In his 1968 book, The Technology of Teaching, B.F. Skinner wrote:
Errors are not a function of learning or vice-versa nor are they blamed on the learner. Errors are a function of poor analysis of behavior, a poorly designed shaping program, moving too fast from step to step in the program and the lack of the prerequisite behavior necessary for success in the program. – BF Skinner
And that sounds great. It also sounds like a lot of pressure on the dog trainer. Never fear! In this episode, we discuss what errorless learning actually means and how to apply the principles in our real life training sessions.
In this episode we discuss:
- The original research on errorless learning by Dr. H.S. Terrace form 1963.
- What is the difference between errorless training and trial-and-error trainings?
- Why we care about training with errors – what’s in it for us dog trainers?
- Designing our training sessions so that we don’t rely on extinction as a training tool, and in fact actively structure our sessions to minimize our dogs’ experience of extinction.
- Strategies allow us to train more errorlessly
- Fading in discriminations – early and often
- Using back-chaining, even when shaping small behaviors
- Introducing a new element between the click and the treat
- Setting your minimum criteria to earn a click to behaviors your dog is doing frequently, and avoid raising criteria to something your dog has never done, or only done once.
Links mentioned:
- Episode 97: Errorless Learning Primer
- Discrimination Learning With and Without “Errors”
- Tsk, No, Eh-eh: Clearing the Path to Reinforcement with an Errorless Learning Mindset
Support The Show On Patreon!
Thank YOU for being a part of our community! I couldn’t do what I do without you. To become a tangible supporter of the show, please visit our Patreon page and share the word in your circles.
Episode Transcript
Hannah Branigan: [00:00:00] So, you know, something like a sit or a down? You think that that’s a single behavior. But if you’re sufficiently nerdy, you can see how many different movements go into the complex beauty of a perfect fold-back down.
[intro]
Hey there, fellow training nerds! You’re listening to Drinking from the Toilet. If you like to geek out about combining the science of behavior with positive reinforcement philosophy in real life, you’ve come to the right place. And I’m your host, Hannah Branigan: teacher, trainer, podcaster, and author of the book Awesome Obedience and its companion, Awesome Obedience: The Field Guide, which are both available from clickertraining.com.
In this episode, I wanna talk about errorless learning and how we can apply that in [00:01:00] our training.
And I know errorless learning is a topic that we’ve spent some time on before, but I felt like it was time to revisit it. It’s the sort of thing that I have to revisit for myself in my own training on a regular basis and I do think that the way that I think about it evolves – hopefully clarifies – over time, and I thought that might be true for you as well. So we’re gonna dig into that.
But before we do, I gotta send a shoutout to some awesome folks who support this podcast on Patreon. So big thanks to Divya E, Katie H, Olnagh[?] G, and Delinda D. I love you, you’re awesome and I really appreciate you.
If you would like to support the podcast, get your questions answered, and get access to some super-secret extra podcast episodes, you can go to patreon.com/DFTT. In fact, our next patrons-only Q&A I believe is on Thursday, July 20th. We have them every [00:02:00] month, so if you wanna get in on that clock’s ticking.
[Episode begins]
Okay, so let’s talk about errorless learning.
So whenever we’re having this conversation, I really like to start with talking about– well, at least what I know about the history of the procedure and where the term came from.
Okay. So in 1963, Dr. Herbert Terrace published an article about teaching discrimination tasks without any errors. So, of course, for context, traditional discrimination tasks were taught using that old trial and error method, right? You put out the right answer, they guess that one, they get a reward. You put out the wrong answer, if they guess that one, nothing happens. And so you figure out what is the right answer by just guessing right or wrong and collecting rewards for the right answer in theory – or the right behavior, or the right response, whatever the thing is.
And so Terrace was exploring how to teach this discrimination task without using the [00:03:00] errors. So he did a couple of things that were really different.
Now, the task that he was working with was teaching a pigeon to peck a button when a light is red and not peck the button when the light is green. The idea here is that if the light turns red and the pigeon pecks the button, he gets reinforced. If the light turns green and he pecks the button, he’s not reinforced. So using a trial-and-error teaching strategy, the pigeon should learn to stop pecking when the green light was on because he never gets reinforced. That should extinguish, right? And then only – or mostly only – peck when the red light is on. So that’s the idea.
