Comparing apples and oranges and coming up with magic beans

That blog post you link to is weird, because it kind of lumps Cesar Millan in with dog trainers. While I don't like Cesar Millan's methods, there is one thing we both agree on, he's not a dog trainer. Cesar Millan's methods won't train your dog to sit, or roll over, or anything. So with regard to that, "alpha theory" won't give you any results.

Later he says:

<cite>True, the alpha theory has very little basis in science. But it's starting to look like learning theory isn't as scientific as we've all been led to believe it is either.</cite>

Okay, this is kind of confusing. It's true, alpha theory does have little basis in science. But his next sentence doesn't make any sense because learning theory <em>is</em> scientific in that it is actually studied by scientists using scientific methods. How much more scientific can you get? That doesn't mean it's correct, that's why it's a theory, but at least we know lots of people are constantly look at it and improving it. The same can't be said for alpha theory.

It feels good to vent about his crazy post, but in the end I've seen "this guy" before. He's the guy who writes blog posts that don't make any sense. You can go through sentence by sentence and find at least one logical error in each one. I don't know if he's cognitively impaired or crazy, but he's one or the other, or some combination of the two.

__________
doxienews.com

Usually "this guy" doesn't have a blog on Psychology Today, alongside Mark Bekoff and Stanley Coren. I'm used to crap science in "mainstream" publications, but....seriously.

--------------------
Eric Goebelbecker

As usual, great post Eric!  That link was rather amusing.  His whole argument is that learning theory isn't working because Ian Dunbar and Nick Dodman haven't taken the world by storm?  He forgets to mention people's propensity to be coercive and their own reinforcement for punishing.  Anyone who has ever read a psych book is laughing at this guy who hasn't a clue.  His examples are ridiculous and CAN be explained using Operant and Classical conditioning. It gave me a laugh!

Liz Catalano, MA, CPDT, CDBC www.thelittlewhitedog.com

Hi, Erik,

Thanks for reading my article. You make some interesting points. I was especially happy to hear you say that you think learning theory is one-dimensional, which was kind of what I was arguing at the end of the post when I said, "there's a (perhaps) tiny piece of the smooth-running operant conditioning machinery that's simply, well ... inoperative."

Given your point that learning theory is one-dimensional, how can you fault me for agreeing with you?

However, your apples to oranges argument is a bit off. I'm comparing two competing theories of canine behavior, so I'm sorry, it really <i>is</i> an apples to apples analysis.

Using your gravity metaphor, it would go like this:

There are two theories of lunar behavior. One theory proposes that the moon orbits the earth because the earth is dominating the moon (which is actually true on a certain level). The second proposes that the moon orbits the earth because it's a self-reinforcing behavior (which is also true, sort of, though it doesn't really explain anything). However, neither theory is based on the laws of nature. If the earth dominates the moon it's because the earth has more mass than its lunar "pack mate," not because the moon recognizes the earth as its "leader." And if the moon's orbit is self-reinforcing, that's only because the law of gravity is creating whatever reinforcement, if any, is need to keep it spinning around us.

The question is, can animal behavior be explained through the laws of nature (which I'll explore in my next article), or are we simply stuck with what we've got? And I don't think we <i>are</i> stuck.

Also, Erik, I fail to see how a <i>conditioned</i> response to seeing scraps of food on the ground (i.e., the scavenging behavior), followed by a conditioned stimulus (being praised), which results in a conditioned behavior (of giving up the scavenging) adds up to classical conditioning. If I'm missing something, please explain it to me.

And as for doxienews and Liz Catalano, saying that my posts don't make sense or are full of logical errors (doxie), or that my examples can be explained via learning theory (lc), may I just suggest that what you've written isn't quite the same thing as actually pointing out my errors in logic or describing <i>how</i> learning theory can be used to explain my examples. It's one thing to label something as classical conditioning, as Erik does. But it's quite another to fully explain how it falls into that category, which no one has done yet, not here, not on the comments page at PsychologyToday.com.

From what I've gathered, no one here believes that learning theory is a 100% accurate model of learning, which kind of makes the first part of my argument for me. So I'm not sure I understand the hostility to something that all of you already know. Perhaps I needed to take an extra step, add another argument to my syllogism to make it more cohesive.

I still think the fact that +R training hasn't dominated the marketplace speaks for itself. These things operate on a self-emergent level, just as the wolf pack does. So trust me, if someone had a training system that always worked, with all dogs, all the time, it wouldn't take long for people to be knocking down that system's doors. They would be orbiting around you like the moon orbits the earth!

I will back down on one thing I said. I think I <i>do</i> understand the hostility that my post has engendered here: if you can't refute the message, you attack the messenger.

LCK

Fine, clarify the examples that you give and I will gladly explain them according to learning theory. I stand by what I said. The Alpha theory was based on erroneous information.  Learning theory is scientifically defensible.  I believe that Eric calls it one-dimensional because animals also learn in other ways, i.e. classical conditioning, social facilitation, etc.  The only hole is that there are some people who don't believe in using all of the quadrants. As for what works, if you give even the most wonderful Martha Stewart recipe to a lousy cook - you'll still get rotten results.  

Liz Catalano, MA, CPDT, CDBC www.thelittlewhitedog.com

Lee, I'm not going to argue with you because that would be a waste of time. However I am going to respond because I am experimenting with dealing either online or in person with people that use your debating tactics.

You use a shotgun approach to logic. You throw all the information you have out there very sloppily and then you expect the people who disagree with you to sort it all out for you and present you with those results or else you claim you have won the debate.

But people aren't interested in doing that.

For instance you link to an article that says alpha theory is based on a misguided premise, but then you seem to say it is the better theory. Or at least I think you do, you say it is winning, but you don't really give proof of that.

This is one of probably hundreds of instances of confusing logic in your blog post. We aren't going to sort your thoughts out for you, we just recognize this behavior, state that you are crazy, spend a brief time in wonderment of your craziness, and move on.

__________
doxienews.com

Doxienews, You said it much better than I could!  

Liz Catalano, MA, CPDT, CDBC www.thelittlewhitedog.com

I haven't interfered as of yet, and I'm always glad to see spirited debate, but please keep it on point and contructive.

Thanks,

Kelly Gorman Dunbar Editor, Dog Star Daily

R+ is losing the training wars? It doesn't seem that way to me: R+ training is the primary method taught at both PetSmart and PetCo. I have no idea what their market share is in dog training, but they are a lot of stores teaching a lot of dogs. There are scads of books, websites, and training centeres based on R+ as well. In fact, every training center in my area uses at least some R+. They may mix that in with corrections, but it should also be noted that traditional training a la choke collars, etc., does use operant conditioning as well, it just focuses on a different quadrant.

R+ is also on television, with both Victoria Stillwell and Superfetch, and while neither has the popularity of Cesar, I don't know if that has to do with methods so much as charisma. (Not having cable, I can't really comment on the shows much beyond that, although I gather that R+ just doesn't make for "good" television.)

Perhaps operant conditioning is one-dimensional, but I don't think that means we ought to dismiss it out of hand, nor do I think that means it is "losing" the war.

And on a personal note, I rather take offense at the notion that I'm treating my dog as a lab rat, and not as a dog. I am using principles tested on lab rats, sure, but many things are tested on lab animals before they are used on humans; medications would be an excellent example.

Crystal and Maisy RL1X AOE-L1 CGC St. Paul, MN

First, they might both be "competing theories of canine behavior," but one of them is also a theory of behavior for just about every organism that has a nervous system, while the other is only a theory of canine behavior. They are not apples and apples, especially when you can describe many traditional training techniques in terms of operant conditioning. You tried to pit one against the other, and it doesn't work.

I don't know how "a conditioned response to seeing scraps of food on the ground (i.e., the scavenging behavior), followed by a conditioned stimulus (being praised), which results in a conditioned behavior (of giving up the scavenging) adds up to classical conditioning."  I'm not the one who describes the praise as changing Freddy's emotional state over here. Maybe you should publish and clue us all in on how you are able to determine that is the case, and how you did it without classical conditioning (or medication.)

In my comment over on your blog I mentioned that after reading and thinking a bit more I'm changing my theory: I think it's either DRO or DRI. Chances are the praise somehow made paying attention to you more rewarding than scavenging. Sounds like a great dog!

But you know what? You can have your one example. I'm not going to balance past 60 years or so of behavioral science on a dog, a chicken breast and a brief explanation. It's clear that your mind isn't going to be changed. You're having too much fun using the very terminology you reject to dispute possible explanations.

You said:

From what I've gathered, no one here believes that learning theory is a 100% accurate model of learning, which kind of makes the first part of my argument for me.

Except, well, it doesn't. The first part of your argument is:

Of Mice and Mutts: Why Behavioral Science Is Losing the Training Wars

And the second part of your argument is a link to a wikipedia article that covered both Operant and Classic Conditioning.

And then rest of your article pits operant conditioning on its own against pack theory 

You even wrap up with:

And what they're showing us, in these silly training wars of ours, the thing they've detected that our cleverest scientists have overlooked, is that there's a (perhaps) tiny piece of the smooth-running operant conditioning machinery that's simply, well ... inoperative

I only know a few clever scientists, but I'm pretty sure all of them know about that thing with the dinner bell and the dogs.

