your comment Ian! I do believe that RRNR is high effective, negative reinforcement always is if done right. And I am even more impressed that you have thought of a R- technique that truly is non aversive!
I just can't get over the annoying aspect of having to repeatedly cue the dog until you get there in the beginning stages. It may not be the same as nagging as nagging doesn't produce any results but your RRNR is effective, but it comes off with the same type of vibe.
I have thought about it in the week since your seminar and have come around to grant you that it does have a place in pet classes. I originally thought that clueless Joe already has a dog who is used to tuning out his actual nagging so he needs to learn to shut up all his droning and work on just rewarding attention and using premack. But people just aren't good at that. So I can see where teaching them to effectively "nag" (sorry!) their dogs, still using premack can be an easier change on them while getting good results.
Do you think though that if you're students were better trainers, would you still have the repeated exercise (to get it on the 1st cue) be at the original distance? Or does it just not matter at all in your experience? Or do you think that you even need to repeat the exercise at all (immediately) since you already negatively reinforced compliance and could go on to trial #2?
Anyways, thanks for a great seminar! You really are a great presenter and I love having to think about things!