Thursday, December 16, 2010

Ian Dunbar Seminar: Types of Reward-Based Training (and Which One is Best?)

Ian Dunbar is widely known as a reward-based trainer, and is often credited as revolutionizing dog training. But what people don’t talk about as often is which type of positive training he uses, and why. Today, I’d like to share with you the four types of reward-based training that Ian identified, and what he likes- and doesn’t- about each of them.

Plan A: Lure-Reward Training
Okay, let’s cut to the chase. Ian likes lure-reward training best, and I can understand why. Lure-reward training is generally very easy and fast to do, making it ideal for the average pet owner who just wants the basics, and wants them now.

Indeed, that’s exactly why Ian prefers it. Since dogs will predictably follow a lure, Ian says that you get a 100% response rate right from the beginning. (For what it’s worth, I think that’s an over-estimate. Some dogs are tragically difficult to lure into a down, but even so, you do get a high rate of response.) Because the dog’s success rate is so high, you can easily pair the behavior with the cue from the first repetition, making it the fastest way to get a behavior on cue.

Since the behavior is so predictable, Ian says that you can train with a differential reinforcement schedule from trial two in order to improve, not just maintain, behavior from the beginning. Ian also likes that lure-reward training comes with a built-in hand signal, and that you can work on several behaviors in the same session.

The downside to lure-reward training happens when people do it poorly. When people fail to phase out the lures in a timely matter, they become a bribe. Instead of using the treat to instruct the dog, people often become dependent upon using food to coerce the behavior from the dog. Ian also says that lure-reward training doesn’t always work with adolescent dogs, especially if the lure wasn’t faded out when he was a puppy.

Plan B: All-or-None Training
All-or-none training is Ian’s go-to when lures aren’t working because the dog sees the food as nothing more as bribe- and one less interesting than whatever is going on in the environment. All-or-none training seems to be Ian’s term for “capturing” a behavior. You wait for the dog to do the desired behavior, and then give a reward.

All-or-none training is easy. Either the dog is sitting, or he is not. It doesn’t take much sophistication, and as a result, is well-suited to basic behaviors and novice trainers. The down side is that since you aren’t giving any instructions (like with a lure), it’s hard to predict when the dog is going to do the behavior. In turn, this makes it much more difficult to get the behavior on cue.

Plan C: Clicker Training
Ian rarely uses clicker training, and he never uses it to teach the basics. Instead, he says clicker training is for anything you can’t get through luring or all-or-none training. Since he believes you can lure everything a pet dog needs, he doesn’t introduce it to his students until a level 3 or higher obedience class. Those dogs already have reliable behaviors, and he introduces the clicker to help refine the behavior, make it more precise, or make it flashier.

He doesn’t like clicker training because it is hard to attach a cue. He also believes that people click too often; because he thinks differential reinforcement is the best schedule to use, he believes that it slows down training if your dog is getting clicked more than 50% of the time. This number seems low to me- I’ve heard clicker trainers say that your dog should be getting it right 80% of the time before you increase your criteria.

Plan Never: Physical Prompting in Training
Physical prompting involves applying pressure with the trainer’s hands or by manipulating the collar in order to get a behavior. He includes tools such as shock collars and the Gentle Leader in this category.

Ian says that using physical prompting involves a lot of skill, more than most students have, and that in his experience, gentle prompting often turns into “physical splatting.” He also believes that physical prompting is a crutch which is incredible difficult to phase out.

Honestly, I agree. Although you can use physical guidance to help get behaviors, it is notoriously difficult to get rid of, so I generally avoid it. What I found interesting is that Ian doesn’t think props (like a physical channel to teach the dog to back up straight) are a crutch, and doesn’t find it hard to phase them out. I can’t comment on this- Maisy generally finds props scary, and it’s easier to find a way to teach a behavior without a prop as it is to desensitize her to the prop.

I have used all four of these methods. Most of Maisy’s foundation behaviors were taught with lures, because that’s what we were taught in our big-box store puppy classes. Once we began competition training, I started using a clicker to shape behaviors. I’ve done relatively little capturing and physical prompting, but I have used both. I don’t think any one of them is “best.” Instead, I choose my method based on the task at hand. I will admit that I tend to use shaping a lot because I think it’s the most fun. Still, Ian’s probably right that lure-reward training is easier and faster for the average pet owner.