--------------------

Eric Goebelbecker

And  by the way, you're one who decided there's a war, and apparently you're the one who decided who's winning based on some mysterious criteria.

There are plenty of skirmishes between different camps of trainers — hell, that's what the Internet is for. 

But in the background, many, many, dogs are being successfully trained, in real life, outside of labs, using that lab rat stuff. 

Isn't that what really counts?

--------------------

Eric Goebelbecker

The other blogger mentions "cracks" in operant conditioning, making it clear his knowledge of the subject is lacking. I just want to make the point that those "cracks", are variables to be duly noted and accounted for. Variables do not negate the basic laws of learning.

Mira Jones, CPDT, CDBC

Your thinking on the plane analogy is revealing of the larger problem you have with Lee Kelley's article.

You treat gravity as a known variable. Because planes do indeed fly, therefore A is true and B is false. In dog training things are not so clear. Proponents of +R and Alpha  don't agree on "gravity" and have set out to build two different models for building airplanes. Neither design flies true 100% of the time but both work to an extent. Alpha "pack theory" gives us a model to follow in Cesar, while +R provides us with little bells and whistles(literaly). Of course both sides believe that they have the correct model and both sides blame the owner or environment for the crashes in their camps. The consumer is left with an end product that is not structurally sound.

Correct me where I'm wrong but the following isn't 100% true... "Operant conditioning is observable, measurable, and reproducible." I think this needs to be amended to include the caveat... "in some contexts and depending on the subject at hand". You can't take all dogs and praise them and stop them from scavenging. Each dog is different and each context is also different. You can add another acronym and claim you guessed right all along, that the inverse is true because of context. Though that sounds like you're moving the goal post to fit the model. "Cracks are variables" is just mumbo jumbo for we might get the analysis wrong. All you are doing is applying a label, documenting ex post facto. This has limits in practice by the average owner because you haven't explained anything, just stated what becomes obvious after the fact.

Also, you state "Comparing the two on an equal basis is apples and oranges" but why is there a distiction between "stimuli" and "interactions between n+ organisms"? Why would organisms learn/behave differently if a machine dispenses kibble or when interacting with another dog/wolf etc? The same principles should be in play either way, right?

I think that there is some validity in questioning Operant Conditioning and comparing it to techniques employed by Alpha trainers. You state... "Operant conditioning describes how the consequences of voluntary behaviors modify their frequency. Punishment makes a behavior occur less often. Reinforcement, more often." Ok, so in the Alpha model, punishing certain behaviors is also used to make them occur less frequently. It might be based on a false concept but it's the same end result and a lot of times those behaviors come back much worse. Why?

You then state... "no knowledgeable trainer would try to solve a serious behavior problem with just operant conditioning... [they use or know of other tools such as]classical conditioning, as well as the specifics of canine behavior". Why not? If the principles are not universally applicable, at what point do they break down and not become useful in training? And what are these "specifics of canine behavior" you speak of which somehow refute critique w/out example? If isn't useful in such circumstances, when do they lose their worth, why should one believe that they are useful at all, or that it is the operational premise for the behavior to begin with?

I don't think that anyone here (Mr Kelley or Mr. Goebelbecker) has the actual estimates on measured popularity of one method over another. I think Mr Kelley's piece was meant to be somewhat provocative and challenge some long held beliefs and reveal that there is a lot of 'faith' within the behavioral sciences when applied to training techniques used by average owners in the 'wild'.

Dogs do form associations and trainers can influence behavorior this way to an extent, but not for all things at all times. This I feel, is where Mr Kelley is equating the "Myth of the Alpha" with "Myth of +R".

This is a fascinating discussion!  Lee, thank you for starting this.

Operant conditioning IS an "incomplete and inaccurate" description of how organisms learn.  It is not synonymous with "learning theory".  It is only a part of learning theory.  As Eric pointed out, there is also classical conditioning, and as Lee's link to Wikipedia pointed out, there is also cognitivism, constructivism, etc.  "Learning theory" is the theory of how learning takes place and no we don't have a 100% accurate model of this.   But pretty much all dog training is the effort to increase some behaviors and decrease others.  Sometimes we go about this by changing underlying emotional states (scare the dog into not reacting, or make it associate the situation with something the dog likes).  But when we effect the behavior itself, we are almost always applying rewards (R+) or applying punishers (P+), BOTH of which are part of operant conditioning. 

What I find interesting is that pack theory is a GREAT model for describing how individuals in a social group might interact with each other, but how can that effect a human wanting a dog to sit, or to stop eating food that the human clearly doesn't want?  In the non-human species where there clearly IS a social hierarchy, individuals don't train each other to follow commands, or try to modify their natural behaviors.  You don't see high-ranking chickens preventing low-ranking chickens from scratching for food or clucking.  Dominant elephant seal bulls don't get submissive ones to walk next to them - they can't even keep them from following females around; all they can do is keep them from entering their territory.  Dominant animals just want to keep the important stuff to themselves - they otherwise don't really care what the subordinate animals do.  

(This is probably NOT true among humans, who seem to often form dominance hierarchies that extend well beyond access to resources, at least among males.  We do order each other around.  But not all ordering, or behavior modification, among humans is due to dominance.  I don't "dominate" the waitress who brings me food; I don't even "dominate" my child or my husband.  Yet these individuals often do my bidding.)

(The earth is "dominating" the moon?  That's just... lunacy!  Sorry, couldn't help it!)

I thought you might like this subject line.

I don't think of operant and classical conditioning as 'bells and whistles', they are the principles of learning.  It is important to remember that the 'concept of Alpha' is simply a model for why dogs do behave as they do, an incorrect model but a model.  Trainers who follow pack theory, still employ the use of operant conditioning.  The Alpha concept and Operant conditioning are not mutually exclusive, one is based on ethology the other is based on psychology. When used in training, they do not function independent of each other.  Using the concept of alpha simply supplies the reason for applying +P or -R.

Correct me where I'm wrong but the following isn't 100% true... "Operant
conditioning is observable, measurable, and reproducible." I think this needs
to be amended to include the caveat... "in some contexts and depending on the
subject at hand".   

Operant conditioning is observable, measurable and reproducible.  Yes, there is variability in what some dogs will be rewarded by: praise, food, toys, or even an action as Lee says "acting submissive".  If it is something that the dog likes, it is a reward.  But that doesn't change that it is still a reward and when it is applied at the right time it will reinforce behavior.

You can't take all dogs and praise them and stop them from scavenging.

No, rewarding a behavior (i.e., praise) will not stop a behavior.  Punishment stops behavior.  The question is: do you teach the dog a different behavior so that scavenging is less rewarding than the new behavior and scavenging goes away (training an incompatible behavior) which would be +R; or do you punish the behavior to stop it +P?  The distinction is according to Alpha Theory is that you would punish the behavior because the dog is scavenging to be dominant.

Also, you state "Comparing the two on an equal basis is apples and oranges"
but why is there a distiction between "stimuli" and "interactions between n+
organisms"? Why would organisms learn/behave differently if a machine
dispenses kibble or when interacting with another dog/wolf etc? The same
principles should be in play either way, right?

I'm not sure what you are referring to with 'a distinction between stimuli' or 'n+'.  I don't think organisms learn differently with a machine that dispenses kibble or when interacting with another dog, bear, elephant, or unicorn. The same principles apply. If a behavior gets rewarded with kibble the behavior will continue, if they get what they want out of the interaction with any of the mentioned species, the behavior will continue.  If they don't get kibble, whatever they wanted, the behavior will stop.

Eric was comparing the Alpha Concept to Learning Theory, and they are apples to oranges.  Somehow the discussion changed to  defending coercive/punishment-based training over positive training.  Funny thing is: the BOTH follow the principles of Learning Theory.

Liz Catalano, MA, CPDT, CDBC www.thelittlewhitedog.com

"For instance you link to an article that says alpha theory is based on a misguided premise, but then you seem to say it is the better theory. Or at least I think you do, you say it is winning, but you don't really give proof of that."

I've written numerous articles critiquing the alpha theory, and even state clearly in this article that it's not based on reality. And I never said or even suggested that it is a better theory. In fact, just the opposite. So am I using a shotgun approach to logic (which I admit I have used in the past for the very reasons you've suggested), or do I just need to dumb down my writing a little?

Also, I never said that the alpha theory is winning, though I can now see that I might've needed a bit more clarity in my writing on this point. After all, if behavioral science is "losing," that certainly would imply that the other side is winning, and it wasn't my intention to say that. In fact, in the article I stated clearly that both models are equally successful. The point that I guess I should have made clearer is that since one of the goals of the APDT was to steer training away from the alpha model, and since that hasn't happened, that's how behavioral science is losing.

As to the "why," that's what's got everybody so hot and bothered.

LCK

Liz answered things quite nicely (better than I could have in a few places) but there are a couple of things that stick in my craw.

