The Limits of Clicker Training

ORIGINAL: Scout in Canada

 But you haven't addressed the root of their fear/reactivity.

 
Always my point and always overlooked.
 
(I have to keep my post short so I am not misunderstood anymore and so I don't get into any more trouble)

Isn't *positive punishment* a oxymoron?

The term positive punishment refers to:
+P: Positive punishment results from action that is ADDED to the context or environment affecting the dog. It "gives" the dog something the dog doesn't like. Examples: verbal reprimand, scruff-shake, leash-pop, yank on the leash, hanging, alpha-rolling

Negative punishment:
-P: Negative punishment is REMOVING something from the dog that the dog expected, or wanted. Examples: removal of expected treat, removal of toy, removal of owner from dog, removal of dog from owner.

The terms do not mean "good" or "bad"
 
ETA: So together with reinforcing actions, you have +P, -P, +R, and -R.

ORIGINAL: DPU

ORIGINAL: Scout in Canada

But you haven't addressed the root of their fear/reactivity.

Always my point and always overlooked.

(I have to keep my post short so I am not misunderstood anymore and so I don't get into any more trouble)

 
Sometimes I don't know that you can address the root. If the dog is poorly bred and has a reactive or fearful temperament, sometimes masking the fear with alternate behaviors may be the best you can do.
 
With my dog, who is leash reactive, I am trying two approaches.
 
First, I'm using the clicker to teach alternate behaviors - look at me when you see a dog, instead of lunging and barking.
 
Second, I'm trying to address her fear/defensiveness by socializing her with other dogs more, and spending more time on the leash - to get her more accustomed to it and eventually convince her that there's nothing to be afraid of.

So we're all clear on the terminology, from the Wikipedia entery on operant conditioning:

1. Positive reinforcement occurs when a behavior (response) is followed by a favorable stimulus (commonly seen as pleasant) that increases the frequency of that behavior. In the Skinner box experiment, a stimulus such as food or sugar solution can be delivered when the rat engages in a target behavior, such as pressing a lever.

2. Negative reinforcement occurs when a behavior (response) is followed by the removal of an aversive stimulus (commonly seen as unpleasant) thereby increasing that behavior's frequency. In the Skinner box experiment, negative reinforcement can be a loud noise continuously sounding inside the rat's cage until it engages in the target behavior, such as pressing a lever, upon which the loud noise is removed.

3. Positive punishment (also called "Punishment by contingent stimulation") occurs when a behavior (response) is followed by an aversive stimulus, such as introducing a shock or loud noise, resulting in a decrease in that behavior.

4. Negative punishment (also called "Punishment by contingent withdrawal") occurs when a behavior (response) is followed by the removal of a favorable stimulus, such as taking away a child's toy following an undesired behavior, resulting in a decrease in that behavior.

Also:

* Avoidance learning is a type of learning in which a certain behavior results in the cessation of an aversive stimulus. For example, performing the behavior of shielding one's eyes when in the sunlight (or going indoors) will help avoid the punishment of having light in one's eyes.

* Extinction occurs when a behavior (response) that had previously been reinforced is no longer effective. In the Skinner box experiment, this is the rat pushing the lever and being rewarded with a food pellet several times, and then pushing the lever again and never receiving a food pellet again. Eventually the rat would cease pushing the lever.

* Non-contingent Reinforcement is a procedure that decreases the frequency of a behavior by both reinforcing alternative behaviors and extinguishing the undesired behavior. Since the alternative behaviors are reinforced, they increase in frequency and therefore compete for time with the undesired behavior.

ORIGINAL: Scout in Canada

Isn't *positive punishment* a oxymoron?

The term positive punishment refers to:
+P: Positive punishment results from action that is ADDED to the context or environment affecting the dog. It "gives" the dog something the dog doesn't like. Examples: verbal reprimand, scruff-shake, leash-pop, yank on the leash, hanging, alpha-rolling

Negative punishment:
-P: Negative punishment is REMOVING something from the dog that the dog expected, or wanted. Examples: removal of expected treat, removal of toy, removal of owner from dog, removal of dog from owner.

The terms do not mean "good" or "bad"

ETA: So together with reinforcing actions, you have +P, -P, +R, and -R.

Which are all in play when you apply to scent hounds to positively reinforce the seeking of a smell that is repulsive to the dog?

What sort of smell would be repulsive to a dog?  And how owuld you know the dog found it replusive?  Sorry.  OT OT OT.  This has now gone off the rails into a discussion over the terminology and how it should be applied.... A sticky with houndlove's quote might be useful in future, but as for this thread, I see no hope!