The standard way to teach that is that first, you teach the pigeon to peck when there’s a red light on, so turn the red light on and the pigeon pecks the button, gets the reward, pecks the button, gets the reward, pecks the button, gets the reward. And then you add in a green light [00:04:00] and the pigeon will peck and then you just don’t reinforce.
And so initially the pigeon is pecking the button no matter what. Doesn’t matter what color light is on. Then there’s a lot of errors and a lot of extinction. But since the machine is set up so that only pecks when the red light are on get rewarded, hopefully fewer and fewer green light pecks happen in time.
Now, Terrace comes in and he designed a different procedure that had two major differences here. First, and I get very excited about both of these– First, he introduced the green light really early in the process, so instead of building up a huge reinforcement history for pecking when the red light is on and only having red lights in the picture, he started to add the green light into the picture really, really early in the process, and he started off with that green light at a much lower intensity. It was not as bright, it was a shorter [00:05:00] duration, and I think he was messing with the wavelength as well, so it was further away, so it looked the least like the red light, right? Then gradually over time, he increased the intensity, so made it brighter, made it more green, I guess, and longer duration while continuing to reinforce when the pigeon was pecking the [red] light.
And so the results were that the pigeons that were trained with his errorless discrimination procedure only made about 25 errors. So 25 times an ish, they pecked the green light or they pecked the button when the green light was on, compared to the pigeons that were being trained by the more traditional strategies, which made on the order of thousands of errors, a hundred times more errors than the pigeons that were trained without errors, with the errorless procedure.
So that’s part of what’s really interesting to me, right?
So introducing it [00:06:00] very early, introducing it at a much lower intensity. And then here’s the other part that I always find very fascinating about this sort of stuff. So you end up with two cues, right? This is a stimulus control thing to a certain extent. You have a cue of the red light that means “pecking will be reinforced because the red light is on,” and then you have a cue of a green light, which means that “no peckings gonna get reinforced.” So we have the SD and the S-delta, the discriminative stimulus and then the “I’m so sorry, store’s closed” cue.
Pigeons that are trained with the trial and error approach exhibited emotional responses when the green light was turned on. So when the light that came on that indicated like, “Oh, you’re not gonna get rewarded now. You can peck all you want, but it’s never gonna work,” they would exhibit emotional responses. They would flap their wings, they would attack other birds, behaviors that, from the pigeon’s perspective, we might describe as frustration. [00:07:00] Even rage. Well, that’s really interesting.
Whereas the pigeons that were trained with the errorless approach stayed calm even when the green light was on and then only got excited when the red light came on because then they would peck. So they didn’t experience the green light as frustrating. At least, that’s the conclusion I’m drawing here.
And that’s really cool because those are things that I think are very applicable to my training. Those are things that I wanna keep in mind when I’m thinking about how I’m training my dogs and honestly how I’m training my human learners as well. Certainly my horses, my child, myself, perhaps. I don’t know. Well, I dunno, that’s a stretch. Definitely my human learners. Anyways.
So this is a really cool concept and we’ve talked about it a lot in the dog world. I’ve seen it come up more and more, which I think is awesome. I wanna keep it in the conversation, which part of why I’m talking about it now.
Because I think that a lot of what we can take from this research [00:08:00] really looks a lot like good training when we’re already doing good training. But there’s also some confusion at times and sometimes some arguments, but I think a lot of those arguments come down to just not really not being on the same page, not really knowing what each other are talking about (which is of course the case for many arguments in the human world).
One of the things that I’ve heard other trainers say is, “Oh, well, you know, some dogs can learn with errorless training, but some dogs learn better through trial and error.” Ehhh, maybe. But at this point, we’ve studied this with pigeons, with rats, with lots of other species, including humans in lots of other applications. There’s some really cool stuff that I ran across using errorless learning principles working with people who were recovering from traumatic brain injury. Really cool stuff! You can look that up yourself. There’s things on the internet. So whenever I have a training concept that I know works across [00:09:00] species, like we have good evidence for that, then it seems very likely that it’s probably going to work within the same species as well. So, you know, german shepherds and labradors aren’t any more different than dogs and pigeons or rats and pigeons or humans and rats. So there’s probably something there.