You treat gravity as a known variable. Because planes do indeed fly, therefore A is true and B is false. In dog training things are not so clear. Proponents of +R and Alpha don't agree on "gravity" and have set out to build two different models for building airplanes.

First and foremost, operant conditioning and R+ (or +R) are not the same thing. R+ is only a part of OC. 25%, as a matter of fact. As I pointed out in my comments both here and on psychologytoday.com, Mr. Kelley bounces around between criticizing learning theory, operant conditioning and in his own comments, "R+ trainers." Pick one.

Second, yes, I treat gravity as a known variable. Just like I treat OC as a known variable. I'm not going to pretend, even for the sake of argument, that it's not an accepted scientific theory. (Like many scientific theories it may actually be replaced by something else someday. But not by a single dog and a chicken breast.) Maybe in some alternate universe (there we go with the crap science again) someone got a plane to fly despite their disbelief in gravity. Afterwards, the guy who did believe in gravity took a pencil and a slide rule and explained why their craft flew.

Correct me where I'm wrong but the following isn't 100% true... "Operant conditioning is observable, measurable, and reproducible." I think this needs to be amended to include the caveat... "in some contexts and depending on the subject at hand". You can't take all dogs and praise them and stop them from scavenging. Each dog is different and each context is also different.

Ok. You're wrong. I can set up situations where each accepted aspect of operant conditioning is provable. (That would be why each aspect is accepted, by the way.) I can't say the same thing for "alpha theory." (Before I could even try, I would probably have to launch a research project into defining what exactly the theory is, but that's not important right now.) I can't really observe, measure, and reproduce how being the pack leader made a dog sit more or bite less. There's more variables and obscure meaning in that sentence alone than an entire hard disk of Windows(tm) source code.

That is what I was referring to in my original post. The scientific validity or a theory vs. the lack of same in folklore that is based on science that has been discarded. In a scientific context, observing, measuring and reproducing includes controlling for contexts and different subjects. I guess I should have thrown in some more science-y language to be more clear.

You can add another acronym and claim you guessed right all along, that the inverse is true because of context. Though that sounds like you're moving the goal post to fit the model. "Cracks are variables" is just mumbo jumbo for we might get the analysis wrong. All you are doing is applying a label, documenting ex post facto. This has limits in practice by the average owner because you haven't explained anything, just stated what becomes obvious after the fact.

When you have a scientific model you try to stay within it. I don't know why this is a problem. It has cured disease, won a World War, and got us to the moon after all. You try to define things in terms of it. I'm not really sure what you or Mr. Kelley expect us to do with a vague anecdote. Maybe we should just throw around vague terms like "positive and negative energy" or "polarity?" Or how about this: "The dog moves in strange ways?"

Being able to look at a situation and determining what is going on is very useful in real life. Your dog jumped up on the counter once and got a slice of pizza. He's going to try that again. Every time your dog barks, you pay attention to him. He's going to try that again too. When you pay attention to your dog instead of your blackberry, he stops looking on the ground for chicken breasts....

Ok, so in the Alpha model, punishing certain behaviors is also used to make them occur less frequently. It might be based on a false concept but it's the same end result and a lot of times those behaviors come back much worse. Why?

Let me get this straight: you want me to explain why something that's based on a false premise doesn't work? It's because it's not the same end result. Pinning a dog to ground because he's acting "dominant" doesn't punish any specific behaviors and fails to address any underlying emotions. It's, to borrow a beautiful metaphor I saw someone post anonymously, shooting at the fence and hitting the barn. Sometimes, the correct behavior actually does get punished, and that's when it does work, but even an un-programmed VCR is right once a day.

If the principles are not universally applicable, at what point do they break down and not become useful in training? And what are these "specifics of canine behavior" you speak of which somehow refute critique w/out example?

Seriously, pick up a book. OC addresses behavior. Classical conditioning addresses emotions. I think this is covered in high school psych in some public schools? As I and at least one other person now have pointed out, Mr. Kelley peeled off a subset of behavioral science, showed how it fails to address everything, and then declared behavioral science deficient. The fact that OC doesn't address anything isn't a flaw. It's a limitation known to everyone that knows what they are talking about.

As for the specifics of canine behavior...are you serious? Sigh. Probably. Are we discussing two "theories" or has this moved into you proving that some trainers you don't like don't know what they are talking about? Sorry. No sale.

--------------------
Eric Goebelbecker

"The point that I guess I should have made clearer is that since one of the goals of the APDT was to steer training away from the alpha model, and since that hasn't happened, that's how behavioral science is losing."

Yes, that point was lost on me, so thank you for clarifying that, although I'm not sure what, exactly, you think behavioral science is "losing." It seems to me that the vast majority of dog training is based on either OC or CC, regardless of whether the trainer is using mostly R+ or P+ (or the other quadrants). If that's the case, then behavioral science can hardly be "losing" anything since it's in widespread use.

So far as the APDT "failing" (my word, not yours, but that seems to be what you're suggesting)- well, it has certainly grown in popularity, as has R+ in general, so it would seem that while the APDT has work yet to do, it hasn't failed in its mission to promote dog-friendly training. I don't think it's reasonable to expect that all dog trainers will ever agree on training methods. As long as trainers are using well thought out, well timed methods that are fair and understandable to the dog, I have little to quibble about.

Crystal and Maisy RL1X AOE-L1 CGC St. Paul, MN

Great discussion topic!

First of all can we stop throwing the term 'theory' around - in science, since this is what we are tailking about, this term has a specific meaning.  A theory is as reliable as a statement can be - it is usually a group of hypotheses that have been proved so many times that they are thought to be true - this is as close as we get to certainty in science.
So to repeatedley attribute the term theory to alpha schmalpha stuff is really starting to grate - but then I guess this whole area is synonymous with misinterpretation and misuse of terms.

Secondly, I have read a series of blog posts on LCK's blog regarding dominance in domestic dogs and found it to be very good - so well done there!

But I do have to take exception to several of the points raised in the same blog (over different posts) regarding learning theory. 

You list a couple of different situations as experiments e.g. your version of 'alpha rolling' and your dog scavenging trash off the street.  These are not experiments, there is no control, there is no provision made for variables that are likely to affect the perceived results so your observations of a rather small sample are not really worth that much.  So to use these examples as reasons for discounting a huge area that is backed by countless trials is a little weak IMO.

Also you make a mistake in your labelling of your praise to your dog (the scavenger!) as positive reinforcement.  Positive reinforcement by definition is something that increases behaviour and this example, the scavenging behaviour, was not increased.
Something is positively reinforcing to an individual if it motivates them to repeat behaviour and is clearly linked with that behaviour.

There are so many other things that could possibly be going on with your dog that may or may not explain his/her behaviour but it is not positive reinforcement.  If there is a decrease in behaviour, punishers must be involved. 
Whether we know it or believe it the four quadrants are a force affecting the behaviour of animals, it is most certainly not the be all and end all in relation to learning or modifying behaviour but it is pretty influential especially when manipulated by someone who knows what they are doing.
Even people who have no clue about operant conditioning may still be using it and providing operant consequences for their dogs e.g. as highlighed, users of alpha/dominance as a training paradigm.

As the original post points out OC is not the only port of call but rather forms the basis for many training programs but there are soooooooooo many other factors involved, not just at a species level but also when dealing with the individual.

Anne Rogers
Pet Central, Ireland

Surely this is a manifestation of instinctive drift. Problems with operant conditioning occur when you enter into the realm of instincts, evolutionary shaped behaviours. In the chicken training camps, no-one ever trains the chicken to stand still, they always dance ... and these problems must surely happen in dog training too. Although you can apply operant conditioning, when you encounter food gathering behaviours, which for a dog I guess are chasing/biting, you can have problems. I think LCK is looking at ways to tackle the Misbehavior of Organisms in the dog training world by looking at emotions and harnessing feelings of flow, where rewards are internal.

Anne Rogers: "You list a couple of different situations as experiments e.g. your version of 'alpha rolling' and your dog scavenging trash off the street.  These are not experiments"

Of course they're experiments. The resulting data may be somewhat limited by the scope, but they still qualify as experiments. After all, Einstein came up with relativity by doing a gedanken experiment while riding a train.

LCK

Not wanting to flog a dead horse here but LCK, with all due respect you have quoted me out of context (or just cut the quote short) plus I have not been as specific as I should have been.  These are not scientific experiments because it is very difficult to replicate them - lack of control, accounting for variables etc.etc.  This making your observations weak especially in relation to the tonnes and tonnes of replicable work on learning. 
Many great discoveries hatched in seemingly mundane, non-scientific, out of context situations but before publishing or at least before using these developing ideas to debunk other more credible solid work they have been worked on and worked on to improve air-tightness.   

Thanks for responding to all the ins and outs here but I can't respond any longer in a discussion that is likely to go round and round and round rather than tackling the topics at the core here.

Anne Rogers,
Pet Central, Ireland

You summed it up quite well Anne.

I have largely dropped out of the conversation because it seems like a lot of you know a lot more about training than I do. But I have a pretty simple question I think any of you could answer. LCK said:

The point that I guess I should have made clearer is that since one of the goals of the APDT was to steer training away from the alpha model, and since that hasn't happened, that's how behavioral science is losing.