What about you guys? Which method do you use most? Does it differ if you’re just starting out with a behavior? How do you choose which method to use? Do you think one is better than another? Let me know what you think!


Elephant said...

I'm inclined to agree with Mr. Dunbar regarding props. In my example -- training "wipe your feet" near the wall to avoid having Maisy *back up and then wipe her feet* -- the prop can be phased out simply by gradually increasing distance from the wall.

Now if only my training skills were up to the same standard! ;)

Crystal said...

Very good point. I was thinking of props in terms of platforms, targeting sticks, chutes for straight fronts or recalls, that kind of stuff... More trouble than they're generally worth. But when I have used things like walls, they haven't been hard to phase out at all.

Then again, I didn't have trouble phasing out physical prompts the one (and only) time I used them with Maisy: to teach shake. I just picked up her foot and gave her a treat a bunch of times. It didn't take her long to offer the foot on her own.

You are brilliant, husband-man. (That's right everyone, my husband is a brilliant dog trainer in his own right. You may all look upon me with envy now.)

Crysania said...

I've mostly used lure/reward with Dahlia. It was the only thing I really had a grasp on when I got her and it was what was taught in the basic obedience classes we took with her (at Petsmart). She responds well to it. Not having any clue what kind of training she might have had before us, if anything, it ultimately made most sense.

Since then I've been trying to work on shaping behaviors and the class we're taking for agility is all shaping based. Dahlia doesn't get it 100% and is often waiting for me to show her what to do. She's brilliant at getting things and FAST when I lure and reward, but she's slower at shaping still. Some of that may be me. I'm not good at coming up with little steps to get there and I'm horrible at rewarding at just the right moment to get those little motions toward the final result.

Crystal said...

It might be you, but I'm sure part of it is that she's used to being told/shown what to do with the lure. It's a pretty big leap from that to "hey just start doing some stuff."

Laura, Lance, and Vito said...

I taught my pet classes with luring. People can do it, they get frustrated with shaping. That being said, I want the first thing my puppy to learn is how to think. I do a ton of shaping right from the start with different objects so my puppies know that it is their job to get thing started, they control the training. I am not a free shape only person, I lure quite a bit as well in the initial stages, but I do switch pretty early on from a lure-shape combo to shaping only.

If I am shaping I am deinetly clicking a ton, I break down things small so the dogs aren't getting frustrated. When I'm working on perfecting an already known behavior, I still click more than 50% of the time. In his example of fast recalls I said 7 out of 10 (ok I actually wrote 5 since I thought that was the answer he wanted, but thought 7) should be rewarded. I can see how he would say 4-5 and how it could work. But I also would be afraid of the dog not wanting to try again or having an even slower response since they do't know what's being rewarded yet and might be afraid of getting it wrong. Of course it could be a completely unfounded fear of mine and maybe we really are better off only rewarding 40% of the time when tightening criteria.

Sam said...

Hmm.. this is an interesting question! I definitely used luring the most in the beginning for all of the simple behaviors. She hasn't had a problem with fading the lure and working just off of verbals, for most behaviors (but some tricks do still require the hand movement).

Now, I think it's definitely changed, and I use marking/clicking a lot more. But ultimately, it depends on the behavior - some things are more suited to luring (like teaching Marge to beg), and some more suited to shaping/clicking (like picking objects up with her mouth).

I think that shaping is better for the serious trainer who would rather not go through the trouble of fading the lure and/or hand signal when it's time to start competition training. Shaping also helps with the tiny little details of a behavior (like sitting straight in heel position), IMO.

Louise Kerr said...

Another good post Crystal. I mix together luring and clicker when teaching novice dogs and owners (many of who are cross over dogs/owners. I get the owner to lure the simple behaviour such as sit and then mark the bit of the behaviour they like with the clicker and reward. I also do a bit of shaping in about week 2 or 3 as we move onto more complex tasks. I think it is important to use whatever works so long as it is positive. I use more shaping along with a clicker to mark the behaviour I want in my own dogs who are more advanced. I don't think it is a one or the other arguement but pick what is going to work and do it.I have just read a blog by Nicole Wilde on dogstar and she says much the same thing- think outside the box.

Dawn said...

I have used all 4. I used a lot of physical prompting for conformation training. Happy legs with the pups hand moving the feet into position, all help teach the dogs to get there legs under them. I have used a small stick to encourage Magics butt to get over to learn a straighter sit. I dont think physical props are at all a bad thing. Like any form of training they have their uses.