Now, when we’re talking about behavior, it’s always complex, so it’s probably not gonna be super simple. What is an error in a particular case? What is the problem that we’re trying to solve? What do the behaviors look like? There’s a lot of details that go into there. But it probably isn’t that some dogs learn better through trial and error. So if we find ourselves being caught up with that, that explanation, maybe we wanna look a little bit deeper. There might be other stuff going on there as well.
And then one of the other things that I see come up a lot – and sometimes this is a criticism of positive reinforcement training in general – but [00:10:00] it’s the idea that, “Oh, well if you’re using errorless learning, it’s because you’re never challenging your dog. You’re not ever taking them out to the real world where mistakes happen.” And again, I think that’s a misconception: both that positive reinforcement trainers don’t ever take dogs out into the real world. There’s a whole nother podcast maybe series in there. That’s different. I don’t wanna get too far down that rabbit hole. So that part’s a misconception.
But also the idea that if we’re training using errorless learning principles, that we’re never gonna challenge the learner. And I would circle back to Terrace’s original experiment where it’s not that he never turned the green light on. He turned the green light on and he went from it being very, very low level, very short duration to full intensity, full duration, just like the pigeons that were trained with the trial and error condition.[00:11:00]
So it’s not that. We get to the same place! But I do think that that’s one that we need to look a little bit more carefully at ourselves as well, because I do think sometimes when we’re trying to think in terms of errorless training, we can get kind of bogged down. And this is more to do with I think just like human psychology and probably our childhood trauma and less to do with anything specific to dog training.
But when we’re trying to do things The Right Way, and we’re like, “I wanna train errorlessly!” And so we can get really hung up on the idea of, “Well, if I let an error happen, it means I’m a terrible trainer and I’m doing a terrible job.”
And that’s not true either! Again, we wanna get to the same place and I think part of it is in the language. I think that’s one of the places where we can get a little bit tripped up. So when we hear the term “errorless learning,” it really sounds like it’s zero errors. Like the errors that occur are none. That is the only thing that is [00:12:00] error-less. It’s gluten-free dog training, right? There can’t be any errors and it would still count. And that’s not in fact the case! With Terrace’s pigeons, his pigeons did peck when the green light came on – they just did a lot less and he wasn’t depending on the lack of reinforcement when the green light was on to do the heavy lifting for the training plan, right? He was primarily fading in that green light incrementally and that’s really what was doing the work of teaching that skill to the birds.
So if we’re thinking about errorless learning as never letting your dog make a mistake, that does imply a lot of rigidity and a lack of motion and a lack of progress. Because I mean, the best way to never make a mistake is to never try, right? I mean, I’ve tested that one out myself! But that does keep the world very small and never challenges the learner and yeah, you don’t get anywhere. So that’s not [00:13:00] what we want. That’s not really the solution.
So, there’s a big difference between never letting your dog make a mistake and deliberately using mistakes as a training strategy, whether with extinction or punishment, depending on how you’re training. There’s a lot of space in between that and it’s sounds like the same words in just slightly different arrangement in the sentences, but it’s a very different meaning.
Alright, so before we get into what that looks like in practice, I also wanna spend a little time thinking about why I care. Why is it so important to me? Why do I spend so much time and energy trying to design training sessions around errorless learning principles?
Well, the first and probably one of the most important reasons is what we were talking about at the beginning with regards to the emotional behaviors that we see with learners who are in a trial and error situation. It’s frustrating. Extinction is aversive. [00:14:00] It’s very unpleasant. I hate feeling frustrated. I think frustration is probably one of the experiences that I have the lowest tolerance for. I’ll take pain over that! Like pain and then like maybe itching and then frustration and then nausea. Because I think nausea is the worst. But anyways, I digress.
The experience of extinction is extremely unpleasant and we know that again intuitively ourselves, but we can also make some guesses based on what we see with animals. We know that if you give these pigeons a means to turn the green light off, they will. So if you give them another button and they pick that button, it turns the green light off, and then they can have their red light back on and they can peck their button. They’ll choose to turn the green light off, which indicates that there’s probably a negative reinforcement contingency in play there so they can avoid or escape that green light situation. And who wouldn’t? I would do exact same thing!
Okay. So there’s that. When I’m [00:15:00] training my dogs, I don’t want the signal that reinforcement is not available to be experienced as frustrating. Both because I care about my dog’s emotional experience – I care about their welfare and what their experience of life with me is like – but also a lot of the behaviors that dogs and horses and children and people (children are also people), a lot of the behaviors that they do under frustration/extinction conditions are really unpleasant to experience from my side as well.