My question is: Does anyone actually recommend "training" a dog using the alpha model? Cesar Millan doesn't train, and doesn't claim to, and I think the "New Monks" just recommended punishment, but didn't necessarily talk about "dominance" from my very little knowledge of them, which may be wrong. My interpretation of what LCK said was that the "New Monks" and Cesar were the same, but I don't think they are, are they?

To clarify my question:

Does anyone suggest training a dog (sit, down, stay) using alpha theory (dominance theory)?

__________
doxienews.com

Yes, since Millan came onto the scene there's been an explosion of young trainers advertising themselves with "be the pack leader," "be the alpha," etc. Plus, many of the older dominance-style trainers never really went away.

LCK

Great blog, Eric. The article you're referring to is an unfortunate but perfect example of someone who knows enough to sound knowledgeable but not enough to be useful. In fact, similar to the DW, he knows enough to be dangerous. Loved your response.

Pat Miller

Gravity does not equal Operant Conditioning (OC) in the thought experiment.  Gravity in the example would be the training regime (either Alpha or +R).
Just to be clear I'm not saying +R is equal to Operant Conditioning either. Throw a plane in the sky and see if it flies. Throw a dog into
a training regime and see if s/he's trained. There is unanimous agreement on gravity in the airplane building community, whereas this is not so within the dog training community. We're not talking about OC in the lab, just to be clear.

Therefore, in the example, "Gravity" is NOT a known variable. Both training regimes act differently on the dog and produce output that isn't operated
on by a constant, as is the case with gravity and planes.

It's a silly example but your usage of it to belittle Mr. Kelley is ironic.

Also, I meant bells and whistles literally, not pejoratively. Those are the training tools that +R give us to work with, bells, whistles, clickers, etc.
My POINT was that each training regime has a different design, and hence we see different approaches and different tools coming out of each camp.

I haven't said that OC and +R are the same thing. But +R is the training regime extracted from OC. This might be the origin of the "confusion".
Learning theory, Operant Conditioning and +R each involve the concepts of schedules of rewards and/or "punishments" to manipulate behavior and/or
emotion. Each concept might be distinct (+R doesn't involve type I punishment), but they involve some of the same concepts
in a high level discussion like this one. However, I suspect that the issue here isn't really about the misapplication of the labels.

I think most are aware of the definitions and what they mean on a technical level. But the problem is that in the nitty gritty world there is NOT
 uniform diagnosis and a linear one to one mapping of stimuli to behavioral output. I think this limits its use as an effective training tool. (**Note: this doesn't translate into an endorsement of Alpha theory/methodology)

"""
Also you make a mistake in your labelling of your praise to your dog (the scavenger!) as positive reinforcement.
Positive reinforcement by definition is something that increases behaviour and this example, the scavenging behaviour, was not increased.
Something is positively reinforcing to an individual if it motivates them to repeat behaviour and is clearly linked with that behaviour.
"""

It seems like the whole exercise is one concerning labeling, rather than of explanation. Praise in one context is a reinforcer,
"good boy sit" and not in another "good boy get the trash". The example might be trivial, but for the purposes of the
discussion let's just run with it. In the first example the behavior is reinforced with praise and in the other it is extinguished.
The stimulus takes a new label in a new context, but for some reason the stimulus yields different output.

Good Boy (praise sitting) = more sitting

Good Boy (praise trash eating) = less trash eating (behavior)

Why?

I don't dispute the labeling! You may call it whatever you want and be as specific as you like. But again, by identifying a
label it is just an exercise in documentation. By stating that praise in a new context acts as an extinguisher doesn't address the
nature of the behavior and/or why it is performed and then not performed in the other context.

I have some additional complaints with your Goebelbecker's answers to my questions.  
"""
I don't know why this is a problem. It has cured disease, won a World War, and got us to the moon after all. (In reference to Science)
"""
+R training (not OC) hasn't done any of the following or anything as grand. There are millions of dogs with behavioral issues. This wouldn't be an acceptable benchmark for any of the other sciences (Please don't say I'm equating +R with OC again, you'd be missing the point).
Not all are dogs who are +R trained have behavioral issues, but A LOT of them do.

I know too many people who have done clicker training and +R obedience work who have "whacked" out dogs that don't listen to their commands, especially when there are distractions.
I blame the training regime on this, not the people (they sought professional help). I think this is why the 'science' in it seems thin.
Dog training is not equivalent to the mechanics of flight and a planes relationship to gravity!
And while +R in OC might make sense in clinical applications, highly controlled scenarios, the "science" in +R doesn't carry over so nicely to the field.

"""
Being able to look at a situation and determining what is going on is very useful in real life.
Your dog jumped up on the counter once and got a slice of pizza. He's going to try that again. Every time your dog barks,
you pay attention to him. He's going to try that again too. When you pay attention to your dog instead of your blackberry,
he stops looking on the ground for chicken breasts.
"""
I agree it would be very useful. But +R doesn't provide that. You supply very simple, almost clinical examples. I agreed with
you in my earlier post. It is a nice framework 80% of the time. I also don't think dogs have a sense of "jealousy" over a Blackberry either. However,
that's a different discussion.

"""
Let me get this straight: you want me to explain why something that's based on a false premise doesn't work?
"""
Haha, you know how to pick cherries! You do have it twisted, so let me straighten it out. That wasn't my question. Alpha uses OC (unknowingly
as you admit - and I'm not saying +R = OC just to be clear) and Alpha training fails on the whole to 'train' dogs. Failure means the behavior
(perhaps temporarily extinguished) manifests into aggression in another area of the dogs life.

I did NOT give this example(for some reason you think I did): "Pinning a dog to ground because he's acting
"dominant" doesn't punish any specific behaviors and fails to address any underlying emotions."

The question I posed involved punishing (pinning down,smacking him, etc) for a CERTAIN behavior (getting into the trash). I agree dominance trainers
do the dominance routine for the sake of dominance and that is out of context a lot of times and does damage to a dog.
But that isn't what I'm addressing in my question.

HERE IS MY QUESTION: Why doesn't the delivery of punishment (from +R trainers perspective)
teach a dog to simply stop performing a behavior? Interestingly, why do a lot of dogs who receive punishment become
aggressive? (I know that +R doesn't "punish" in this way, but I'm curious for the interpretation from the +R perspective.)

Now, from what I gather there is no real answer for this. There is just language that defines the results, categorically, as side effects. I understand
that punishment "for some reason" is not the same as reinforcement.

But my question is "WHY IS IT NOT THE EQUAL OPPOSITE"? It seems like +R trainers are saying that dogs are doing a basic arithmetic and making
associations when engaged in behaviors via reinforcement, but that they can't do this when punishments are involved. But this suggests that dogs
learn in TWO different capacities.

For instance, a dog only ever takes food off a piping hot stove burner once! It's a horrible punishment and one that I would not ever
like to administrate. However, it is tremendously effective and the dog seems to 'learn' very quickly via punishment in this context.
Why wouldn't I beat the living hell out of my dog if he gets into the trash based on the same concept?

(**I'm not advocating this in any way.It's just a question!)

Keep in mind that if there is an answer for this I would like to here it!
Don't just tell me to open a book and that I don't understand another acronym. I'm sure that there is an acronym for it and that
a lot of scientists agree on how many words it should have. But an answer, please?

"""
"Ok. You're wrong. I can set up situations where each accepted aspect of operant conditioning is provable.
(That would be why each aspect is accepted, by the way.) I can't say the same thing for "alpha theory."
(Before I could even try, I would probably have to launch a research project into defining what exactly the theory is,
but that's not important right now.) I can't really observe, measure, and reproduce how being the pack leader made a
dog sit more or bite less. There's more variables and obscure meaning in that sentence alone than an entire hard disk of Windows(tm) source code."
"""
You think I'm proposing that Alpha people have a better model? That would an incorrect reading of my critique.

I agree that there is more scientific validity in +R and I certainly agree that it is accepted by behavioral scientists. I also agree that it is
a much preferred, humane method. But Just because it's accepted by a lot of people doesn't for a minute prove its validity.
I agree that it's the best we got at the moment, but this is an acknowledgement that there is something missing.
My point is that the science is wrong or incomplete (in +R). Wrong might be a strong word,
but it breaks down when the environment is complex. There are too many variables introduced and it becomes a guessing game even for
experienced trainers and for average owners. This does not prove "anything", that it is wrong or right.
This might suggest that the infinite flexibility of the model is NOT a feature, but rather a bug.

"Seriously, pick up a book. OC addresses behavior. Classical conditioning addresses emotions. I think this is covered in high school
psych in some public schools? As I and at least one other person now have pointed out, Mr. Kelley peeled off a subset of behavioral science,
showed how it fails to address everything, and then declared behavioral science deficient. The fact that OC doesn't address anything isn't a flaw.
It's a limitation known to everyone that knows what they are talking about."

Great, I'm glad we're in agreement. +R is of limited value. I'm glad that all are aware of this. I'm so surprised by
the hostility towards Mr Kelley's piece then.

