Clicker Training Discussion Thread: Capturing, Shaping, Cue's and Stimulus Control

Via request I've created a separate thread for the discussion of some of the finer points of clicker training.

Specifically the nuts and bolts of getting behaviour via capturing and shaping, getting behaviours on cues, transferring known behaviours which are on cue to new cues, getting behaviours under stimulus control for the purposes of sport competition/ service dog work/ working dogs/etc.

If you are unsure of what clicker training is check out the Sticky thread at the top of this section labeled: Clicker Training 101 to get all of your basic question's answered. That said this thread is for everyone, so please feel free to join in

In another thread we were discussing the changes/ stages dog's seem to go through in terms of "awareness" (I use the word loosely) of the clicker. For easy reading purpose's I've bolded each stage and then described my particular observations of each.

Everything below is simply my 0.02.


STAGE ONE

The first stage of "awareness" by the learner appears to happen fairly early when pairing the click noise emitted by the clicker with a primary reinforcer. This is where you will see dogs go from basically ignoring the clicker to snapping to attention when the click noise is emitted.

Primary reinforcer's usually refer to these four basic needs of social animals: Food, water, play and sex. For non-social animal's I would argue that play would be an irrelevant reinforcer.

The biggest thing to note about any reinforcer is that the learner decides what is reinforcing to him/her.


STAGE TWO

The second stage occurs when the learner further discovers that not only does click=goodies, but that the learner's behaviour can control the reinforcement flow.

This moment of "awareness" is commonly referred to as the "light bulb" moment. Some animal's appear to become visibly excited at this moment.

Side Note It is important to note that this second stage of awareness that I am describing can only be reached by allowing the animal to become an active participant in the training without luring, or physical restraint/ manipulation. A dog that is lured to sit, will still do so with enthusiasm, and still be very reliable- however it will never exhibit the same deliberate movement that a dog who has been taught to sit by capturing in the breakthrough "Ah ha if I do this I can make her click".

After you reach this second stage of awareness the whole process becomes much faster. Your dog will become more and more proficient at acquiring new behaviors during shaping session's and learn new behavior’s taught via capturing much quicker than the first three or four. This is where the "accelerated learning" really becomes noticeable.

STAGE THREE

Once your dog get's a few of these new learned behaviour on cue the third stage of awareness is learning that the cue is the only green light, when performed in conjuncture with x action that will earn reinforcement. This is the goal of most trainer's: to get their dog's important known behaviours (the ones which get polished to perfection for sport, or work, etc) under stimulus control.

To be under stimulus control an on cue behaviour must meet these four conditions:

-The behaviour is performed when cued
-The behaviour is not performed in absence of the cue
-The behaviour is never performed in absence of the cue
-Only the behaviour which is cued will be performed (Example dog doesn't lie down when cued to sit)

While practically I do not think every single behaviour a dog learns should be put on cue, you can certainly see how it would be helpful in a variety of situations. Specifically, behaviours which are under stimulus control tend to reduce the frantic "behaviour flinging" which some clicker trained dogs tend to do in shaping sessions. Again to each there own, personally It doesn't bother me in a shaping session because I feel it gives me a wider variety of behaviours to work with and very quickly you can narrow it down to only clicking for what you want, and thanks to extinction law (stating that unreinforced behaviours will eventually disappear-usually right after spiking) everything else should very quickly disappear.

Anyways that's all my fingers can manage for now, please add thoughts, questions, experinces, discuss!

Kayla

Why would you want the behavior to never be performed in the absence of a cue?

What behaviors do you think don't need to be put on cue?

And you didn't tell us how to teach the cue.

Nice post, Kayla. Thank you.

I was a clicker trainer for several years before training assistance dogs, and to be honest, I always thought stimulus control was kind of silly and not relevent to most behaviors. :lol-sign:

In training assistance dogs, I can tell you that stim control is THE most important the dogs learn. It's also the hardest thing to teach, the most time consuming to teach, and IMO a pain in the neck to bother with. But relevant, it is.