So here’s an example that I think is very similar to this red light/green light situation. A question that comes up a lot– It’s been asked at some of the patron Q&As and I’ve demonstrated how I handle it, and we talk about it in Zero to CD as well, is how to teach your dog to wait their turn. Turn taking is a big deal in my house. I always have multiple dogs in training at any time and I’m just one person. So while I’m training one dog, everybody else [00:16:00] has to wait their turn. And it’s very difficult for me to focus on training one dog if the other dogs are screaming and carrying on behind me. So when I’m working one dog, usually the other dog is popped in a crate or they’re on station depending on a lot of factors, such as the age and training of that particular dog, and they’re waiting their turn. But that means that when I pop you in a crate and turn to address the other dog, that means that reinforcement is not available to you. So you could be offering all the behaviors that you want while you’re in the crate and they’re not gonna get reinforced.
And I could teach that – and for “scientific purposes,” I have in fact taught that mostly using extinction, right? Like that’s what we say when we wait it out, “I’m gonna wait the puppy out until he stops screaming before I open the crate door.” And so if I put you in the crate and I’m training the other dog, the other puppy is screaming and screaming and screaming and I’m not gonna let him out until he stops screaming. I am hoping that crate screaming will [00:17:00] extinguish. So I’m hoping extinction will do that job for me. And then only rewarding when he’s quiet and then he only gets reinforcement from me when he’s quiet.
I don’t love that. I don’t love that strategy. For one thing, it has resulted in some fallout, some behavioral and emotional fallout with dogs that I’ve worked with. When it’s time to wait their turn, more screaming happens, other things happen. I’ve even worked with clients whose dogs redirected on the other dogs in the household under those circumstances. Dogs definitely will resist going back in the crate. Like you’ll see a lot of stuff show up there and I don’t want that at all. I want my dog’s experience of “now it’s your turn, now it’s not your turn. Red light, green light, green light, red light.”
(The color of the lights gets me tangled up. So if I’m swapping them while I’m talking here, it’s because in my mind, I feel like green light should mean go, and red light should mean stop. But my understanding from the reading that I’ve done is, it [00:18:00] was the opposite of that in this case. But hopefully you’re tracking me here.)
Okay. So another way that I could do this is I can make a clear cue picture for when reinforcement is available and when it’s not and fade that into the picture. And that’s exactly what I do that I find to be the most successful. And so what I usually do is, let’s say I have a new puppy. I pop that puppy in his crate and I turn my back on him briefly momentarily. He doesn’t even have time to register what’s happened and then he’s back outta the crate and he’s getting rewarded. And now he’s back in the crate, I turn my back very briefly and then he’s right back out and he’s being rewarded. And he’s back in the crate and I turn my back again just for a second and then he’s back out. And then maybe I pop him in his crate, I turn my back, I give maybe an older dog a little bit of attention, maybe a few treats for training and then turn right back to the puppy and I let him out.
So I’m keeping that loop very clean so that [00:19:00] hopefully me putting the puppy in his crate and turning my back so that I can address another dog has a really clear cue associated with it and that cue is not associated with all of the experience of frustration and screaming and biting the bars and all things that can happen if I’m waiting and depending on extinction in that context. It goes actually fairly quickly as long as I’m honest with myself and I keep those loops fairly clean. I can extend that up in a fairly short amount of time to the point where I’m able to do a real training session with one dog while that puppy is waiting his turn in the crate and then he comes out and does his.
And it’s both going in the crate and my turning my back to the puppy and then turning back towards the puppy when it is time to be rewarded. That’s the red light, green light. Two different cues that make it very clear when reinforcement is available and when it’s not. And the “reinforcement not available” condition is never associated with a lot of frustration because it’s never associated with extinction. And the way that [00:20:00] that works is, again, by manipulating the duration there. So that’s an example where it matters.
But I think that can be true in a lot of different places, right? There’s a lot of places– I really don’t want a lot of frustration in my training relationship with my dog in general because if frustration doesn’t look like screaming and biting the bars of your crate, it can look a lot like sniffing and wandering off to find reinforcement elsewhere. So again, not something I want as part of my training picture. So I’m very, very careful with that.
