The Limits of Clicker Training

ORIGINAL: Awsomedog

ORIGINAL: gunny1

We are different people thier A-dog. And you do understand that you can use +R, -R, +P and -P all together.

Yep, I'm just intested now in what that *clicker* crowd thinks about that. This should really be interesting.

All of the above are part of operant conditioning.  The so called +R people simply choose to base their training more heavily on the +R, -P, -R quadrants and way less on the +P.  Our philosophies exist on a continuum, like any other training method.  For example, you might use the occasional leash pop, but might elect not to use the old Koehler "hanging" method.  So, both of you use aversives, but one is less harmful than the other perhaps.  With +R trainers, there are those who might use the occasional aversive, and those who never do so, but overall the method is designed to be as dog-friendly as possible.  Positive doesn'e mean permissive, even for those who don't use aversives, and many of them (notably, people like Angelica Steinker and Pat Miller) are very skilled at communicating what is required to the dog without having to resort to those.
A discussion of someone else's methods is not grounds for the rest of us to be "converted".  gunny chooses how to train his dogs, and I choose how to train mine.  Sometimes, I learn interesting things from him, and I'm sure he does from me.  We aren't disrespectful of one another, and I can tell him straight to his face if I disagree with something he has said, because I don't attack him - I simply say why I agree or disagree with the point at hand.  No "can't wait to hear that one" or "bahahahahah's".  Just discussion.  To be honest, I appreciate that gunny has the guts to come to the clicker threads and say how he is using the device to his advantage.  Maybe he will hear something, or take note of a link, that will help him out.  Maybe he will drop a name, or mention a tape or DVD that will help me.  But, I do know that he won't call me "rude" or pick my posts apart line by line, nor will I do that to him.  That's why he's a respected member here on these threads, regardless of his philosophy.

Gunny, thanks for explaining that to me in detail.
 
It is certainly beyond anything I've tried or am familiar with, as my dog only has titles as CH Foot Warmer and CH Running Around Like Crazy. []
 
But I can see what you're describing and it's applications for a really drivey sport like schutzhund.  Certainly with the 'stick' you're describing, you've basically got 2 options - make the dog get excited about it, or have them want to avoid it and be fearful.

Gunny thanks again for your posts.  Especially explaining about the e-collar and leash pops in more detail. 
 
I would be interested to learn more about how it would be possible to "build drive" with a clicker.  You can encourage a hyper dog to be calm by "clicking" for calm behaviours so that they are offered more often and by offering the calm behaviours frequently the dog actually calms down.  So I would assume in theory that it is possible to "click" (Read: mark) for intense, "drivey" behaviour as well..... ????
 
I think a few people get confused as to the meaning of operant conditioning..... it's not something some animals learn better than others, it's not something that "won't work" on certain animals.  It's the science behind how all animals learn.  Every trainer uses it.  I prefer to go easy on the P+ and use a combo of the others. Perhaps, in a totally perfect world, with all rainbows and kittens and suchlike it would be possible not to use P+ at all, but we humans are not perfect and we do not live in a perfect world.  When I have to resort to P+, I usually see it as a failure on my part - in hindsight it could have been avoided if I had managed the environment better, if I had caught the dog earlier and redirected into a  good behaviour, if I had done such and such....  So what I am saying there is that a definite "limit" of marker training is the person using it.....

Ok gunny and others,,, I have been thinking over your collar post.  It appears to me the collar becomes an SD (discriminative stimulus).  The strategies you described all appear to be using a positive reinforcement approach because behavior is not disappearing.  The play, put away etc.  all result in an increase in behavior.   By the time you start using pops, I think there is a good possibility that the collar is not an aversive.  Many folks (previously myself) felt that prong collars were in essence a negative training tool (negative reinforcement since the dog worked ot avoid the pressure or punishment because it stopped a behavior).  Neither of those scenarios actually apply in the way I read your post.  I think we are talking SD.... Collar comes out, good stuff happens, so it is more anticipatory in nature, rather than a consequence as a reinforcer.

So what I am saying there is that a definite "limit" of marker training is the person using it.....