To start - teaching the cue: Once the dog has learned that his specific behavior is the behavior that is being clicked/rewarded (Kayla's Stage Two), then you simply name that behavior. The dog should be repeating the behavior fairly rapidly, and it should be apparent to the trainer when the dog is about to perform the behavior. When you can see that he's about to do the behavior, say the cue; then c/t when he does it. You should say the cue with the right timing relative to the behavior, so that to someone watching, it looks like your cue causes the dog to do the behavior; though actually, of course, you know that the dog would do the behavior whether you gave the cue or not.

Once he knows the cue, then you can start teaching stimulus control. Kayla's four conditions of stim control are, of course, correct (though I would challenge that #2 and #3 are redundant). But not every behavior NEEDS to meet all four conditions, in the real world; in fact, in the real world, no behaviors need to be on stim control, it just depends on what your goals are for your dog. For example, for sits and downs, I don't mind if the dog does it without being cued; BUT, when I cue a sit I don't want him to down; and when I want the dog to come I don't want him to sit away from me.

There are a LOT of behaviors that I don't want the dog to do without being cued, with the dogs I train. Retrieve, opening doors (including refridgerator doors), turning lights on/off (my org had a service dog who would spontaniously turn on lights in the middle of the night, then "forget" how to turn them off ), hearing dogs alerting to a "sound" when the sound is not going off. Search and rescue dogs should have stimulus control with their "alert" behavior so that they don't do it in the absense of the cue (which would be whatever they're searching for). Police dogs should not chase and grab someone without the cue.

Personally, I don't want my dog to "speak" without the cue. She also does this cute thing that her previous owner taught her, called "clap" - she stands up on her back feet and "claps" her front feet together; it's cute, but the only thing she knew when I got her so it got a little annoying, I would LOVE it if she only did it when cued.

Stim control is very difficult for dogs to learn, because it takes a lot of concentration. With my dogs at work, almost all of our cues are verbal, which of course are harder for dogs to differentiate than physical cues, so they have to be very careful to listen to what we are saying. When we are teaching the dogs to distinguish between two or more cues, it is very obvious that the dogs get tired quickly, so we have to keep training sessions very short.


Very interesting discussion, thanks Kayla!

I now feel very stupid. I could have thought of those, but apparently my brain decided to go to sleep long before my body.

I have behaviors that I definitely only want on cue but lots of behaviors that I want the dogs to do where I don't have to tell them. Of course there really is a cue. (contextual) But I want it on "default" mode. In other words, I don't want to have to tell my dogs wait while I open the door before they go out.

What do you normally do when she does this cute trick spontaneously? I bet if for a while, you beefed up the reinforcement for when she does it only on cue (like give her a tasty treat and a fuss) and turned away and ignored her when she did it spontaneouly, those which are spontaneous would fade away and she'd learn to wait for a cue. But at first, I'd be quick to get the cue in there if she looks like she's thinking about doing it. Beat her to the punch. LOL.

Clicker training is indeed interesting. My dogs have back chained. Not only do they perk up when they hear the click, but if I so much as pick up my clicker, it's all click, click, click of little nails on the floor. I suppose I could pick it up and then put it away, but I don't have the heart. LOL. The same thing happens when I pick up the Dremel. My dogs are nuts.

Yay discussion on

Great question's CP I'll address them individually.

Lizzy posted some excellent examples, one other I can think of is putting some annoying behaviour like digging on cue, create a sand box outback for the dog to get it's daily outlet, but never without the cue.


I generally don't put anything that pops up during 101 things to do with a box on cue. I just free shape and select new things as they appear. Duke is still a novice to the game and often feels like offering default behaviours in which he remains for quite sometime so there is an example of where on cue behaviours can slow down a free shaping session. To combat this I usually wait for him to get out of position and then jack pot him and begin shaping him for anything, eye contact with something, walking around, whatever.

I've also shaped duke to file his nails on a board with heavy duty sand paper attached. I haven't put this behaviour on verbal cue because I really don't see a need, the second the board goes down Duke runs over and begins to scratch it.

Basically if I don't need a behaviour for some purpose, I usually won't put it on cue.

