The fabulous Sara and Layla demonstrate "breathe."
Believe it or not, Sara actually taught Layla to take a deep breath when cued.
Kathy’s session on developing cueing skills changed my life. Okay, maybe that’s a little dramatic, but it was possibly the best session I attended all weekend. I have to admit, I suck at getting behaviors on cue, so I knew going into it that this would be full of great information. Even so, I was not expecting to feel so devastated about my subpar skills. In a panic, I begged my way into her lab on cueing skills so I could see it in action. I'm glad I did!
In my last entry, we talked about how to get behaviors. That is the first job of a trainer, and despite my feelings to the contrary, Kathy said it is the most difficult part. Personally, I think getting the behavior itself is far more fun than the tedious process of getting the behavior on cue. This is probably why Maisy has virtually nothing on cue (at least not reliably). That’s not her fault, of course- as Kathy pointed out, the reliability of the dog reflects the reliability of the trainer.
Despite the not-fun-ness of it all, it’s important to develop reliable cues so that you can get the behavior you want, when you want it, and also so you aren’t getting the behavior when you don't want it. (This is also called "stimulus control.") Who hasn’t seen the clicker dog offering behaviors willy-nilly? It’s kind of exhausting of watch, and having one of those dogs myself, it is also kind of frustrating. Still, I created this monster, so I can’t get mad at her. It’s time to fix (um, okay, get) those cues!
Let’s start with some information on cues in general. Cues don’t make the dog do the behavior; astute readers will remember that in Kathy’s talk on shaping, she said consequences drive behavior. Those consequences- the reinforcers- provide the motivation for the dog to perform the behavior. Cues simply provide the clarity of “now would be a good time to try that behavior.”
Throughout her talk, Kathy compared the idea of cues to a green traffic light. If you’re sitting in a car at a green light, you don’t go because the light makes you, you go because you want to, and the green light is a cue that gives you permission. What’s more, you can’t go at a green light if you don’t know how to drive, so you always have to get the behavior first, and then attach the cue.
This is different from how people used to train. In the past, trainers would say a word like sit and then make the dog do it by physically manipulating or luring him into position. Kathy calls these words “commands,” which she distinguishes from “cues.” Commands carry an implicit threat: do it or I’ll make you. Cues are simply an opportunity; if the dog doesn't do the behavior, he won't be forced. However, he needs to do do it if he wants to earn reinforcement, and assuming the dog has been adequately reinforced in the past, he should be excited for that opportunity.
It is this last point that absolutely fascinated me. If cues are an opportunity to earn reinforcement, then they should be pretty awesome things, right? Trained well, a cue should be like a release word. They tell the dog, you no longer have to wait, you can do that behavior now, and when you do, I’ll click and treat you.
The cues therefore become reinforcing in and of themselves because they become what’s called a tertiary reinforcer. Primary reinforcers are things that the dog likes inherently. Dogs don’t need to be taught to like hot dogs, they just do. Secondary reinforcers are things that predict a primary reinforcer. Clicks or marker words tell the dog a piece of hot dog is coming through the process of classical conditioning. The tertiary reinforcer predicts the secondary reinforcer which predicts the primary reinforcer. The cue predicts a click which predicts a piece of hot dog. Cool, huh?
Anyone who wants their dog to do complex behaviors, or a chain of behaviors, will recognize the value in having a cue act as a reinforcer. Agility or obedience dogs often need to do a series of behaviors with no primary (food) reinforcer. Being able to reinforce a behavior simply by giving the cue for the next exercise or obstacle will help sustain motivation and ensure that the dog continues to perform over the long-term.
However, there is a catch- in order for a cue to act as a reinforcer, that cue must always predict a good thing. If you mix in corrections, the cue is then sometimes associated with unpleasant things and no longer acts as a reinforcer as a result. Therefore, Kathy said that training positively is not about fairness to the dog- it’s about giving yourself more tools. Losing a reinforcer by using punishment is a big price to pay.