Actually, I have worked well in a no aversives set of conditions with a dog once. I got a lot done and made a seriously reactive/ aggressive dog into a dog that could go on a walk, walk by other dogs and small children on a leash, and play with other dogs under certain conditions.

The limitation for that dog was biting, and that was her limitation, not mine. Her bite-to-fight ratio was high--if she fought, she would generally maim, and as she got older this tendency got worse and not better. I was able to decrease the power of her triggers as I found them, but new ones popped up--dogs became okay and children became a threat. People coming over to dinner was okay, except certain people, and parties were definitely NOT okay and people coming over became more of a problem.

Training that dog was like playing Whack-a-Mole.

Can't clicker train a dog not to bite--nobody can train a dog not to bite, not all the time, guaranteed. And you can't train (clicker or otherwise) a dog that bites really badly not to use maiming force when she bites. Not reliably. Was I able to decrease the number of times she did that on any given day, so that I was pretty sure the day would be okay? Yes. But pretty sure isn't good enough when you live in a city with a bunch of other people and you know that your dog has the potential to rip a kid's face clean off.

The trick, then, is to have a sound that can be heard for at least 100 yards, or so, such as a shepherd's whistle.

The shepherd's whistle, and I think those hunting whistles, appear to be audible by dogs at nearly a mile away. Possible more so. There's an old story about a famous trainer of sheepdogs who left one dog behind to watch a packet of sheep. walked for a while to check for strays, then spotted a packet of his neighbor's sheep had strayed into his sheep. He started directing the dog he had with him to push the sheep away over the hill to his neighbor's pastures.

When he turned around to return home, he didn't directly pass the fold where he had left the first dog, so he whistled him off those sheep. To his irritation, the dog didn't show up. "Gone away home," he thought, annoyed.

When he arrived back at the homestead, his dog was indeed there, but so was his neighbor on the other side. ""Ah've brought tha dog home," he said with a grin.

He said he had looked out the window and spotted a dog running about in his field. He started to get his rifle, but then noticed that the dog, which he recognized, seemed to be working, and that the sheep weren't his.

The dog had heard his master's commands from a mile away, then had driven the sheep almost an additional mile in the opposite direction (his whistles were backwards of the other dog to allow them to work as a team). Then the dog had left the sheep, but then looked a bit lost.

"Ah've brought tha dog, but Ah've left tha sheep for thee to get outta ma pasture." The neighbor concluded laughing.

We sometimes have an interesting discussion, since working at these distances is not exceptional for these dogs. How are they responding to our commands, considering the speed of sound? We blow a whistle and the dog appears to take the command almost immediately - but at 800 yards the delay for the actual sound to arrive should be significant (two or three seconds!).

You increase drive in a positive way by withholding the reward a little bit at a time - except in really prey driven dogs, where you can simply physically or mentally block or restrain them within sight of the goal. We use this in flyball - someone holds a dog while the owner teases the dog with the toy or simply calling.

If true work ethic could be increased through positive punishment, herding people would have discovered the method, trust me. [] Nothing beats good breeding however. Then it's just a matter of encouraging what is useful and getting rid of the trashy stuff. I can't believe it's THAT much different in the other working venues. ETA: It does occur to me that breeding for protection is still in its infancy compared to breeding sheepdogs, and breeding hunting dogs has been around ten times as long as either put together.

ORIGINAL: gunny1

A-dog, If we can a civil conversation here. I will explain. First when a dog first wears a pinch collar, or an E- collar would you agree that collar means nothing positive, nor anything negative? This is called a neutral stimulus. From here on out the way the dog views that collar is up to how it is used. Now traditionally the collars have been used to cause pain, or discomfort. So when you hear the words the pinch or E-collar. You immediatly believe the collars are used to cause pain or discomfort. What i do I assure you causes the dog no discomfort nor pain. Even the E-collar, I use an innotek on level one only. with special porongs that even further reduce the stimulus of the electrical.

I understand exactly what your saying here more than you know, and we can have a civil discussion as long as there's no insults. And I'm finding this very interesting and I think you'll soon see why.

the Stim is so mild I guarntee you would not be able to tell me when the collar was being activated and when it is not., if it were to be place on your own skin.