It seems my brain went to sleep before my body as well. Lizzy posted an excellent description. Basically when I'm willing to bet ten bucks that my dog is about to repeat the behaviour I have been developing (usually by shaping as I'll admit I do not currently have the patience for capturing) I will say my chosen cue. After repeating this over a few sessions I begin saying the cue earlier and earlier, then once I can say the cue, and the dog will perform the behaviour I will start alternating between letting uncued response's go un reinforced and then immediately cueing the behaviour and jackpotting.

Like Carrie mentioned, the uncued response's seem to vanish rapidly after alternative reinforcement for cued responses is on the table.

I also just wanted to talk about, how crossover dogs (meaning dogs with previous training history, generally where, at some point corrections have been used) tend to react to clicker training vs a dog with absolutly no previous correction based training history.

When I first enrolled in the Karen Pryor Academy program I had done a bit of clickering with Duke and figured it would be pretty easy, after all we had gone through three levels of reward/correction based OB classes which he excelled in.

It was a real shock to me then, to discover that when I took away luring, and correcting, Duke during shaping sessions would frequently begin to show signs of frusteration, via barking before shutting down (which to me is showing outward signs of stress such as panting, calming signals and a sluggishness to respond) if he did not immediately recieve a click for something we were working on.

At my first workshop I believe we had four or five behaviour to shape/capture and put on cue. The academy strongly pushes students to be able to identify where they are in the training process so, though a little bit dissapointed I could identify that Duke had really only just reached a stage two awareness and only one of my five behaviours was on cue.

We also had to bring detailed training notes and show how we worked on a typical session. At this point I knew better then to try and shape Duke for more then one or two miniutes tops because it just sent him over the top if I tried to raise my criteria too rapidly where old criteria was no longer being clicked and thus into panic mode he'd go. I explained this and showed typically how we would work. Next it was my classmates turn. My oh my did I feel stupid-my classmate's dog seemed to LOVE shaping and would be wriggling with joy the entire time. In one session both of my classmates could raise there criteria enormously because when there dogs did not recieve a click/ treat instead of spirling into frusteration they'd simply offer new behaviour until something worked.

Over the next six month's Duke began to recover from not recieving a click and he also learned to offer new behaviour when something wasn't working. However looking back I can identify that his previous training history played a large roll in his initial inability to work past failure.

This of course again is purely my opinion however I would like to present the idea that even elegant use of corrections (minimal and only after the dog understands the behaviour 100% and some distractions and duration have been layered in) does negatively condition offering new behaviours during a training session. To such a dog while the first behaviour he was been reinforced for was safe, lack of reinforcement does seem to throw them into a anxiety based state which to me makes sense because a correction trained dog has learned that doing anything other then the "safe behaviour" usually leads to a correction.

There is a lot of supporting scientific evidence which is currently being conducted by Dr. Jesus Rosalez Ruiz at the University of Texas about poisoned cues. Over thousand's of trials dogs were taught the same behaviour, using two different methods and both were put on seperate cues.

The first behaviour was free-shaped, when the dog did not perform the correct behaviour, nothing happened, and if necessary the criteria was temporarily lowered to help the dog succeed before rapidly bringing it back up to par.

The second was free-shaped to the point of being put on cue however after on cue, if and when a dog offered an incorrect response a correction was employed.

The result's were rather phenominal, while cue 1 was responded to with a wagging tail and promptness, cue 2 produced a sluggish dog. Furthermore for an additional thousand trials cue 2 no longer employed corrections and yet the response remained the same.

To me this furthur shows that while rewarding what you want and correcting incorrect response's make's perfect sense to us; dog's, which are excellent discriminators may not make the proper connection. Instead of learning that breaking a sit/stay, or not coming when called will produce an aversive event it seems more likely that they tend to learn that the entire cue itself is not always safe and thus the sluggish/ response.

As always this is just my 0.02.

Kayla

Well, in the four years that I've had her, I've NEVER rewarded her for the clap with a treat.... I've been hoping that one of these days it'll extinguish, and I think it has definately lessened. Occasionally, I do cue it (only when I'm showing off to other people) but even then her only reward is me saying "good girl," not even very enthusiastically; she's extremely treat motivated, and NOT very praise motivated, so it's really almost like no reward at all.

Yes, I think your advice would work.... but after 4 years of her doing it, honestly, I'm just not at all motivated to work on it.