Which are all in play when you apply to scent hounds to positively reinforce the seeking of a smell that is repulsive to the dog?

That would probably fall more under the realm of classical, not operant, conditioning. You would counter-condition the dog to view the bad smell as in fact a good thing by continually associating that smell with things the dog already likes. I'm currently working on counter-conditioning one of my cats because she had come to associate the top floors of the house with the presence of dogs, which she does not like. She already very much likes being groomed, so I've begun grooming her on the top floors of the house only, so that she has begun to associate the top floors of the house more with the good thing (grooming) than the bad things (the dogs). It has actually worked like a charm and she's begun to spend the entire night upstairs, sleeping in bed with me, which she hasn't done for almost a year, since we adopted Marlowe.

When you "load the clicker" in clicker training, you're using classical conditioning to move a stimulus that most dogs view as neutral (and some view as pretty loud and bad) into the good column by continually pairing the sound of the click with a greatly desired reinforcer. That is the part of clicker training that falls under the umbrella of classical conditioning, but the rest of the process is operant conditioning.

ORIGINAL: brookcove

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]

 
So was I, because of the behavior of some extremists in the dog training community. Clicker-training should not be blamed for where the "Positive Only" crowd has taken it.
 
Bob Bailey himself has addressed some of these folks and has said that although he uses "clicker-training", he is not a "clicker-trainer". He is not a "Positive Only" trainer.
 

ORIGINAL: Scout in Canada

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.

 
You're right. It does not address the behavior. There is a difference between shaping a behavior, addressing a behavior, and supressing a behavior.
 
But there are those who believe that distracting, redirecting, shaping, or manipulating the dog's behaviors (symtoms of instability, frustration, or confusion, IMO) is all you need to do, and that "social learning", leadership, fulfillment by design, has little or no value.
 
It depends on who you are talking to, and their personal choices and/or beliefs. []
 
 
 
 
 
 
 

But there are those who believe that distracting, redirecting, shaping, or manipulating the dog's behaviors (symtoms of instability, frustration, or confusion, IMO) is all you need to do, and that "social learning", leadership, fulfillment by design, has little or no value.

 
I do believe in some cases, certain dogs, the distracting/redirecting/etc is all you can do. Depending on their genetics, early socialization/trauma. I don't think being a good human leader and giving the dog a secure knowledge of their place in the pack is going to cure fear aggression or many other issues. But certainly leadership and security are a prerequisite to making changes to a dog's issues.

Hmmm.

I would like to speak to addressing the behavior, and to the idea of teaching a dog not to do something, and how that works in terms of marker training.

Addressing the behavior:

I do think that teaching an alternate behavior is a legit way to "address the behavior" for two reasons: the dog can't speak English (so you can't really get into a Why Session with your dog no matter how much you want to try. It's simply not possible.) and the dog needs to know what *to* do when it wants to, say, bite, or freak out on a leash.

Leash reactivity is a good example because it's so easy to address the behavior using operant conditioning. I had great success using operant conditioning to get a completely buck-wild dog to walk on a leash on crowded city streets with other dogs and children nearby. She needed to know what *to* do in this situation and was more than happy to do it my way--she relaxed a lot on leash once she and I worked out what the appropriate routine is. Here's the trick: her 'root cause' or whatever seemed to be arousal based, so in order to address that root arousal, I needed to work at her level, not at an already-aroused level, and build toward decreasing the space between her and other stuff very slowly, with her giving the cues that she's ready to take more stimulation.

I do think that this is addressing the root cause of the behavior, which I see as arousal. She wasn't supressed on leash, she was happy to know how to do it right and happy not to be aroused. I don't see this as suppression, although I do see that the "whack-a-mole" picture makes it sound like suppression.

I lived with this dog and I see a different thing. I saw a *very* confused, undersocialized animal with a very low tolerance for stimulation and one dangerous way of handling too much stimulation that was obviously self-reinforcing. Every time I isolated a situation and gave her new tools for handling it, she *blossomed*. She wanted the relief that I was giving her every time she and I figured out a new scenario. But dogs don't generalize well, and every single new situation (or situation that she decided was no longer tolerable) created serious potential for danger.

Honestly, if I wasn't in a city that is full of arousing things, and if I had a private yard so that I didn't have to subject my dangerous dog to my whole community every time she had to pee, then I could imagine getting to a point where maybe she could generalize calmer responses as a default, the way she generalized sitting in front of doors or not lunging after stuff that falls on the floor. But I could not guarantee that in a way that is fair for my neighbors.