And the other reason that I pursue errorless training as a kinda a core strategy for how I think about dog training is in addition to those emotional behaviors, we just have the errors themself. And one of the things that we know when we think in terms of loopy training, which is compatible with errorless training, is that anything that happens in the loop is gonna tend to get built into the loop. So if my dog tries two or three incorrect responses and then gets the correct response and then gets [00:21:00] rewarded, those errors tend to get built into the loop, which I think is really interesting.
And I’ve certainly had this experience. I think I’ve talked about this before. When I switched where I was working, this was years ago. This was more than 10 years ago and I had to turn a different way out of my driveway. I repeatedly would leave in the morning, I would turn the wrong way out of my driveway, the old way, and then have to go around the block to make a 270 [degree turn] and end up going the other direction. And I think I worked at that new clinic for I don’t even know how long and I would continue to turn the wrong way, do the 270, ended up going the right way. And that just became the new behavior. But that’s not the most efficient way to solve the problem. But it got built into the loop!
And we do that a lot with our dogs as well. The error may diminish, it may attenuate to some extent, but a lot of times it doesn’t really go away.
One of the resources that I found that I was talking about errorless learning with traumatic [00:22:00] brain injury patients, he said it seemed like they remembered what they did and not what was the right answer, when they were trying to do whatever the treatments were to help them rebuild their memory and their skills with their memory. If they were using trial and error and they would make an error, and they would say, “no, that’s not it,” they’d make an error and they’d say, “no, that’s not it.” And next, they would get the right one, and they’d say, “yes, that’s it.” And they would come in the next day and they would ask the same question and they would get the same incorrect response. It was like they were practicing the incorrect response before the correct response. And I feel like I’ve observed this happening with my dogs as well, so I was like, “Oh, that’s really interesting.”
So I wanna keep clean loops because that keeps my training cleaner. It also means that while I am fading in any aspects of the environment, any discriminations that may be part of the behavior that I’m trying to train, but I don’t have to clean out a whole bunch of junk. I find that it’s a much slower [00:23:00] process to clean out junk behavior that gets built into a loop. It’s much harder to do it without taking everything apart than it is to train it cleanly in the first place. And that’s one of the things that we talk about when we say, “Oh, you know, taking the time to do it right.” But I’m gonna say that taking the time to do it right doesn’t actually go slower. It gets us a stronger outcome, yes. And it’s also not slower. I think it’s actually faster. It takes more cognitive effort on the part of the trainer though, and that’s not nothing. That’s a real thing.
Alright, so we’ve bought in, we wanna get where we’re going faster, we want clean behaviors that don’t have errors built in, that don’t have junk behaviors built in that we don’t like, and we want all of that with without a lot of frustration, certainly without aggression. So we’re committed to that. We want to design our training sessions so that we’re not relying on extinction as a training tool. But in fact, actively structuring the sessions so that we minimize our dog’s experience of extinction in that context.
So [00:24:00] this does mean changing how we think about things like shaping. I know that a lot of our early pictures of what shaping should look like, at least those that I was exposed to in the dog training world, involved a lot of sitting around and waiting for the dog to just spontaneously emit the right behavior. And a lot of dogs, and I think humans, can get frustrated in that situation. Again, very trial and error with that. I did a lot of shaping that way. And guess what? Those poor dogs, bless their hearts, they were so good to me. They did not eat me while I was asleep and they could have and should have. It would’ve been deserved.
But they were wildly guessing, I was waiting, and they did end up with a lot of frustration-based behaviors that were looped in. Typical clicker trained dog, 300 different behaviors, none of them on particular stimulus control, lots of them involved noise and general frantic behavior because of the training process.
And I think that that also contributes to the myth that shaping [00:25:00] is hard and that clicker training is unfair because it’s not clear because it’s lacking the information of having both sides of the hot and cold game.
And I think that honestly, those are valid criticisms if we’re looking at shaping and clicker training and positive reinforcement training if that was our baseline, but we can do better than that. We can do a lot better, but we do need to think about it a little differently.
So I wanted to talk about some specific strategies that I find really helpful. And this is not an exhaustive list. These are a couple of the strategies that I find really helpful in allowing me to train more errorless, right?