Since most of the dogs I work with come from shelters, we don't know anything about their past training history. But we do sometimes get dogs who have a VERY hard time with shaping/capturing, it tends to stress them out. Another trainer I work with had this problem with a new dog we just started training, she would wait for the dog to offer a sit or down (can't remember which one she was working on in this session, but of course she had only one), and the dog would just pace (it was a very small area - though not too small - and not distracting at all), whine, yawn, lick lips, jump on the trainer, etc. It seemed clear to us that she was waiting for a "command," and the fact that it seemed that we weren't telling her what to do was very stressful.

I honestly don't know what this trainer is doing, so unfortunately I can't report on how the dog is after the past couple of weeks of training; though I do at least know that the trainer is pleased with her progress. In the past, when I've had dogs like this, I do "play" exercises with them, where I reward basically anything they do. We have yards where we exercise the dogs off leash, and we put various obsticles in the yards for the dogs to climb up, crawl through, etc., and I would reward the dogs for any interraction with any of the obsticles. One of my favorites, I had about 1 1/2 years ago, a little boston terrier who had a difficult time offering up behaviors in training sessions. After just a couple of weeks of training, where he learned probably a sit, down, and the beginning of loose leash walking (plus how to offer behaviors on the obstacles in the yards), I played "101 things to do with a box" with all of my dogs, and this boston was by far the best at the game.... even better than Luna!

But I think the key is to kind of earn the dogs' trust.... show them that they can do stuff and get rewarded for it; it builds their confidence for training. NEVER correct a dog like this, until you can get him readily offering behaviors, and even then you have to be very gentle with corrections for a while until he understands what corrections are too (and by "correction," I mean any kind of positive or negative punishment, no matter how mild). Otherwise you will undo any progress you've made.

I had another dog, a golden retriever who came to us from an organization, so that he already had all of his basic obedience, and we just had to teach him his service dog behaviors. I believe that the organization where he came from did not do any positive punishment, but I also believe that they relied heavily on luring rather than shaping/capturing. This dog, though, was extremely sensitive.... if he got any idea that what he was doing was wrong, he would completely shut down. For the service dog behaviors, we teach most of them with shaping rather than capturing, so I ended up not capturing a lot of behaviors with him. And when it came to shaping, I had to do a lot of microshaping, to help him keep going and keep his rate of reinforcement high. He was VERY frustrating to train, it took him much longer to learn because we couldn't take ANY shaping "shortcuts."

I skirted this question before because it was late and I didn't want to think about it, but I would like to bring it up again to hear others' thoughts.

Basically, it's hard to say that a behavior is not on cue, because there are so many cues to dogs. When we go outside, I want my dog to sit at the door without "being cued," although my behavior of going to the door and standing is in itself a cue. I want my dogs to automatically sit - without a cue - when I stop when we're loose leash walking, but actually me stopping IS a cue. I'm sure there are dozens of these examples.

One behavior that I can think of that I want my service dogs to do without a cue is to make eye contact with me. I will almost always reward spontanious eye contact, because it is so important that their attention is on me at all times.

Thanks Kayla. Great info!!

What about a crossover dog who has always been lured trying to learn capturing and shaping? The dog is used to always being shown/told what to do and shows considerable stress at not being told to perform something.

This is not purely your opinion. Your opinion is shared by many, myself included. This is something I've read about for some time and noticed myself. This is why I do not subscribe to "balanced" training. You know....a collar yank here, an electric shock there and some treats and praise thrown in for good measure to balance things out. It totally inhibits so much potential to learn. This is the basis for why I get in debates about training using force or strong aversives. They're going to have aversives anyhow....just the disappointment of not being reinforced for something or having something taken away that they formerly had...the chance to have the door opened is wooshed away for a moment until they simmer down and wait. Weah, weah. Then there's the dog that barks demandingly for you to throw the ball. No ball if you can't stop yapping at me. Weah, weah, the dog sobs his eyes out. It's the way it is. (the "real" world. lol) Some people on forums sometimes get up in arms and say they're not abusing their dogs by giving a mild collar correction so get off your high horse and get in the "real world." LOL. But that is not the point.