Here is what I cannot imagine. I cannot imagine +P doing anything but messing this dog up more, because the problem was arousal, and she found my initial correction-based approach extremely arousing. She would never have bitten me, but she would redirect bites after being given a correction, lunging after the first smallish or frightened-looking or fast-moving thing she saw. Corrections in this situation were dangerous.

So my point: I saw this as a pretty straightforward leadership issue. I had to tell the dog what is appropriate, and when the dog figured out what is appropriate, then that situation was all good. I don't think this dog had leadership issues as much as it was kinda crazy--she knew that I was in charge.

The problem was simply that the dog could not handle the only environment I had to give her. What, in this case, would "addressing the behavior" mean to you?

ORIGINAL: Angelique


But there are those who believe that distracting, redirecting, shaping, or manipulating the dog's behaviors (symtoms of instability, frustration, or confusion, IMO) is all you need to do, and that "social learning", leadership, fulfillment by design, has little or no value.

It depends on who you are talking to, and their personal choices and/or beliefs. []

Well said Angelique.

Bob Bailey, certainly a well rounded expert to say the least.

ORIGINAL: fisher6000

I do think that teaching an alternate behavior is a legit way to "address the behavior" for two reasons: the dog can't speak English (so you can't really get into a Why Session with your dog no matter how much you want to try. It's simply not possible.) and the dog needs to know what *to* do when it wants to, say, bite, or freak out on a leash.

IMPO One certainly get to the why by understanding dogs natural behaviors, their back ground, what's happening with them now, and what the dog shows us. And it is this that makes it possible to completely change the dogs behavior to one that is...balanced. How else do you explain the fact that some of us are very successful doing so?

ORIGINAL: mrv

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.

You have explained this quite nicely!  I think that some folks have trouble understanding that there is a way to introduce the prong as a non-aversive, as gunny seems to have done.  The difficulty is that some dogs will never perceive it that way - we are talking about dogs that have a considerable amount of natural and built drive.  As Becca points out, their breeding has a lot to do with that, and it is unlikely that someone could build that much drive in a dog that was genetically predisposed not to have any.
Positive reinforcement builds behavior, period.  So, if the collar becomes a discriminative stim, it isn't an aversive in this case.  But, as we can see, some people here are having a difficult time understanding the whole concept, and would probably benefit from a discussion of the basics of operant conditioning first.
As to whether someone can train a dog not to bite, I prefer to think of it this way.  You can lessen the chance that a dog will bite, but aggression is never gone.  Just because a dog lives out his life without biting does not mean that he isn't capable - it simply means that the trigger was insufficient to remove his inhibition, either because he was steadily rewarded for being calm, or severely punished enough for biting.  But, to say that a trigger sufficient to induce a bite would never pop up is assuming a lot.

Teaching a dog not to do something:

I have had success with installing default behaviors. My dog sits at the door because we practiced that one a lot. My dog poops as soon as we go out because I reinforced it heavily when he was a pup. My last dog got really good at impulse control because we practiced it a lot: you could drop a steak on the floor and she wouldn't touch it until you told her it was OK.

I do think that these default behaviors are very powerful because they come from the dog's desire. It's like having an employee who wants to do a good job because of their own integrity instead of having an employee who wants to do a good job because you are watching them. But I do think that there are some behaviors that are so self-reinforcing that they cannot be replaced 100% reliably with default behaviors of your choice.

With my last dog, this was biting and chasing. Too powerfully self-reinforcing to get rid of entirely. Chasing I worked around by turning it into a reward--I dropped her leash and allowed her to chase after she did what I asked of her. This, paired with other legal prey-chase outlets at home, decreased unwanted chase behavior (ie, going after bicycles) a LOT, but not 100%.

Biting was similarly re-directed into games like tug that had strict rules, and turned into a reward, but this was not 100% either, and became kind of dangerous as tug was really stimulating for her--my husband got bit pretty badly during a tug game and we decided that this wasn't such a good idea.

But many dogs have at least one nasty self-reinforcing behavior: counter surfing, garbage rummaging, barking... Mudpuppy has written about this a lot, and I agree. If the problem is not arousal-based reactivity, and if you've really told the dog what *to* do first, then an aversive like a scat mat or bark collar is the only humane thing to do--for everyone. You've got to be able to communicate completely with a dog, and I don't think it's OK to limit yourself to +R when there are a small handful of situations that really do call for well-placed and well-executed +P.

So to stick with the original intent of the post, I think that's a (small) limitation of clicker training.