So again, we’re not trying to completely eliminate errors. We’re not preventing any mistakes ever from happening, because that’s not realistic. That’s the other thing, right? It’s like, “Oh, I’m gonna keep my dog wrapped in bubble wrap.” That’s not possible because I’m not that good. All things being equal, I’m just not that good of a trainer. I’m gonna ask for a little bit too much at times. So there’s gonna be errors that happen no matter [00:26:00] how hard I try to be as errorless as possible. So me shooting for errorlessness means I’m still gonna get some percentage of of incorrect responses just because I make mistakes with the training. So anyways.
Okay, so more errorlessly, moving in the direction of fewer errors and less emphasis on extinction and frustration.
Alright, so the first one, this is one I talk about a lot in a lot of different contexts. I know we talk about it a lot in Zero to CD, which is to fade in any discriminations or distractions. Fade them in early and often.
So rather than training in a vacuum for a very long time and creating a really rigid training plateau honestly, and then trying to add a distraction afterwards, we wanna introduce those elements right from the very beginning, like second session, maybe the first session, depending on what the thing is. If you know that there’s going to be something that’s gonna be part of the final picture that your dog is gonna need to discriminate, we want that [00:27:00] present in the environment from the very beginning.
So go ahead and train your A-frame with a tunnel in the picture, but make it pretty far off. Maybe turn the open ends down or scrunch it up so it’s not inviting. Turn it to face the other way. So it’s still there, it’s around, but it’s not inviting. It’s not really lit up. It’s very dim and it’s not on for very long. Right? So we introduce it at a lower intensity.
Do you have a behavior that is going to– If you’re teaching a behavior where you know that another human is gonna be part of the picture and you want your dog to ignore that person, include the humans in really low key ways from very early on in the teaching process, but again, at a distance, very low volume, very low intensity. So just doing the training with other people around.
It’s not that hard if we’re thinking about it. I’m thinking we end up doing that often anyways, just when we’re going– if you’re lucky enough to have friends to train with, you just go happen to be training and your friends happen to be also training their [00:28:00] dogs. That’s a good example. But then you can fade that person in closer and closer and closer and make them “louder.” Also, actually, possibly just loud without air quotes, possibly in fact literally louder. But you can fade it in.
I do think that that’s huge. Don’t wait until you feel like the behavior is perfect and then try to add a distraction to it. Have them present in a very low level kind of way from the very beginning. It makes a huge difference.
So that’s one of them. Another strategy that I love that allows me to train more errorlessly is to use back chaining. So you can back chain just about anything because even behaviors that you think of as just single events like a sit can be back chained if you consider the components of the movement.
I recently was having a conversation with my child and I had taught her to memorize our address by back chaining. So she memorized the last part first and then walked her way [00:29:00] forward. Because I always like to be prepared for whatever catastrophe. Safety first. I’m a safety girl. I’m a risk-averse, high-anxiety person. But anyways, worked really, really well. And then I did my phone number the same way.
So you know, something like a sit or a down, you think that that’s a single behavior. Oh, but if you’re sufficiently nerdy, you can see how many different movements go into the complex beauty of a perfect fold back down.
I love to back chain fold back downs. In fact, that’s most of the time how I’m approaching it with most dogs these days because it’s just so quick and it’s so elegant.
What do I do to back chain a fold back down? Well I start with the dog in a down position and I’ll reward them there so they’re in the sphinx down and I can lure them to arrange that to make that happen. So they’re in the posture that I want and I feed them with their chin low and their back flat because that’s the last part of the action needed to fold back into the down. [00:30:00] And I know I have videos of this on YouTube, so if you don’t know what I’m talking about with regards to a fold back down, it’s that beautiful– it looks like they’re on hydraulics. They just push back into the down; their feet don’t come forward at all. It’s one of my favorite behaviors.
Anyways, so they’re in that sphinx down. I’m rewarding with their chin low. I often use a chin target to teach this behavior, so I’ll put like a lid or a coaster or something between their front legs so that they’re putting their chin on the ground between their front legs. And then I deliver my reward a little bit higher over their toes. So they’re having to lean up and forward to take that reward. And then they lean back down into that perfect sphinx down, and then I feed them up a little bit higher than that and they lean up and forward. And then as they swallow the treat, they just relax their muscles and they go right back down because gravity does the work for us.
So they’re pretty close– It’s almost impossible to make an error unless you’re gonna really fight gravity here. And then I just increment how high I’m delivering that treat so they’re coming more and more into a stand position before letting gravity pull them [00:31:00] back down to chin their chin target or to get their elbows onto the ground and rewarding there.