Kristen said...

How we train a behavior depends on the handler and the dog...for skill level, experience, and dexterity.

We very very very rarely use prompts or props. That sort of thing often makes the owners think the dog 'knows it'....when in reality they don't. Not to say this is always the case.... but esp with new dog owners...

I've found new owners do better with luring-clicking, than luring alone.

Crystal said...

LAURA- I agree that I want my next puppy to learn shaping games from the start. I love that Maisy THINKS, even when she sometimes outthinks me. For example, now that I'm getting some pretty solid three second holds WITHOUT the front paw lift, she's decided to add something new: backing up and wiping her feet. Sigh. But it's really funny.

SAM- I'm glad you brought up that some of your tricks still need hand signals. I think the major downfall of luring for us is that I've had a hard time phasing out hand signals and other body cues. Not because of Maisy, necessarily, but because of ME.

LOUISE- I agree that the way you get the behavior doesn't really matter (assuming, of course, you aren't hurting or scaring the dog). Shaping, luring, capturing... even using some light physical prompts or props all work.

DAWN- I can definitely see where you'd use a lot of physical prompting for conformation. I know people do use shaping for stacking too, but physical prompting seems easier when you have those short little legs! I know I have difficulty seeing what's going on with the short ones, although maybe it'd be easier if I put Maisy on a table or something.

KRISTEN- Could you talk a little more on why you think new owners do better with luring-clicking than just luring alone?

Ninso said...

Cool stuff! I use all 4, depending on the situation, but I would agree with using lure-click for pet dog trainers. I (almost)always use a clicker or verbal marker, whether I'm luring or not. I think luring is easier and faster in a lot of situations, and cuts out a little bit of the timing piece of clicker alone, but I also think it's important for the dog to have a marker to tell him WHAT he is getting rewarded for, and helps set up foundation for later clicker work. I would use at least a "yes" as a marker, if not a clicker.

I disagree with the lure being easy to fade though. In some cases, it is. With my BCs it is, usually. With my cattle dog it is not at all. I can fade the food lure, but I still need an empty hand lure. He is not good at putting names to behaviors and uses contextual clues more than anything to figure out what I want.

With Lok and Jun, I used a lot of luring at first, and will lure occasionally still. They know what it means and it works well for them. I had a foster that I could not use luring with at all. He knew how to sit when he came to me, and that was it. I could lure him into a down, but when food was in front of his nose, his brain was OFF and he'd be in a down, but have no idea how he got there or why he was being rewarded. Weeks of working on this and he still didn't get it. When I switched to shaping/capturing with the clicker, he started thinking a bit and it was a lot faster to get some of those basic behaviors.

With my dog Elo I have free shaped (almost) EVERYTHING he knows with the clicker, as an experiment, and also because he suffered a bit in the beginning from the (food out=brain off syndrome). He came to me knowing nothing. I really like the results! I'm not sure if it's the way I've trained him or just the way he is, but he is the easiest dog to shape (and gets easier the more I do with him). He's really creative and truly understands what the click means. He will "ask" if something is right, by making a small motion in a particular direction and then waiting to see if it gets clicked. I can click very tiny shifts in weight and he will take it and run with it!

As for physical prompting, I don't use it often and it wouldn't be my first choice most of the time, but this is one I've been experimenting with. Lok seems to respond well to it. Maybe because he's blind feels more secure if my hands are on him? I've recently taught him a "lie on side" by placing gentle pressure on his shoulder--not enough to push him onto his side, but enough to notice. I clicked him for giving to the pressure. I'm not sure about fading the physical cue, but I don't know that I want to. I'd like him to be able to have verbal AND physical cues, just like dogs who can see have verbal and hand signal. I've also been using it recently to teach Jun to balance a treat on her nose (holding her nose in place), and as I mentioned earlier, I ended up resorting to it to get the first couple seconds of Lok's dumbbell hold. Probably a few other things too. As for props and targets, I find them difficult to fade out, but it's probably just a deficiency on my end.

Like someone else said, different things for different dogs/behaviors. And like you said the other day, the cool thing about training is if one method doesn't work, there is always another to try!

Crystal said...

NINSO- Thanks for the detailed comment! I've only really worked with one dog, so it's always interesting to hear about how it's worked for different people and dogs. I've experienced that OFF brain with Maisy, not with food, but with tennis balls. Good grief, she canNOT think when that thing's out. We've been working on it for over a year, and she's just starting to think. A little.