This part though...very questionable. You use the e-collar for correction (not as a reward, as far as I know) but I can't feel it?

The process of teaching the dogs exactly what the collars, training means is teaching them what the communication means. It goes beyond the basics of avoid the aversives and work for the good. I have found it to be much more detailed and in depth.

Trainers I have worked with and talk to who use clicker training and +R training, will do so and talk about avoiding aversives like, leash pops, pinch collars and e-collars. I'm not saying they don't correct, but just read most books on +R training by the people champion the method and you will not see, pinch or e-collars mentioned.

We working with animals that does not have verbal detailed vocabulary. I am trying to brindge the commnication gap. The click or mark teaches the dog, at that precise moment you are correct here is a reward. reiforcing the behavior. And would you agree any behavior that is reiforced is likely to occur again. The gentle pop on the pinch tells the dog that is wrong, or you need to do something else. I even teach my dogs that pops from above mean to look at me, pops to the side mean heel, pops downward me lay down.pops forward means move up or come here. I assure you the dogs do not show no aversion, pain, discomfort, or stress.

We are completely on the same page here!!! And I like you (when I'm doing OBT) use a word, not a click.

As for R+, You can use R+....and ignore the bad....Ignoring the bad is called -P. You remove the rewards as a punishment.

Correct.

As for CM's methods that is a whole another forum. And lets keep this on the workings of Operant Conditioning.

I'll have to go back and look, but I don't think I mentioned him in our debate. But sinse you brought him up...Here's why I find this so interesting. The same people who bash his methods, seem to be ok with yours. I understand what your doing and that we differ on bite and scent work, I'm just shocked by the +R crowds responce of...*gunny1 with pinch and e-collars good...Cesar helping dogs doing nothing to hurt them...bad. Odd very, very odd.

I'm glad (even though you think "I'm the worst trainer in the world") you explained more about your techiniques, which make a lot more sense to me now.

I hope you see I'm not picking anything apart, just trying to get to the guts of things. It's my way of trying to understand.

ORIGINAL: fisher6000

Can't clicker train a dog not to bite--nobody can train a dog not to bite, not all the time, guaranteed. And you can't train (clicker or otherwise) a dog that bites really badly not to use maiming force when she bites. Not reliably. Was I able to decrease the number of times she did that on any given day, so that I was pretty sure the day would be okay? Yes. But pretty sure isn't good enough when you live in a city with a bunch of other people and you know that your dog has the potential to rip a kid's face clean off.

Sorry but I've worked with dogs like that who lived out the rest of their lives and never even showed signs of aggession. So while you may not be able to completely change that in a dog, there are those who have and can.

I just discovered the block feature! How cool . . .

So I'm interested in the concept of training a dog not to do something, as fisher mentioned above (I too have been in the Whack-a-Mole situation - so apt!). Is this a limitation of clicker training? I mean, by definition isn't the purpose to increase behavior with R+? But are we talking about the actual clicker technique or the general "low aversive" philosophy?

I'm confused now. [8D]

ORIGINAL: spiritdogs


A discussion of someone else's methods is not grounds for the rest of us to be "converted".  gunny chooses how to train his dogs, and I choose how to train mine.  Sometimes, I learn interesting things from him, and I'm sure he does from me.  We aren't disrespectful of one another, and I can tell him straight to his face if I disagree with something he has said, because I don't attack him - I simply say why I agree or disagree with the point at hand.  No "can't wait to hear that one" or "bahahahahah's".  Just discussion.

 

That's all very interesting, except you yourself have been rude to people that come to this forum at times, and you love to bash CM's methods. Maybe because unlike gunny, CM's not here, which makes bashing his methods a lot easier. I think I've been pretty strait forward with gunny, and yet you and some others come screaming in about how rude I'm being. If you didn't want me to come over here and ask questions and state my opinion, you shouldn't have pm'd me and ask me to come here. What's the saying? Be carefull what you wish for. When I say IMPO that won't work...that's *my opinion* and one should be stronger in their convictions so anothers *opinion* doesn't upset one so. I believe what I believe, but even so, by reading, working with other trainers, and sometimes forums like these I still learn that someone sometimes can show me something different. The ones I learn this from are usually not so bent out of shape because of our differences. But, oh well, back to the OT.