And I can build it all the way up forward, back chaining the action, always moving towards the easier thing. Gravity’s on my side. And it works really, really well. Sometimes I might add other props to, again, contribute to the errorlessness of the session. So I might put their back feet on a low target or low platform, might put their front feet on something. I might put them on an incline if needed for that dog. So there’s a lot of different tools, but the one–
I’m getting excited talking about fold back downs, and what I really wanna talk about was back chaining. I also love back chaining for finding heel position. Same idea. I take a dog that already has some installed motor skills, so they already have the ability to pivot, they can look up at me, we have reinforcement strategies on board. I’ll start them in heel position either by luring them to my left side, or I’ll just put a treat on their nose and I’ll move myself to [00:32:00] their right side. (Did I get that left and right correctly? I think I did.) And I’ll feed there several times and then I’ll deliver that treat just a little bit out to the side. They take that treat and then they just snap back to that heel position. It’s not even a step. I reward there and they snap back two steps and reward there. And then that next treat gets delivered a little bit further out of position and they just–
In this case, it’s a gravitational pull that I’ve manufactured with my reinforcement history. And then I just increment again, increment that treat delivery to be further and further out of position so that they’re finding their way, taking a few more steps, a few more degrees of pivoting to land back in a straight heel position.
It’s clean, it’s beautiful, it’s my favorite. Having screwed up training heel position with a few dogs earlier on and one dog in particular, my dog Stormy, learned heel position with much less elegance and a lot of frustration and it got built into that behavior so that she would bark or squeal or yip or make [00:33:00] some kind of vocalization, some kind of monkey noise when I would cue her to heel and it was a direct result of it getting wrapped into the training process.
But with back chaining heel position, especially with a dog that already has those component skills. I’ve not had that happen. It’s a much cleaner way to approach it. Much safer.
I also use back chaining for things like putting on a coat or blanket for animals, for dogs and horses.
So I can, for example, teach Rugby that I’m gonna snap that last snap around his waist that closes it and then back up to lifting his paws and then putting his head through. And so putting it all together: he puts his head through, I lift his paws, I snapped the snap. And that’s the blanket on, but I teach it through back chaining, which makes it familiar. Again, very errorless, less junk, less frustration, less weirdness.
Another strategy that I really love to use is to introduce any new element between the click and the [00:34:00] treat first. This is something I learned from Kay Lawrence and I love to use this. I was using it this morning in a training session. An example of something that I love to introduce in this way is any kind of handler movement. So here’s an example with handler movement.
If I’m the handler and my dog is my dog, if I’m going to need to be able to move, so on a sit/stay or stand/stay for example, or any kind of stay, I’ll say sit/stay for this example. I wanna be able to move around my dog and be able to move my body in all kinds of different ways and I want my movement not to be a cue to do anything. I want my dog to continue to hold the sit/stay no matter what I do with my body once I’ve cued that behavior.
And so I might cue the behavior and then mark when my dog has met criteria for whatever I’m working on. And then between my click and the treat delivery, I’ll take a step back. So this maybe hard to visualize here, but I’ll take a step back while reaching forward with the treat. So my body’s moving [00:35:00] backwards, but the treat is moving forwards.
And then I’ll do the same thing. I’ll mark and I’ll take a step to the side, even as I’m moving my hand out to deliver the treat so that my dog’s head is straight. And I’ll do it the other side. I’ll mark and I’ll take a step to the other side and then I’ll still deliver the treat in the same place, but I’m moving my body during that whole food delivery.
So that he’s seeing that it’s a distraction really. He’s seeing that distraction, but he’s seeing it between the click and the treat on that side of the loop, not between the cue and the click. And once we have that now, again, it’s really kind of back chaining. Now I can just as he’s completing the behavior, I can take that step back, click, and then come forward with the treat.
And then eventually I cue the behavior, I take the step back, I take the step to the side, I mark, and then I come back in and reward. Assuming that I’m feeding in position, which I don’t always do.
Now, I will also frequently include the dog’s behavior between the click and the treat. And this is another one of my favorite ones. Let’s say it’s a moving behavior and I want my dog [00:36:00] to continue in a particular trajectory. Well, that’s where I can deliver my treat so that he continues the behavior.