And btw, The nature of your 'mind' isn't scientific at all. If I'm unclear in something I've written, inform me and attempt to make an honest effort
for what I'm aiming at. It's easy to belittle someone because they might not be an expert in a subject. I don't profess to be. But you
don't even make an attempt to understand what Mr Kelley (or myself) are attempting to explain or are confused over.

Thanks, LCK. I've had a gander at your blog and plan on spending a lot more time there. It's curious that people get so defensive about these things. From your comments and responses that I've read, thanks also, especially, for being so affable.

Your examples are really interesting, and I know you are someone who appreciates scientific thinking and all that it can do for us. However, I don't see how anything described here goes against or disproves learning theory. When you start with the four quadrants it's a pretty basic model, which I think we can agree is correct. Then when you add complex motivations in real-life scenarios, it's not like learning theory suddenly fails or contains inherent mistakes, even though someone like me (or maybe you, maybe everyone) can't necessarily explain it. Behavioural science isn't invalidated by our current inability to explain what is going on with principles of behavioural science. I find a lot of dog training/b.mod principles are counterintuitive, but that doesn't necessarily mean they're illogical or wrong sometimes. Basically, I think we're mostly all making a very similar argument!

Except here: "So trust me, if someone had a training system that always worked, with all dogs, all the time, it wouldn't take long for people to be knocking down that system's doors. They would be orbiting around you like the moon orbits the earth!" Not true. People can be lazy. People can want to stay stuck in a bad situation because it's easier to blame their "bad dog" than admit that THEIR behaviour is causing problems. People want a quick fix, even if they suspect it doesn't actually solve things. Dogs don't come with the disclaimer that they are likely to be MAD WORK, POSSIBLY FOR YEARS. Also, dog owners don't always realize that that work should all be fun.

Pat Miller wrote: "The article you're referring to is an unfortunate but perfect example of someone who knows enough to sound knowledgeable but not enough to be useful. In fact, similar to the DW, he knows enough to be dangerous."

Really? Let me use your form of discourse -- an ad hominem attack -- to suggest that your response is a perfect example of someone who doesn't know enough about her own form of training to dispute ideas that contradict it, and as a result has to resort to attacking the messenger not the message. (Mind you, I'm not necessarily saying I believe this, but imagine someone who's on the fence in this debate and is reading your comment; has it convinced them that I'm wrong, or just given them the impression that you're dismissive of and discourteous to those who disagree with you?)

You say I don't know enough to be "useful," but useful to whom? If you mean useful to reinforcing what you already believe in, then you're probably correct. But from what I understand (within my limited base of knowledge), you're a well-known quantity in your corner of the +R world. Is that what you want to be known for, attacking people who disagree with you rather than disproving their ideas?

In actual fact, I would guess that you have no idea what my knowledge base is or isn't, just that what I've written in this one article goes contrary to your personal ideology. I've gotten positive feedback on several of my other articles at PsychologyToday.com from far more knowledgable persons than Pat Miller, even when what I've written has disagreed with their points of view. In fact I've changed Marc Beckoff's ideas on why dogs urine mark, as well as certain aspects of play in canines, something he's far more of an expert about than you are about dog training. Stuart Brown has actually posted links to some of my articles on his personal blog. (I'm not saying any of this to brag, just to dispute your position that I only know "enough to sound knowledgable.")

And you also say I know "only enough to be dangerous?" It would be nice to know why you think what I've written is dangerous, rather than using that old "don't listen to this guy" tactic, without telling anyone why they shouldn't listen. Especially when what's really dangerous is not being open to learning new ideas, no matter who or what the source is. (I've learned most of what I know from dogs.)

Here are some more articles for you and for anyone else who cares to look at them. (BTW, the first 4 would make good additions to anyone's library-of-links on why dominance training is unscientific.

https://www.psychologytoday.com/blog/my-puppy-my-self/200904/is-your-dog...

https://www.psychologytoday.com/blog/my-puppy-my-self/200904/is-your-dog...

https://www.psychologytoday.com/blog/my-puppy-my-self/200904/is-your-dog...

https://www.psychologytoday.com/blog/my-puppy-my-self/200905/pack-leader...

https://www.psychologytoday.com/blog/my-puppy-my-self/200908/the-canine-... (Stanley Coren has not sent me a nice note yet on this one...)

LCK

But my question is "WHY IS IT NOT THE EQUAL OPPOSITE"? It seems like +R trainers are saying that dogs are doing a basic arithmetic and making
associations when engaged in behaviors via reinforcement, but that they can't do this when punishments are involved. But this suggests that dogs
learn in TWO different capacities.

************

I'm not sure who said what you stated above.  The way I see it, what is being said (in the real world of training, not just here) is that there IS an equal but OPPOSITE effect.  Yes, associations are being made in the same way whether you are rewarding or punishing.  But I would think you would have a different emotional response to having a person offer you cheesecake and having a person offer you a slap in the face, wouldn't you?  I'm confused as to why this would be such a mysterious concept.

The question is NOT of which works, punishment or reinforcement.  That's silly.  They both work.  The question is, what are we willing to do to dogs, what associations are we willing to instill in order to get the results we want.  Do I want my dog to decrease the behavior of scavenging for food through fear and intimidation, or do I want to increase the behavior of leaving the food alone through rewarding an alternate, incompatible behavior.  Both work, it's my choice. 

So...

1.  There is no choice to be made between Pack Theory and Operant Conditioning.  Pack Theory merely justifies the use of harsher punishments, intimidation and fear as a way of punishing behaviors (because it supposes that is the nature of the beast).  This justification is misguided because it is provided by a flawed and thoroughly refutted psuedo-theory.  Case closed.

2.  Since one cannot intelligently use Pack Theory as the reasoning behind the use of punishments that are painful, intimidating and fear-inducing, we are left with the question of why we would CHOOSE such techniques when the same results can be gained through the use of rewards and kinder punishments while also producing more positive associations.

(By the way, even if Pack Theory were sound, it still wouldn't negate the sentiments of statement #2.)

To be more specific, I don't choose to use lure/reward training to teach loose lead walking because choke chains are ineffective.  I use lure/reward training because it is (at least) just as effective as using a choke chain AND it has the added benefit of creating positive associations with me, the leash, walking and the world around my dog.

Here's a question for you.  If reinforcement and punishment are equal but opposite, both effective in achieving the same result, for what reason would you choose punishment over reinforcement?

Cindy Bruckart, CPDT-KA

www.regardingrover.com

I would like someone else to answer my question (although I think I already know the answer) because LCK gave just the kind of nonsense answer I would expect from him.

My question is:

Does anyone suggest training a dog (sit, down, stay) using alpha theory (dominance theory)?

LCK answered:

Yes, since Millan came onto the scene there's been an explosion of young trainers advertising themselves with "be the pack leader," "be the alpha," etc. Plus, many of the older dominance-style trainers never really went away.

This is an odd answer since the people riding the Cesar Millan wave would be odd to call themselves trainers since he doesn't even call himself a trainer. (This is why I'm very mistrustful of anyone who adopts the title "behaviorist".) I've seen some of these copycats and they say the same lame stuff Cesar says and accomplish as little. Also, how does "be the pack leader" train a dog to sit? Cesar Millan doesn't do that, and doesn't claim to.

Once again in only two sentences LCK manages to throw a ton of faulty logic out there by being as vague as he can. "many of the older dominance-style trainers". So are you suggesting there are older trainers using alpha theory that actually do train dogs? I think what you are referring to people who openly promote punishment, which is very different from dominance theory. We realize Cesar is using punishment, but he would say he isn't. ("This doesn't hurt the dog.") 

An explosion of young trainers does not mean anything. As a sentence it lacks precision, as a concept it lacks logic. The Thigh Master(tm) was very popular for a short period of time, you could even say it was an "explosion" but nobody knowledgeable at the time would have said the thigh master was the way to go over traditional fitness. In fact, informed people of the time could have told you flat out that the Thigh Master would be unlikely to help anyone in any noticable way.

Dog training is very susceptible to "easy fixes" that don't work, just like fitness. When doing the right thing takes actual work and learning, that is always the case. In fitness we don't say these lame solutions are "winning", we know they are traps for the uninformed or desperate. Anyone who spends any amount of energy or actually needs results usually figures out what really needs to be done. 

So in a very "survival of the fittest" way, positive training is winning. Survival of the fittest doesn't mean that all lesser competitors immediately die off, they die off over a long period of time as challenges arise that they can't overcome.

__________
doxienews.com

Thanks for you comment Cindy. I agree with you on the moral grounds over +R vs. Alpha model 100%. To answer your question, as you've stated it, I don't believe that there would be be a reason for choosing punishment over reinforcement in training. However, I was under the impression that punishment doesn't work in training. That it is NOT just matter of effectiveness. 