I understand that you trained her to do this trick with no treats. Heck....I use to train all my dogs with no treats. They can learn all kinds of things without treats. But what I'm suggesting is the idea that you can really drive home the contrast between doing the trick on cue as opposed to doing the trick whenever she feels like it. The high value treat all of a sudden (where it wasn't before) being paired with the cued response and then the complete ignoring of the spontaneously given behavior may put the two things at further ends of the spectrum for the dog. Lots of practice with cuing the behavior and lots of instances (if they come up) where you ignore the un-cued behavior will raise the odds that she'll increase the cued trick. It probably would only take a couple of days or even less of intermittent practice sessions of just a minute or so each. I've undone obnoxious behaviors to the hilt. Let me tell you. LOL. I let so many little things slide by and before I know it, OMG!:yikes: My dogs are turning into monsters. Then I say to myself, "Self....Helloooooooo. You are suppose to be a dog trainer, not a brat maker." So, I buckle down and spruce things up a little bit. But I have to say, a lot of it is my son's fault. :lol-sign:

That was a great post Kayla...great information. The whole thread is a terrific idea.

[/B] LOL I am sooo guilty of this, but at the end of the day sometimes training your own dog is the last thing you feel like doing, none the less eventually we all buckle down and get it done.

I think such a dog may show stress initially, after all capturing and shaping really put a dog/animal in the driver's seat where as luring only really requires a dog to be a passenger.

I think this is the most important thing to note about clicker training, it does require a lot of mental work for the animal and as such you should work at your dog's pace.

Karen Pryor has ten formal shaping guidelines which are all very useful for getting started however I've found these four have been the most relevant for me while working with Duke.

1) Keep learning sessions short

This is hard because when your dog really starts to "get it", it's so reinforcing for us as trainer's that we just want to continue on. I think for novice dogs regardless of their training history, 30 seconds to two miniutes is an ideal time to push for. Even if you just get ten-twenty click/treats in each session very quickly you will build up a deep reinformcent history for stress free training sessions.

2) Raise your criteria quickly but in very very small steps.

It took me tons of mistakes to get this one figured out. I would get stuck at one criteria and just reinforce, and reinforce and reinforce without moving on (mainly because It was such an aversive event for me to see Duke get so worked up) and it's almost like digging yourself into a reinforcement ditch, the deeper you go the harder it is to get out and move on

3) End while your dog is still in the game

Again it's so hard to do when you have a huge break through but it's better to end before you push your dog to the point where training becomes frusterating. 30 second sessions may seem silly, but for a dog new to this type of mental puzzle, it's more then enough. You will still get in 3-5 click/treat's in such a session and because they are so short you can get in more each day. Once your dog developes more mental stamina you can increase your training time.

4) When behaviour breaks down, go back to basics
This goes against what anyone who uses, or is used to using corrections will typically do when a previously known behaviour deteriorates but it's a good idea to review your starting point.

It is generally agreed upon, regardless of the type of training method you use, that most behaviour go through a degression when they are close to be learned. Instead of applying an aversive which could furthur inhibit a novice dog, simply review your shaping plan rapidly. For a novice/shy/crossover dog this can do wonders.

Two days before my final assessment at the Karen Pryor Academy, my ten part behaviour chain which we were required to train and present fell apart because of one cue. While most of the behaviours I used were well established and meeting at least three conditions of stimulus control. A few had just been shaped a month prior and were just beginning to meet one or two conditions of stimulus control. Among these included Duke's informal retrieve. My criteria for it invloved Duke picking up the object, removing it from a box and then moving towards me with it- I did not care if he delivered it straight to my hand.

I was devistated when he seemingly "forgot" the cue so I spent 48 doing very short sessions rapidly moving through my previous criteria, putting it on a new cue, teaching him to wait for the cue, etc.

On assessment day my chain worked out but it taught me a valuable lesson on how effectively reviewing the basics works- all without any stress being applied to the dog.

If the corrections I've used have produced a sluggish dog, then I wouldn't be able to physically keep up with him without those corrections. Tyr is lightning fast and his tail wags fiercely even while receiving a collar pop.