So let’s look at a spin example. Now I could wait and mark when my dog does a quarter spin and reward, and then wait and mark when he goes half of a spin and then mark and reward. Or I can capture that first head turn and then deliver my treat so that what happens between the click and when he contacts that or when his little nose touches the food in my hand is that he continues a couple more degrees, another quarter turn around that spin, and now I mark and I can deliver the treat again ahead on that path, and I mark and I deliver the treat ahead on that path.
So, I’m not making a half spin the criteria to get a click and a treat. I’m still marking when he does just the initiation, but then delivering the treat so that he continues [00:37:00] on that path. Again, then it’s very easy for me to just slide that click forward in time (backwards in time? forward in time?) so that he’s already in motion. He’s already got momentum going. So he’s making some moves and then he’s at the point where he would expect the click and then follow through for the treat. But I get one more step and then I click and I continue the treat delivery on that trajectory so the dog’s movement can continue such that the behavior that I am trying to shape occurs between the click and the treat before I start to ask for it to occur between the treat and the click or the cue and the click in a cued loop like that.
I love using this one. I use this one all the time. Again, I think if you’re combining this idea with strategic reinforcement, it unlocks a lot of stuff that makes shaping way less hard than it might otherwise seem to be. And my favorite thing about that – and this is something that I talk about was shaping a lot – is [00:38:00] it reduces their frustration, it makes it more errorless because you’re never raising criteria to something that your dog has never done before. Because he just did it. He just happened to do it on the way to the treat.
And again, I think that’s where we can get in trouble with shaping. We raise criteria or even start a shaping session. “I’m gonna wait for my dog to retrieve a beer from the fridge.” We may wait a very long time! That was an extreme example, of course. We’re waiting for either a first step that the dog has never done before or maybe he’s only done it once, or we get the behavior going and then we’re gonna raise our criteria to something that the dog has never done before. And again, we’re gonna be waiting a very long time.
So if we can arrange our training session, the placement of the props, the using something like back chain, using the placement of the reinforcement so that it is much more obvious he’s already done that behavior, then it’s much more likely to happen. And then I’m gonna add in good shaping strategies like while I’m [00:39:00] continuing to shape, even if my dog does spontaneously do the next layer of whatever the behavior is, but he’s only done it the one time, I’m not gonna suddenly raise my bar to “that’s my new minimum.” So if my dog happened to do a half of a spin and I clicked that, but he’s only ever done that once, I’m not going to hold and wait out for that thing that’s happened one time to now be the only way you can get reinforcement again. You’re gonna get a lot more errors.
So I will continue to reinforce the earlier step, which was happening a lot more often, and anytime he goes above and beyond I’ll reward that one as well so I can move the whole range of reinforcement behaviors in that direction. (I’m gesturing to the right, but that doesn’t help you at all.) But moving the whole range of behaviors without cutting off 60% of the dog’s responses and dropping the rate of reinforcement, which increases a lot of errors and is gonna increase some frustration.
Okay. So I have [00:40:00] more strategies. Maybe we’ll do another episode to follow up on this one with strategies that I love to use that help reduce my dog’s experience of extinction, that reduce errors, that allow us to train more errorlessly.
And if you’re interested in hearing more of those strategies, let me know. You can find me on like all kinds of social media these days. I don’t know where I am, I don’t know what day it is, who are any of us right now! But if you can find me on social media, for example, if you were able to locate the post that goes along with this episode on Facebook or Instagram, let me know if you wanna hear more and let me know what your ideas are for training more errorlessly.
[outro]
Thanks for listening! If you liked this episode, well, you have good taste and I hope you’ll hit the subscribe button on your podcast app to make sure you don’t miss the next episode. It might be even better than this one.
If you are already subscribed, well, thank you. I really appreciate it, and there are still some ways that you could reinforce me if you were so inclined. You could always leave me a five-star review on iTunes or Stitcher [00:41:00] or wherever you happen to be listening to this podcast. And you can also check out and support the sponsors because they help make the podcast possible.
You can find links and information about them and the other things that we’ve talked about in this episode by going to the show notes, which can be found at www.wonderpupstraining.com/podcast. And while you’re there, you could also pick up a free PDF training template to help you plan your training sessions. There’s also some other articles and previous podcasts and that sort of thing, which you could always find if you were interested.
So until next time, happy training.