If I hit my dog to correct a behavior it turns nasty. It doesn't grasp that it was the behavior I was correcting. I hear what you're saying about punishment having emotional impact on the dog. But then why isn't this the rule: dogs always learn via an emotional context? Why are there TWO different processes that seemingly run a dog? Rewards map to 'thinking' associations and punishments map to emotional responses? Why isn't it one way or the other? It seems like it's a matter of definition and syntax on one hand. However, dogs don't learn to map punishments with behavior, so why would this lead us to believe that they do map rewards with behavior?

Why not a 100% emotional model for the sake of consistency? Dogs can clearly grasp the context of punishment via an emotional input/output, as you suggest in your comment. It makes sense that they do the same in all other facets of their lives as well. The distinction between thinking and emotion seems arbitrary in the case of the dog. Furthermore, the fact that dogs demonstate emotional  awareness to stimuli(the punishment), demonstrates (to me at least), that there is little need for the other 'thinking' associations that can't form awarness to punishments.

From the language I have so far heard, +R claims the 'thinking' association making as the secret sauce in training. But if emotional capacity in the dog is great enough so as to inform it to behave(aggressive towards people) correctly(from the dogs perspective) in the future it seems to contradict the premise that associations are formed via thinking at all. Perhaps not contradict, but suggest that there is 'redundent' technology that the dog is equipped with. 

The subject is obviously very nuanced and in depth, definitely beyond the scope of this forum. But thanks for your input!

Cindy Bruckhart asked: "Do I want my dog to decrease the behavior of scavenging for food through fear and intimidation, or do I want to increase the behavior of leaving the food alone through rewarding an alternate, incompatible behavior."

The problem is that I don't think there was an incompatible behavior involved.

Here's the scenario: the dog is used to sniffing and investigating things on his walks. He would also take direction from me: "Hey, let's go this way." or "Wait..." or "Okay to cross!", etc. Along with these very normal behaviors, he had also gotten used to scavenging.

So as an experiment (yes, an experiment), I tried doing the opposite of what most trainers at the time (this was 17 years ago) would have done; I praised him instead of correcting him verbally or pulling him away from the tasty tidbits. After doing that, and only that for about 3 days, his other behaviors didn't change at all. He was still sniffing and investigating his environment and listening to my directions when necessary, but he was no longer interested in scavenging. He might sniff a pizza crust or chicken bone, but that's it. There was no "replacement" behavior, unless you count not scavenging as an incompatible behavior, which it isn't: it's not even a behavior, it's a non-behavior.

So if it's not classical conditioning, and it's not teaching him an incompatible behavior, what is it?

LCK

"However, I was under the impression that punishment doesn't work in training. That it is NOT just matter of effectiveness."

Well, all I can say is that you're under the wrong impression.  To say that punishment doesn't change behavior would be to say that reinforcement doesn't change behavior.  In fact, the only way an aversive can be truly categorized as a punishment (within the realm of training) is if it does, in fact, decrease behavior.  Anyone who says punishment "doesn't work" is either under-educated, lying or in denial.

Maybe I'm dull, but it seems that the issue is being someone stirred around and convoluted.  We can both learn what is right or wrong (what gets us the best result) and form an emotional association during that lesson at the same time.  The two are neither mutually dependent, nor mutually exclusive.

If I get several tickets(an aversive), over time, for speeding (the behavior) and it causes me to decrease the behavior of speeding, then it is a punishment.  If the tickets don't decrease my behavior of speeding, then it may be aversive, but not aversive enough to change my behavior, so not qualifying as a punishment within the realm of training.  Regardless of what I learn, to stop speeding or to watch for cops, I am probably going to build a negative association with cops, and/or flashing lights, and/or the location where I was stopped, and/or driving.  (We never know for sure what the association will be atta ched to.)

If instead, a cop pulls me over and gives me chocolate when I'm driving UNDER the speed, I may or may not learn to drive under the speed limit (depending on how rewarding I find chocolate compared to how rewarding it is to get where I'm going faster), but I will certainly, regardless of learning, build a different association with the cop, lights, driving, location mentioned above.

The emotional part is classical conditioning.  Classical conditioning isn't about learning behaviors.  It's about creating conditioned emotional responses. 

Another example.  Cesar Milan can shush a dog, stare at it and punish it into standing still while a cat walks by.  Yes, the behavior has changed for the moment.  Does it change over time if this is repeated?  If it does, we can call it punishment and it has worked.  If it doesn't, we can call it either aversive or completely ineffective in changing behavior.

At the same time, the dog MAY change his emotional associations along the way, regardless of what he might have learned or not learned about chasing the cat.  The dog could conclude that when the cat is present, humans punish him.  He could conclude that this particular person, Cesar, is intimidating and will punish him.  He could conclude that cats are dangerous and to stay away from them.  We don't know.

I can't re-write the already well written books on these subjects, but let me leave you with this.  Teaching a behavior and changing a conditioned emotional response are two different things, two different processes.  As a trainer, I have to know which I'm intending and plan my sessions accordingly.  At the same time, it is impossible to teach something without the emotional connections happening with or without our concent or intent.

Of course, you might want to take a look at this:  http://web.mit.edu/newsoffice/2009/successes-0729.html

As I see it, no living thing can be not doing one thing and also not be doing anything else.  Behavior is constant.  We, and our dogs, are constantly "behaving".

So, when your dog wasn't scavenging for food, what was he doing?  We can't say that he simply wasn't scavenging.  He had to be doing something.  He was sniffing, looking around, looking at you, turning away from the food...but something.

More to the point, what happend right after you praised the dog during the training phase.  So...the dog is walking along, sniffs toward some food, you praise the dog...then what?  Somewhere in there, the dog is being rewarded in some way for doing something other than going after the food.  Honestly, I'd stake my life on it.

Here's what I would venture to guess.  Your dog sniffed toward the food, you praised your dog, your dog did something else (turned away from food, looked at you, sniffed the next spot, took a step) and whatever you did after that behavior (continued the walk, petted the dog, gave more enthusiastic praise, or even witheld any punishment if that had been used previously) was rewarding enough to your dog to increase this new behavior (again...looking away or walking on or whatever) which was also incompatable with scavenging.

Thus, what you thought was praise (at the moment your dog went toward the food) became a cue for looking away or moving on.  The real reward was what happened in response to that behavior, which I would love to hear about. 

Additionally, simultaneously...and possibly unintentially...a positive asscociation to you and walking with you was made, of course.

The answer regarding DRI was rejected coming from me as "throwing labels and acronyms." Maybe you can explain it better.

--------------------
Eric Goebelbecker

Gravity does not equal Operant Conditioning (OC) in the thought experiment. Gravity in the example would be the training regime (either Alpha or +R). Just to be clear I'm not saying +R is equal to Operant Conditioning either. Throw a plane in the sky and see if it flies. Throw a dog into a training regime and see if s/he's trained. There is unanimous agreement on gravity in the airplane building community, whereas this is not so within the dog training community. We're not talking about OC in the lab, just to be clear.

You're not saying OC is +R. You just keeping jumping back and forth. First it's +R, then it's not OC in the lab.....

Throw a plane in the sky and see if it flies. It won't if the pilot is not skilled. Throw a dog in a training program and (s)he won't get trained if the trainer is not skilled. If you want to talk about the relative skills of trainers, we're going to need more blog posts.

Therefore, in the example, "Gravity" is NOT a known variable. Both training regimes act differently on the dog and produce output that isn't operated on by a constant, as is the case with gravity and planes.

Are you talking about R+ or OC? If the dog is not trained, it's because the trainer screwed up, not because there's something wrong with OC. Someone who understands OC can train a dog to do quite a bit (not everything.) Someone who also understands CC can do a lot more. Literally thousands of people are doing this. Every day. There's more to it, but "behavioral science" certainly isn't in crisis due to "cracks."

It's a silly example but your usage of it to belittle Mr. Kelley is ironic.

I wasn't belittling anyone. My opening made fun of his argument, specifically his reference to a "war" in training and the fact that's there some sort of debate over operant conditioning. There's a difference. I am not belittling Mr. Kelley. I'm sarcastic. I'm cranky. I even fly off at the handle at times. But I am not in the business of belittling people.

And it still seems valid, because almost every attempt to refute it jumps over to reference to "+R trainers." You want to argue over the efficacy of training with only R+? Find someone else. I'm not trying to make that point.

--------------------
Eric Goebelbecker

This thread has gotten a bit too out of hand for me so I'm going to stop reinforcing anyone's posting.  But I just found a nice resource on learning theory that I thought all of us might enjoy. 

http://scienceofpsychology.com/prospectus/chapter-07.pdf

Hi Ezdog,

I thought I might try to shed some light on your questions:

...I hear what you're saying about punishment having emotional impact on the dog. But then why isn't this the rule: dogs always learn via an emotional context? Why are there TWO different processes that seemingly run a dog? Rewards map to 'thinking' associations and punishments map to emotional responses? Why isn't it one way or the other? It seems like it's a matter of definition and syntax on one hand. However, dogs don't learn to map punishments with behavior, so why would this lead us to believe that they do map rewards with behavior?

From the language I have so far heard, +R claims the 'thinking' association making as the secret sauce in training. But if emotional capacity in the dog is great enough so as to inform it to behave(aggressive towards people) correctly(from the dogs perspective) in the future it seems to contradict the premise that associations are formed via thinking at all. Perhaps not contradict, but suggest that there is 'redundent' technology that the dog is equipped with.