Whew! Who knew there was so much to say about cues? But there’s so much that Kathy talked about them for 90 minutes, and then did two 90 minute labs about them! I only attended one of those labs, but in my next post, I’ll share her cue tips with you, as well as information about how to add a cue to a behavior. In the meantime, I can't wait to hear what you guys think. Have you ever used a cue as a reinforcer, or does this concept just blow your mind? Let me know!
19 comments:
I think so, am I? I often will use a retrieve (get it) cue as the reward for good heeling. is that right?
It is boring to get behaviors on cue! And it also implies getting far enough along with the trick that it's perfected enough to name it!
I also don't worry too much about stimulus control. Lance is a prime example of a dog who throws out behaviors to other people, especially his beg and pose. It doesn't really bother me and he doesn't do it too often to me. I figure the only time I would care would be in the ring but then he's given other cues for duration behaviors that would override any offerings.
I love the analogy of a green light! But at my house I would say that while most of their behaviors are cued, some are commanded (sit, down, stay) as I will follow up and enforce them instead of just withholding the reward. Luckily in most situations withholding the reward also enforces it, as in a sit to go outside.
I have always wanted to train a new trick with a cue as a reinforcer but haven't done so yet. I also can't think of too many times I purposefully use a cue as a reward although I know it is done all the time doing agility with start lines, weave poles, and stopped contacts. Times that come to my mind are Lance's rollover in the obedience ring and also occasionally used during shaping sessions. In obedience the prime example is using a finish and a reward for a front but I rarely do it. I almost never finish my dogs after a front and instead reward in position and break them out happily. But perhaps I should be utilizing the concept more.
Dawn, assuming you trained the retrieve positively, yes, I think you've got it! Do you find that works? That's the ultimate test of whether something is reinforcing or not- if the behavior increases.
Laura- I have also wanted to shape a behavior using cues as the reinforcer, but that would require me to actually have things on cue! LOL
Cues as reinforcers: People inadvertently use cues as reinforcers all the time. We've all seen this situation in classes: Dog pesters handler. Handler cues down, clicks and treats. First opportunity dog pesters handler again. Rinse and repeat. Does the dog learn to stay down and leave the handler alone? No, because the "down" cue is reinforcing the unwanted pestering behavior.
Getting stuff on cue: Your dog can learn the concept of a cue -- that once a behavior is learned, the cue will become part of it. And once they get that bit, they learn cues insanely fast. I once taught Finn a cue in 12 clicks. An hour and a half later I took him to class and demonstrated it. Not only did he wait for the cue, he was also able to recognize it when I interspersed it with sit and down. (This is back in the day before Finn quit school.)
I love love love cues and cueing. There is so much power there. Don't even get me started . . . !
Oh, Jane. Would you take Maisy for a couple days??
Just wanted to say I am loving these posts! Thank you so much for sharing all the info and writing them.
This sounds like it was so awesome - I wish I could have been there! I use 'touch' as a reinforcer around scary things (to get Lola to move closer to something - she approaches, I say 'good! touch!', and she bops my hand for a treat), and I use fetch as a reinforcer for miscellaneous tricks when we're out and about, despite the fact that's a cued, learned behaviour. :)
I love the idea of cues as a green traffic light!
I'm not so great at getting things on cue either. Getting better with practice, with Dragon. I'm always interested in hearing tips about that!
Dragon's cue to start searching during nosework practice reinforces whatever he was doing before -- usually waiting at the door as I open it.
Crate games also use cues in a behavior chain as reinforcers. If we're playing and I ask him to go into his crate, he does it happily because he knows that he will then be asked to exit it again and play will resume.
I don't usually consciously think the CUE as a reinforcer, but I do use lots of life rewards and premack and so I'm thinking of those activities as reinforcers.