ORIGINAL: brookcove

But are we talking about the actual clicker technique or the general "low aversive" philosophy?

I'm confused now. [8D]

 
When I started the thread, I meant the actual clicker (read: marker) technique.  But now... Who knows!!!  I am confused too.  I think in places the discussion has dissolved into the disadvantage of the clicker itself which misses the whole point of the original thread and makes it even more confusing.  But anyway.
 
The thing is that kind of training is great for teaching a dog to DO something and is only good for teaching not to do something by teaching an incompatible behaviour - which may not always be possible?  Sometimes you don't have time to allow for teaching the alternate bhvr - the dog has to learn NOT to do that thing NOW (?)  I would imagine that is relevant in working sheep?  So, I think that is definately a limit of the method, although in many situations it can still be implemented with careful management of the environment - (eg. your dog is chasing your cat.)

ORIGINAL: brookcove

I just discovered the block feature! How cool . . .

So I'm interested in the concept of training a dog not to do something, as fisher mentioned above (I too have been in the Whack-a-Mole situation - so apt!). Is this a limitation of clicker training? I mean, by definition isn't the purpose to increase behavior with R+? But are we talking about the actual clicker technique or the general "low aversive" philosophy?

I'm confused now. [8D]

 
I was just thinking about that too. Maybe the limitation of marker training is that you can't teach a dog NOT to do something, you can only teach them to do an alternate behavior. So I can't clicker train my dog strictly NOT to lunge and growl at other dogs, but I can clicker train her to look at me when she sees other dogs, thus avoiding the lunge/growl. You've given the dog a more acceptable behavior to offer by clicker training. But you haven't addressed the root of their fear/reactivity.

I think that's a key distinction: clicker technique vs. a total low-aversive philosophy. When I say I'm a clicker trainer, I just assume (and you know what happens when we do that!) that everyone understands that Clicker Training with a capital C and T is more than one technique but is an overall philsophy of dog training, management and behavioral modification.

Here's an article by Kathy Sdao about the difference between Clicker Training and "training with a clicker":http://www.clickertraining.com/node/642

Anyway, you can train a dog to not do something in a number of ways and still be within the Clicker Training paradigm, and even for some methods use a clicker. We recommend these types of things to people all the time--train an incompatible behavior (such as when I trained Marlowe to sit and watch squirrels instead of chase them, or someone trains their happy golden to sit for greeting rather than jumping up), or use -P to communicate with a dog that doing XYZ will not get the dog what he wants, so best to try something else. That latter scenario doesn't physically involve a clicker (though when the dog tries the something else and that something else is what you wanted, you can c/t) it is still within the realm of Clicker Training.

I think one thing we're seeing here though is that for certain forms of dog work, the making of mistakes, which is pretty much required in clicker training, is not an option, and that's where a limit of clicker training might be found. I don't really know enough about it, but I'd think that the number of mistakes that my dogs made when first learning "wipe" would not be an acceptable number for a working dog learning a job in which a mistake of any kind could be dangerous to dog, handler or stock.

Yeah, thats kind of what I meant.
 
Also, another limit is that with this kind of training you work on each behaviour (or each aspect of each behaviour) seperately, which makes for a good solid result but it can make training a bit like "whack-a-mole" ... I think you do need to incorporate the method into a more holistic approach that tackles things like fear or stress at it's root.

ORIGINAL: brookcove

You increase drive in a positive way by withholding the reward a little bit at a time - except in really prey driven dogs, where you can simply physically or mentally block or restrain them within sight of the goal. We use this in flyball - someone holds a dog while the owner teases the dog with the toy or simply calling.

That sounds a lot like using what's natural to the dog.

If true work ethic could be increased through positive punishment, herding people would have discovered the method, trust me.

Isn't *positive punishment* a oxymoron?