Ninso said...

My primary dog sport is frisbee and a lot of dogs are like that in the beginning with a disc. Have you tried doing some operant conditioning/shaping while the ball is out? For example. Hold the ball out and wait for eye contact, mark and reward with the ball. Hold the ball out and a target with your other hand. Wait for a target and then mark and reward with the ball. It seems to go faster when you let them figure out what's going to get them the ball vs. trying to give commands. I had to re-teach most of Lok's commands this way with toys out before he started gaining the ability to think around the toy.

Ninso said...

Also, I just wanted to say I've been reading through your archives as I have time and really enjoying them, especially your notes on the seminars you've been to!

Crystal said...

NINSO- Once she figures out that we're shaping with the ball, she plays along just fine. But trying to give cues? She can't hear them. She'll follow hand signals with about 75% accuracy. She's an interesting dog.

doberkim said...

really good question - i guess i would say i use all of them - though as compared to others, i do use a lot for physicality for my dogs.

for foundations, i will do a LOT of free shaping - i want offering behaviors, i want them to think and push me to get a reward. i will also free shape a lot of tricks, etc.

blurring the lines, all my free shaping is with a clicker.

however, with my dog its to the point where they are almost too good at it though - they will fly through a variety of behaviors and if its not right, immediately potentially change to something else - or in the midst of one behavior, will start to bring in other ones just to see if they are rewardable. for rah, this happens so quickly that its hard to mark things. and with a clicker, many times i find that even if i reward attention, that my dogs are viewing me as a pez dispenser. instead of focusing on me and working for me to reward them, they are constantly wondering "did she click, is she going to click" and the clicker takes a lot of the personal interaction i need with them out of the picture. i still use it for foundations, but i will eliminate it from known behaviors very quickly. clicker still remains the best and quickest way for me to free shape behaviors because its a clear "keep working, you're getting it right!"

for many behaviors in competition obedience i DO lure - i lure head position for heeling, i lure the hop for finishes, i lure many things. this is because i want perfect position in these from the very beginning, and the physical memory of how they do it is very important.

then the physicality - i do correct my dogs, and i do use corrective gear. both my dogs known electronic collars, and ive used it to clarify what i dont want them to do just as clear as i want them to do something else. for me, i also use it as a bit of safety issue for me as well - i will never take my dogs off leash without the security of the collar, and my dogs are taught off leash recalls from the baby-times, but proofed later in life with the e-collar. i teach my dogs how to respond to leash pressure, how to respond to restraint, and what collar pressures and pops will mean. i play "beat the pop" games for finishes - ill lightly have the collar in my hand and if you are right and fast, you wont get any collar pop.

i also use physical shaping with my hands as a way to get sits, fronts, stands, and downs. i use sticks and boards for fronts.

for me, i want my dogs to be right from the very beginning and have a solid foundation before they start to guess and test things out - if ive taught them right they clearly understand their job. they may not always do it, but if they were taught it, we have something to fall back on.

Crystal said...

Kim, do you have trouble fading out the physical prompts, whether it's using your hands or using a prop? What about using the corrective equipment- do you have trouble fading that out? Is there anything else you DON'T fade- like the e-collar when off leash?

Eliz said...

I love the comments they are really interesting.

I rarely lure anything. The times I've tried I've failed. I'm not sure if its me or the dog (I'm pretty sure it is me, since I've seen other people lure my dog) but personally I just do not have such good luck with luring specifically anything that new or very different then what we've done before. Of course at the rate I dispense treats (I'm widely known as a cookie pusher) maybe I am a lure, who knows.

The first big success I've ever had is capture, but I have a hard time separating capture from shape. One example, I remember not having any idea how to teach mat. So I thought, well I'll just throw a mat down on the hallway and when he steps on it to go from one room to another I'll click. Soon he is stopping on the mat, awesome four feet on the mat and I'll click. Soon he is stopping there. Now, I have a fairly lazy dog so that helps, once he is standing there long enough waiting for the treat he lays down, I click, and then I have mat behavior - all without moving from my couch. But is it capture or is it shape?

Here is a another thought: Sometimes I wonder if dogs often (not always) learn the best the way they learn first. So if start them with luring then that might be how they learn the easiest. Like learning a second language; the language you learn one first that will always be your native language the one you go back to. Just a thought.