Funnily enough I tweeted about this yesterday (14th):

As you wrote there are two different processes, a dogs behaviour is guided by either emotion or thinking. Two sets of structures in the brain share a very important relationship in canine behaviour.  The limbic system is a complex circuit of neural structures involved in the expression and experience of emotions.   The cerebral cortex on the other hand is involved in various cognitive functions including learning, thinking and problem solving.

The limbic system and the cerebral cortex have an inverse relationship.  When either one is activated, the other system cancels out, or rather gets over-ridden.  If a dog is mentally stimulated and encouraged to think, his cerebral cortex will be activated and learning is effective.  In this state he is less likely to experience intense emotional responses.  Likewise, a dog that experiences intense emotional responses has his limbic system activated and his cerebral cortex inhibited. A dog experiencing an intense emotional response - and I must stress the word intense, is reacting to a given stimulus/event and no longer thinking. This intense emotional frame of mind is not a good place for a dog to be because it can push him/her to react instinctively, and not all instinctive behaviours are helpful to a dog in the human world.

Learning is best accomplished when a dog is in thinking mode, in particular, this is why counter conditioning and desensitisation should always be applied while the dog is sub-threshold, when a dog is over threshold his/her stress level is too high, emotions take over, the dog starts to react, and is unable to learn effectively.

 Give this we can see it's logical that punishment, intimidation, etc., is more likely to activate the limbic system, a dog will either display fear (emotion), or react (emotion). Whereas training through play and positive reinforcement activates the cerebral cortex and the dog will think.

Combine positive reinforcement with operant conditioning and you have a thinking dog that has learned a whole repertoire of behaviours which operate on the environment.

I hope this has in some way shed a little light.

http://www.fun4fido.co.uk - clicker training for dogs

http://twitter.com/fun4fido

Hi, Angela,

It depends on what you mean by "thinking."

My feeling is that what ezdog was getting at is that unless and until you can give evidence that dogs are capable of intellectual thought processes such as logic, language, and mental time travel, it would be best (and much more scientific) to use an emotional model for learning instead of insisting that operant conditioning works because it enables or stimulates dogs to think. And given Keller and Marion Breland's experiences, detailed in "The Misbehavior of Organisms," it's probably more effective when a training paradigm uses a dog's instincts and emotions instead of ignoring them.

Also, your belief that when dogs are in play mode they're also in thinking mode (which implies that play operates solely or even primarily through the neo-cortex) is incorrect. Stuart Brown has written recently that an experiment was done with rats, where the connections between the brain stem and the cortex were severed. This had zero effect on the rats' abilities to play, but it disabled their abilities to mate, form social bonds, and engage in other normal rat behaviors.

So if you believe that play is important to learning in dogs (which it is), and if the process of play is controlled by the brain stem, not the cortex, then play, in and of itself, has nothing to do with "thinking," though it can be a vitally important part of the learning process.

And given that the rat's brain is much smaller and less developed than that of a dog, there's still a big question mark on what you mean by "thinking mode." Do rats think? Do they use logic?

There are evolutionary precursors to logic, such as pattern recognition, that exist in both the animal and human brain. Chess masters are said to win their matches primarily through pattern recognition, not logic. It's also arguable that since rational thought processes are slow, time consuming, and require much more energy than pattern recognition, rational thinking would hold no evolutionary advantage for non-human animals. In nature a thinking animal is a dead animal.

This reminds me of a story about bear expert, Jeb Schenk, who has trained thousands of people in survival skills, including how to deal with meeting a bear in the woods. Then one day Schenk and a hiking buddy actually met a bear. Schenk, who'd told is students to never make direct eye contact, was riveted by the bear's appearance; he couldn't help himself, he stared the animal square in the eyes. He'd also taught thousands of people to never run from a bear. Guess what he did? He ran like hell.

Schenck uses that personal incident to illustrate that even though someone has been well trained, it doesn't mean that he or she will process the information correctly in a real-life situation, unless the learning takes place under real or carefully simulated circumstances. This is why conditioning has a tendency to break down when some dogs encounter a situation that fires up their primal instinctive and emotional circuitry. This doesn't happen with a dog who's trained while his instincts and emotions are aroused and used as part of the training process.

Emotion is important in learning. In fact, there are studies showing that even in humans learning takes place much more quickly and is much longer lasting when a subject's emotions are aroused first.

Neurologist Antonio D'Amasio, who's considered one of the world's experts on emotion, has said that "reason may not be as pure as most of us think it is or wish it were, that emotions and feelings may not be intruders in the bastion of reason. They may be enmeshed in its networks."

He's talking about humans here, not dogs. For dogs there is no "2 + 2 = 4," or even simple propositional thinking like "If I sit then I get a treat." For dogs, learning is an emotional process, not a function of intellect.

LCK

Maybe dogs appear to be better at learning while calm because humans are better at teaching when [they are] calm?

Hi Fun4Dog: In regards to the model you've articulated above, are you saying that the emotion is the base portion of behavior, and that it can be modified, moderated or refined by the cerebral cortex, and therefore the greater the size and/or role of the cerebral cortex the more trainable the animal?

Thanks, Kevin Behan

Hello Lee

I’m happy to discuss this with you but let's keep it constructive, comments such as “language, and mental time travel” are not very constructive I find, as we are both aware I didn’t mention anything about dogs being capable of intellectual thought processes such as language or mental time travel. Hmm, I don’t recall suggesting that we should ignore a dogs instincts and emotions either. So for the sake of good debate lets please keep to the facts of what was written. Ok done. Thank you. :)

I find it interesting that you question whether or not dogs have any cognitive function, I’m not suggesting that dogs are capable of intellectual thought processes. A dog doesn’t need to be capable of intellectual thought processes like humans, in order to gather and process information. The n cortex which is involved in such functions as detailed perception, learning, and intelligence, contributes to about 85% of the c cortex in a canine brain, such a waste don’t you think if a dog can’t think!

In fact there has recently been some very interesting studies in this area. For example Brian Hare’s study:

http://www.time.com/time/magazine/article/0,9171,1921614-1,00.html

There is also the new book by cognitive scientist, Alexandra Horowitz - Inside of a Dog

If you were to give your dog a mentally challenging food dispenser, how do you think he might figure out how to access the food? Do you think he would be using his emotional abilities, or his cognitive abilities? I would put money on him learning how to access the food through a process of trial and error, which basically is what operant conditioning promotes, learning through trial and error. And this is why such treat dispensing toys are mentally challenging.

>>Also, your belief that when dogs are in play mode they're also in thinking
mode (which implies that play operates solely or even primarily through the
neo-cortex) is incorrect.>>

I’m not saying that in play mode dogs function solely through the cortex, simply that mentally challenging games, training, etc activate the cortex. This is very different to what you say I imply. And by play I mean challenging play, training games, where the dog has to gather, process and figure stuff out.

>>So if you believe that play is important to learning in dogs (which it is),
and if the process of play is controlled by the brain stem, not the cortex,
then play, in and of itself, has nothing to do with "thinking," though it can
be a vitally important part of the learning process.>>

I’m not saying that play is controlled by the cortex, I’m saying that playing challenging games activates the cortex. This is different.

You seem to have assumed that I don’t think dogs go through a process of emotional learning. As I wrote in my comment on your post dogs are learning through classical conditioning all the time, or if you prefer we can call it associative learning, and making emotional connections.

My point was not to say that dogs don’t do emotional learning, but to push dogs to learn solely through emotions would be misguided and most certainly not doing justice to the intelligence of dogs.

I was also not saying that dogs shouldn’t use their instincts, after all the option for flight (for example), is a very helpful instinct for survival.

>>Schenck uses that personal incident to illustrate that even though someone
has been well trained, it doesn't mean that he or she will process the
information correctly in a real-life situation, unless the learning takes
place under real or carefully simulated circumstances.>>

I would say that this is down to ineffective training, training can be broken down in to the following steps, if all these steps are followed in sequence then dogs have no problem processing information correctly in real-life situations. For the efficacy of reward based operant conditioning you only have to look at assistance dogs. The issue here is not whether operant conditioning fails, the issue is whether the training was effective in the first place.

   1. Learning new behaviours
   2. Generalisation of learned behaviours
   3. Discrimination of cues
   4. Reliability of learned behaviours
   5. Fluency

>>This is why conditioning has a tendency to break down when some dogs encounter a
situation that fires up their primal instinctive and emotional circuitry.
This doesn't happen with a dog who's trained while his instincts and emotions
are aroused and used as part of the training process.>>

I would say that this generally has more to do with genetics (bad breeding), coupled with poor socialisation, and unfortunate negative experiences. Which likely accounts for a large percentage of the domestic canine population. Such dogs will often encounter situations that as you put it fire up their instinctive (fear based) reactions, but when this happens the limbic has been activated and the cortex is over-ridden, hence the importance of being able to shift the dog in to thinking and not reacting in these situations.