My idea of "balanced" is a treat here, a tug toy there, lots of verbal praise and petting and an occasional gentle collar pop as an interrupter instead of a punisher. Based on the responses from my dogs, that collar pop is less aversive to them than withholding or removing reinforcers.

Well yeah....a gentle tug on the collar is nothing more than like when you bump into them by accident or turn around and clip them by accident with a knee or something. It's not punishing so it won't tend to stop a behavior. Get their attention? Yeah sure. But if they learn some verbal cue like, "watch me" or "hey, pay attention," lol....they won't need a physical cue. (not that I never put any pressure on my dogs) Even when I don't try to tug on a leash, it happens. Here I am with one big dog on my left and two little tramps on my right way-y-y-y-y down by my ankles. We stop while one pees and the other one goes over there to sniff, then before I know it, they've got me tied up and bound. As I manuver to untangle myself from the web, the leashes are all getting yanked a little bit and pulled. "Okay...you go this way. And you stay right there while I loop this leash through that one, now hold still while I step over your leash...." (okay...well...a slight exaggeration) :rofl1:

Yup. But there are people (not necessarily on this board) who think that all collar pops are cruel. It's a matter of repetition and high rates of reinforcement for what I want them to do and interrupting unwanted behaviors to develop into habitual patterns. The most punishing thing I can do is simply be displeased. A frown is punishment to my dogs. A smile is reinforcing to them.

Depends on your def of a 'collar pop'... Kaiden found even fairly gentle ones aversive (our lagging to worse the more he got popped.. and he was on a flat collar) I have seen some neck wrenching 'pops' at the obed club-and those dogs were in chokes or prongs.

Well that wouldn't be a gentle pop. And sure, for some dogs even a gentle one is too aversive. For some, removing something they want is too aversive. For some, a loud "no" is too aversive. Mine ignore verbal corrections, mine get extremely stressed and will shut down by the removal and withholding of things. Maybe it's from 15 years of riding horses - I use the leash and collar with the mindset that I used the reins and bit (except that reins weren't slack as a loose leash is) - it's communication.

I also never use collar pops, no matter how gentle, for lagging. In fact, there are a lot of things I would never use a collar pop ~ or any other aversive ~ for. If my dogs lag, it most likely means that my left shoulder is turned towards them, pushing them back. Otherwise it means they're tired. Except Morgan, who I believe developed a negative association with heeling when her hips or knees might've been sore and I didn't realize it. In other dogs lagging could be a matter of they don't know where they should be or they're under motivated to keep up.

I think you point out an excellent point CP. Much like reinforcer's a dog get's to decide what is punishing to the individual. In many cases and especially in high drive dogs like Tyr the reinforcement he is recieving from working with you may trump any mild corrections you are handing out.

Another thing to point out about the use of aversive's in training is habitation where after awhile most living organisims will adapt and no longer find the same aversive which originally stopped undesired behaviours in their tracks as punishing and therefore a harsher and harsher aversive begins to be needed to achieve the same results.

I'm certainly not saying this is always the case, I know some very very talented "balanced" trainers who have achieved many titles on their dogs, so my base argument here is not that aversives in training do not work. There are thousands and thousands of dog/handler teams who have proved otherwise.

What I am stating is that their is now quite a big amount of evidence from the studies being conducted at the U of Texas along with sport competitiors in almost every field as well as working dogs from police to SAR to service dogs who have achieved the exact same results, without aversives and generally produce dogs who do not exhbit the same inhibitation during shaping sessions that aversive trained dogs exhbit.

Great discussion guys

Kayla

lol YOU never said gentle. And some of these people wouldn't say they are being harsh. To them its 'normal' .. the trainer/handlers have become habituated to punishment.. its now a predominatly e collar club now. (showing the progression of aversives quite nicely)

I have often said its what the dog finds good/bad is the issue not what we people think they SHOULD find good/bad.... Sport LOVES to be sprayed in the face with water

True...if you're trying to stop behaviors in their tracks.

Things like withholding reinforcers, turning your back, walking away, etc. don't do that, nor do the collar pops in the way that I use them.


Yes, that's very true. With Tyr, if I walk away from him and ignore him as an aversive, it would be abuse. And yes, being with me is extremely rewarding for him.

Using no aversives at all??

uh...

I was replying to