Actually, now that I think about it, there are a few behavior chains I'm consciously building (like recall, sit; drop, hand target; mat, down), in which case the cue for the second behavior is R+.
More tips are coming tomorrow. :)
Oh, and Joanna- I don't usually think about cues as reinforcers either. Like you, I do a TON of premack, but some of those activities (go sniff, for example) are on cue. The primary reinforcer is the sniffing, of course, and the cue is PERMISSION to go sniff. As a former teenager, I remember how exciting it was to get permission to do something.
Hmm, I guess I use Fetch as a reinforcer for Lassie... if she Sits like I ask, then she gets to chase the ball! But no, I hadn't been using it the way you describe. I think I am learning all sorts of neat things just from your summaries, so thank you very very much for sharing!
I use VERY few corrections, not so much because I won't ever be negative or whatnot, but more because these dogs don't generally do much that needs correction. Well, except Missy and her never-ending mission to kill Lassie, but even that is much easier dealt with in a positive way.
I've been actively working on making one of my dog's cues a reinforcer: her cue to jump into my arms.
I've been using it to build enthusiasm for interacting with me while off-leash, and as a reward for particularly good focus. It's become obvious that what was previously kind of a precarious "I'll do it sometimes but not all the time..." cue has become something she truly enjoys -- she'll launch herself at my arms without much thought of whether I'll be able to catch her or not. (Most of the time I catch her, I swear.)
As has been mentioned, the act of training and interacting with the dog is also a reinforcer, as well as the cue predicting the marker predicting the reward. My girl, being as energetic as she is, is thrilled sometimes just to get my attention to cure a bit of boredom.
Crystal, I will be happy to take Maisy for a couple of days, although you'll need to take Finn. I'll be shipping Tadgh to Robin to give her more teenage dog blog material. . .
I've been thinking about this post for a bit. Lets see if any of my thoughts make any sense. I like the idea of cue as green light. Jun will occasionally refuse to do a behavior (generally a sit or down) and in those circumstances I've had no idea what to do. She will NOT be physically forced into a position and mild physical punishment (e.g., collar pops) has no effect except to make her less likely to want to do it, and more resistant. I've been experimenting with negative punishment (removing my attention, stopping the training session). No solid conclusion regarding whether that is working yet. Sounds like greater reinforcement history with the behavior may be in order. However, these are the first two cues she learned and she has been reinforced for doing them a LOT. So I'm still not sure what the answer is. She seems to resist most often when I'm cuing a sit or down as a set-up move for some other behavior, like a sit in front when we're working on finishes. I think it's just her way of expressing that she would prefer to dispense with the "unnecessary" (in her mind) re-setting. Jun is nothing if not efficient.
And on cue as reinforcer--this is a mind-boggling concept for me, and may be one of those things I won't understand unless I see it in action. It is not really the cue itself that is reinforcing is it? Isn't it the "hope" that the cue will eventually lead to a primary reinforcer that is actually reinforcing? So a click is a secondary reinforcer, but if you stopped treating after the click on a regular basis it would lose it's predictive value, and thus lose its reinforcement value. Right? Maybe I don't truly understand secondary and tertiary reinforcers.
A few people have also commented on particular cues that they use as reinforcers--like Maisy's sniffing, Lance's roll over, or CiDa's dog's jumping into her arms. I use Elo's jumping into my arms as a reinforcer ALL the time. But I don't think it's the cue he's finding reinforcing. Jumping into my arms and being cuddled (and also being higher up with a better vantage point) is something he inherently LIKES. I am seeing a distinction here between a CUE being a reinforcer and a BEHAVIOR that happens to be on cue as a reinforcer. To me, the examples above are cues that give the dog permission to do a behavior that is a primary reinforcer because the dog finds the behavior itself fun. And there may be a further distinction to be made between cues that are inherently fun for a dog (like Elo's jumping into my arms) and cues a dog has been taught and subsequently comes to think of as fun and will do for fun even though it has not been reinforced with a primary reinforcer in a very long time(like Lok's roll over or Jun's nose touch, or certain agility obstacles). I can see how both of these categories of "cues" can be used as reinforcers, but it seems to me it is the behavior itself acting as a primary reinforcer and the cue is just the "on switch". Can a behavior be reinforced with a cue that a dog does not particularly enjoy performing?