>>He's talking about humans here, not dogs. For dogs there is no "2 + 2 = 4,"
or even simple propositional thinking like "If I sit then I get a treat." For
dogs, learning is an emotional process,/ not/ a function of intellect.>>

Not sure why you’re bringing math in to a discussion on learning theory, and operant conditioning. And again I haven’t said or implied that “If I sit then I get a treat” has anything to do with propositional thinking.

Dogs learn primarily through a process of association. They learn very early in life that their behaviour causes consequences they like and those they don’t like. In this respect, the behaviour of dogs will reflect what they’ve experienced. So a dog learns through association (primarily) and experience, that if he sits then a pleasant consequence will occur, such as going for a walk.

Keeping it really simple, dogs want stuff, we show them how to get it.

The reason for my comment was simply to try and answer Ezdog’s questions as he/she posed a questions about:

>>Rewards map to 'thinking' associations and punishments map to emotional responses?
And
Why are there TWO different processes that seemingly run a dog?>>

So I was trying to explain that positive reinforcement training is more likely to activate the cortex, this is not to say that the dog is not also learning emotionally, but rather that the dog is not likely to get highly stressed, go over threshold, and possibly react instinctively.

On the other hand if punitive training methods are used, it is likely to activate the limbic, get the dog highly stressed, potentially push him/her over threshold, and possibly react instinctively.

Given that I work everyday with dogs (and their humans), I know what frame of mind I would advocate activating for training and behaviour modification.

http://www.fun4fido.co.uk clicker training for dogs

http://twitter.com/fun4fido

What is a dog thinking?

Kevin Behan

We seem to be viewing the subject matter from such opposite perspectives that we’ll quite probably never agree. For instance, I would say the idea that dogs learn through trial and error is erroneous, and that when a dog works to get treats out of a food dispenser, he's engaging his hunting instincts, not any higher mental faculties he may or may not have. (I would also dispute your assertion that the 85% of a dog's cortex is dedicated to anything that could be considered thought: it primarily includes things like pattern recognition, sensory input, etc., and what cognitive scientists would call "down and dirty" neural circuits, having nothing to do with "thinking," per se.)

Despite the slim possibility we'll ever agree, let me try to get down to what may be the most bare-bones element of how dogs supposedly learn: cause and effect, and why this bit of folklore, if you will, may be partially responsible for your belief that dogs are capable of certain types of thought processes unsupported by the size and shapes of their brains, not to mention any real scientific data (please read Rational Animals? by Susan Hurley and Matthew Nudds).

First of all, there are three basic forms of consciousness, or more properly, awareness:

1) physical (sensori-motor, body awareness, etc.)

2) emotional (limbic system), and

3) mental/intellectual (neo-cortex/frontal lobes).

Unless and until you can completely rule out all possibility that behavioral phenomena can be explained solely from the perspective of the two lower functions, you can’t (or shouldn’t) automatically assume the higher.

What does this have to do with cause and effect?

It seems to me that the law of cause and effect is an anthropomorphic view of learning, an essentially human mental construct that doesn’t apply or make “sense” to the animal mind. In order to truly understand how dogs think we have to look at their behavior from their point of view, not ours: we have to dogthropomorphize ourselves.

So even though cause and effect are clearly at play when you look at canine learning from the outside-in, they’re not something a dog can have any awareness of. As a result any attempt to explain learning via this model will always lead to a substantial error in logic and understanding. Put another way, while cause and effect are things that happen to the dog, if they don’t actually play any part in how he experiences what’s happening to him, they can’t rightly be said to play any real part in how he learns.

If we look at the following statement – in which I’m using personification to describe the type of awareness a dog would have to have in order to learn through cause and effect – we see that cause and effect are sequential and chronological:

“When I do such-and-such a behavior in such-and-such a context/situation, then something good/bad happens.”

Yet if dogs live totally in the moment, which I think they do, then they would have no sense of linear time. As a result they wouldn’t be able to perceive one thing happening before or after another, let alone in consequence of another. In a dog’s mind events wouldn’t happen sequentially; they’d all happen in the now moment, so a dog couldn’t rightfully be said to learn through cause and effect: without a sense of linear time, the sequential connection couldn’t reverberate or be able to attain any purchase within his consciousness.

But let’s say you disagree; you believe that the dog does experience cause and effect, but perhaps as part of an expanded now moment, the way athletes experience time differently when they’re “in the zone.” Even so, it seems to me that there’s still a second error, which is that cause and effect are both mental constructs, thought-based "phenomena;" they’re neither visceral nor concrete. In other words, you can’t feel, taste, sniff, bite, chase, chew, or pee on a cause or an effect. So how could a dog experience them?

You might say that a dog certainly can experience things like a tasty treat or a correction from a choke chain, and do so on a very real and very physical level. I would say, yes, but a treat or a correction are not in and of themselves a cause or an effect; it would depend on the context. And absent any mental connection linking them, they’re still abstract and thought-based, no matter how they might manifest themselves in physical form.

On the other hand, if the principle of cause and effect can be explained as part of an energetic dynamic, where the dog is only required to experience shifts in his own energy states and adjust his behavior accordingly, then you will have succeed in obeying Ockham’s razor, Morgan’s canon, and the rule against anthropomorphism. But you’ll also have deconstructed one of the primary principles of operant conditioning, that a dog “operates on his environment.” He doesn’t. Without a sense of self, how could he? (I sense another argument looming…) He can only "operate" on his own behavioral choices by feeling the shifts in energy that any outside stimuli might cause to stir within himself.

It may seem as if I’ve overcomplicated things. I’ve actually done the opposite. Some might argue that I’ve engaged in a reductio ad absurdum argument (which, by the true definition, I may have). Be that as it may, science tells us that the simplest explanation, as long as it satisfactorily explains all aspects of a natural phenomenon, is preferred to one that’s more complicated, particularly if those complications are unsupported by any hard evidence. My explanation is simple, and it’s capable of explaining all aspects of a dog’s behavior. Yours requires that dogs “think,” for which there’s no hard data, just guesses and +R folklore. And if you examine the idea that dogs "think" more closely, you’ll see it's too slow and complicated a process to serve any adaptive purpose in animals. Even in humans, most of what shapes our behaviors is done through “down and dirty” cognitive circuits, not conscious thought. This means that not only is there no thinking going on inside a dog’s brain, there’s very little of it in a human’s!

Best wishes,

LCK

Has anyone seen or read Bob Bailey's response to my article?

LCK

Do you have a link?

--------------------

Eric Goebelbecker

Hi, Eric,

There's no link that I know of. I was told that Mr. Bailey sent out a newsletter with my article and his response to it.

LCK

Behavior A - scarfing items up from the street DECREASED - became less likely - was punished.
Behavior B - the behavior of passing by undesirable items without picking up or ingesting (leave-it) INCREASED - became more likely - was reinforced.

Did you accomplish teaching an incompatible behavior?  He can't take it and leave it at the same time, so, yes.

Scavenging was reduced in frequency due to the application of an interruptor, albeit a pleasant one: praise.   Each time the dog approached something tasty it was interrupted by your happy response.  You took notice of what the dog was about to do and responded predictably with happy sounds.  In effect, your response drew the dog's attention away from the item each time, INCREASING attention away from a potentially ingestible item and DECREASING inclination to make contact with such items when he came upon them in future trials.  

Which is what any good "positive" trainer attempts to accomplish: the least aversive method necessary to increase or decrease the desired goal behavior.   The point we all appear to agree on is that it isn't necessary to inflict physical or negative, socially "dominant" energy on an animal to affect a change in behavior.  If "correcting" is defined as replacing with a more desireable outcome, then yes, you "corrected" with praise. 

Point: increase or decrease in behavior is the measure - not "happy" vs "angry" or "good" vs "bad"; lots of people yell at dogs and get an increase in the very behavior they intended to decrease.  

The scavenging (picking up/carrying/ingesting) behavior was replaced by non-picking up/eating/carrying behavior.  There was no classical/negative emotional fall-out - the dog showed no fear, avoidance or defensive response, he simply no longer picked stuff off the ground. 

The normal sniffing and investigating the environment on the walk didn't change because there was no coersive fall-out.  The only thing that became less likely was the very behavior you tried to affect. 

Happy dog.  Happy owner.  Perfect training.

Carol Byrnes

Wow, I have read through as many of the posts in this link as I possibly could and it left my head spinning. Lots of knowledge here! Imagine if dogs could read, what would they think lol! I could imagine them all lying down with there furry paws over there snouts, wondering why us humans have to make things soooo difficult!!

LCK, your final sentence in this post reallllly got me thinking!  Your final sentence:

He's talking about humans here, not dogs. For dogs there is no "2 + 2 = 4," or even simple propositional thinking like "If I sit then I get a treat." For dogs, learning is an emotional process, not a function of intellect.

LCK

I'm no scientist, I teach people how to teach there dogs. I have heard that for Learning to take place that you should activate the pleasure center of the brain. So if learning is an emotional process for dogs, then that would make Positive Reinforcement the best way for dogs to learn. Correct?

Tracey Harrison

Tracey Harrison