Crystal said...

Elizabeth, I agree. The comments ARE interesting.

Re: luring. I do think it's you. I was luring him this morning. Not much, but he was clearly going where the treats were. Actually, I had a lot of fun working with Beckett. He's so THOUGHTFUL, which is different than how Maisy operates. (You should take her during class sometime.) Anyway, I was mixing luring and shaping all at the same time, which is kind of funny.

I think your mat example is primarily capturing, but I can see how it has elements of shaping, too. Based on that, and my experience with Beck this morning, it seems that you can use two or more methods at the same time.

Regarding how dogs learn: I've wondered about that too. But I also think that the way the person learns to train first profoundly affects things. Maisy and I both learned how to lure first, and I find that I often fall back into luring. She understands it, I understand it. Of course, I blend luring and shaping, and sometimes I shape only, depending on the task, but before the seminar, I never worried too much about the "best" way to train, other than to use reward-based methods.

Tegan said...

A thought on physical prompting: This is not my preferred first avenue, either, but I think it's a matter of picking your cases. When I have foster greyhounds, I often use a lot of physical prompting. The other options are not so good (e.g. greyhounds are not food or toy motivated, in case of 'all or none', geryhounds do a lot of nothing and clicker training is hard to accomplish when they are unmotivated). Additionally, the foster greyhounds are racing animals used to a kennel environment and not being in a house they are stressed and unresponsive.

Perhaps, the only thing that is like a kennel environment in my house, IS the physical prompting. Greyhound people don't teach their dogs anything - they physically manipulate them from one kennel to another, drag them by their collar, pick them up to put them in the car, etc.

I have often described having a new foster greyhound as flooding a lot. These dogs are very stressed when they come into my home, and they're overwhelmed with what a home is like. Perhaps, because they are flooded, they also need the physical prompting to encourage their movements around the house.

And I actually haven't found it difficult to get them past the physical prompting faze. Perhaps it is because I use my physical prompts with word cues from the start, so when I pull/push the greyhound outside, I'm also saying, "outside".

On a different note, I think the dog training world is going to get back into luring pretty soon. I think the clicking/shaping thing was a fun experiment, and has its place, but I think even Karen Pryror is starting to reduce her clicker focus? I will be interested to see how things develop.

Crystal said...

Tegan, I think your comments re: greyhounds are interesting. My friend has two ex-racers in her house right now, one permanent, one foster, and I have definitely seen how physical prompting is easier with them. She also does a lot of capturing- she's always said that's what she's most comfortable with, but I wonder if that's because of who she's working with. :)

Your comment that Karen Pryor is reducing her clicker focus is intriguing. What do you base that on? I do think that luring is great for beginners and "just" pet folks. Shaping, while fun, isn't really needed for a well-mannered dog. And while having a "thinking dog" (Gail Fisher's term for a clicker dog) is fun... it's really, really not for everyone!

Tegan said...

Re: foster greyhounds: I'd like to do something else with them, like capturing, but (at least for the first few days) the dogs find nothing rewarding. In fact, many of them cringe at my touch, so my presence isn't even that rewarding. There is probably a lot of negative reinforcement going on - if they comply with my physical prompting, I leave them alone. The first few days that the greyhound is in my home is the most difficult and, once they've learnt to go to their bed and outside (by physical prompting), then there's nothing more they truly need to learn. However, at this time I could use food rewards, rewards of my touch, or, in some of them, toy rewards. But I've got to do something with the dogs in the first few days, when they are unresponsive to rewards and so stressed I can't expect much anyway.

My comment re: Karen Pryor is more a 'I heard' thing. I did try to Google, but it's pretty much impossible considering her extensive involvement in the field. I believe the comment was made from someone else on a forum, and it was more that Karen Pryor was advocating clickers for training behaviours, and that's it. However, it's all hearsay so close to useless. :D

Crystal said...

Advocating clickers for behaviors vs... what, exactly? I don't think I quite understand what you mean. Because I'm not sure what else you'd use a clicker for? Maybe counter-conditioning, I guess?

Leema said...

Clicking for training behaviours, vs clicking to mark in already learnt behaviours. (Sorry for the unclear.)

Crystal said...

Ohhhhh wow how did I miss that? I guess it's because I only use the clicker for teaching and phase it out as soon as I can. Thanks for the clarification.