I think there are some behaviors that are inherently fun and rewarding for our dogs, but I have also found that our attitutde as far as whether it is fun to do predicts whther our dogs experience it as fun to do. In my mind, I want my dog to think of it ALL as tricks. And to that end, I try to behave as positively to a down on recall as I do a "Bang, your dead" and roll over trick.
That's a good point, Elizabeth. I certainly do my best to make ALL training fun for my dogs. However, there are still things they like doing more than others. I think all dogs have their "favorite" tricks and others that they really only do out of a desire for a reward.
But the main question remains . . . is it the cue or the behavior that is the reinforcer? And can a cue a dog is not all that jazzed about be a reinforcer?
Ninso- You articulate some of the same thoughts I've been having as I've read the comments. Unfortunately, I just started a new job this week, so I have neither the time nor brain power to fully respond to them. (Well, that and this is new information for me, too.)
That said, I have some neonatal thoughts I'd love to bounce off you. PLEASE give me some feedback. :)
I think that cues can work as reinforcers in one of two ways.
1. As you and a number of other people note, it's not so much the cue that's reinforcing as the behavior it allows. For example, sniffing is inherently reinforcing for most dogs. I use a "go sniff" cue with Maisy. My cue gives her permission to do something she wants to do. I think the cue IS reinforcing, but only in the sense that a click or marker word is reinforcing- because it signals "awesome time now." Plus, I think getting permission to do something is reinforcing. Being given permission to borrow the car when you're 16 is pretty darn exciting after all. Does that make sense?
2. For behaviors that have no real value- that aren't reinforcing in and of themselves- the situation is a bit more abstract. In that case, assuming positive training only, the behavior itself is associated with a positive outcome. I sit, I get a treat. Therefore, sitting is good. The cue is therefore not permission, as in our first scenario, but rather an OPPORTUNITY. I think opportunities can also be reinforcing. When I was recently job hunting, getting calls for an interview was pretty exciting, and reinforced my behavior of filling out applications EVEN THOUGH I hadn't yet received the reinforcer I wanted- a job. In the same sense, getting a cue is an opportunity to get a treat. Does that work, or am I stretching? I feel like I'm close but I can't tell if I'm describing what I'm thinking very well...
On #1, I agree with you. On #2, I am gonna have to think about that a bit more, but my initial thought comes back to the fact that we don't get to choose what's reinforcing for our dogs. Maybe a cue CAN be a reinforcer, just like hot dogs CAN be a reinforcer, and getting sprayed in the face with water CAN be a reinforcer. If it increases the behavior, it's a reinforcer. If not, it's not. So maybe some to some dogs, some cues are reinforcing and some are not and the WHY behind it doesn't matter as much (though not knowing the why drives me insane) as getting the behavior you want.
I don't know, that was my first thought. If any other epiphanies come to me, I will let you know.
I'm still mulling it over myself. But I agree that we don't get to pick the reinforcer, and that suggests that not all dogs will find cues reinforcing. Kathy said that can happen if there was punishment mixed in, but I bet there's more to it than that. What if it was trained in an uncomfortable environment (busy class or cold room or something), or if the food reinforcement wasn't that great, or if the trainer was frustrated or even sick while training? Or maybe it hurts to jump, and no amount of steak will make up for it. Just throwing ideas out there- no idea if they're right or not.
I think the conclusion I'm coming to is "yes, cues can act as reinforcers, but they don't always for every dog, and whether or not they do is dependent on how the behavior was trained... and maybe other stuff, too."
Post a Comment