I think treat training is cruel. - Page 9

Prager,
Pavlov used respondent or classical conditioning which is about behaviors that are controlled only by antecedent stimuli, without any mention of consequences. So there are two different kinds of behavior. Behaviors that are controlled by consequences are sometimes referred to as "voluntary" behaviors or operant behaviors because they operate on the environment to generate consequences and are in turn, controlled by those consequences. Respondent behaviors are reflexive or elicited by prior stimuli quite apart from the consequences of the behavior. An example of respondent conditioning would be a person is presented with the stimulus of the sound of a bell, paired with the stimulus of ice-cream in the mouth resulting in the response of salivation. The stimulus of the bell elicits the response of salivation. An example of operant conditioning would be a child hears the stimulus of the sound of a bell, volunteers the response of running to the street where the ice cream truck bell rung, and gets the reinforcer of ice cream in his mouth. They are technically, both conditioning, but have significant differences. Operant learning is much more prevalent in training dogs than respondent conditioning. It started to become fairly well know in the 1950's and progressively caught on. I believe way too many trainers don't fully understand principles of operant learning or how to apply them to training. This idea of the dog working for the trainer is old school and somewhat bogus. You can get reliable results with your approach, but you won't get the intensity and the dog is likely to come down in drive much easier. And those issues are not just related to sport. You want a detection dog to remain in high drive while working. You want a police down to fight intensely while working. All bite work is really based on learning. The dog works for what is most reinforcing. You can never prove the dog is working for the trainer, only speculate.  The whole issue of GSDs herding has to do with selection and genetics.  The dog either has the genetics for genetic obedience or he doesn't.  These genetics have largely been lost to the breed.  And the old herding masters did not approve of the emphasis on prey drive for sport any more than they approved of the high lines for beauty.

Apple, love your post. Excellent explanation.
I can certainly relate to your remark concerning herding dogs and genetic obedience.
I've always maintained a dog can't be trained the herd, only trained for control...ie, they either have the genetics or they don't.

And I have dogs with, and producing, both types.

You are fortunate to have dogs with and producing dogs with true herding instincts. IMO, they are rare. I believe you have said you breed for a more balanced dog not so much toward extreme prey drive. The old herding breeders were really against extreme prey drive. That is why they thought schH as well as show line breeding would all but wipe out the herding lines.

Apple, Thank you for sharing your interesting insight into the various learning models. Although classical conditioning may not be as prevalent in training, it is deeply seated and often an adjunct within the operant model.

I think any behavioral sequence is likely to include both classical and operant conditioning. Maybe you could give some examples of classical conditioning in training. I have a little trouble coming up with examples. I can think of praise, where praise becomes a conditioned stimulus. Maybe the sight of a prong collar creating avoidance of having it put on. It gets kind of complex and that is why I believe trainers should at least know the basics of operant learning.

How about my husky sled dogs reacting to the sound of their gangline snaps clinking when I took it out of the box.
No matter how quietly I tried to be, they would hear it. They lived up on the terrace above the house, and behind. My room where I kept the equipment was below them....at least 250' and up above. They were in big yards, two dogs to a on, they would all start barking and jumping around excitedly.
I put there harnesses over the porch railing n order of the dogs in the team. A friend usually would open the gate of the pair I was ready for, heyd bolt out at a full run, come to the house and each go to their harness and wait. One dog used to try to put his head up into the harness while it was draped over the rail, four feet above the ground.

By, all my dogs, gsd, that I have ever used a prong on, reacted same as the sled dogs when I would pick the prong up...never had one go into avoidance.

I think classical conditioning in working with dogs has more to do with changing/manipulating a dog's emotional state as opposed to teaching a new behavior. Changing an emotional state can then facilitate changing a behavior. So with the example of a dog showing avoidance when the prong collar is brought out, repeatedly pairing the sight and sound of the prong with food would be a form of classical conditioning.

But if the dog loves the activity related to wearing the prong, food is not needed.
Same way my husky sled dogs became excited at the slight clinking of the gangline...anticipation of being hooked up and going out with the team.
The prong for my dogs is prelude to going out and doing obedience or tracking, or protection work...causes change from laying around to excited anticipation!

Apple, praise is excellent example of classical conditioning having roots in the operant model.  Can you think of a more powerful tool for a working dog than the praise from his handler?   It's the one thing you will always have with you and can control how much, how little, how often and how long it's given.  My favorite tool in the box for training police dogs was praise.  Rather I was conducting building searches, areas searches, narc or explosive searches, tracking the criminal or watching him down and indicate on a gun that was just used in an armed robbery, hearing that 'good boy' always made that tail wag.  It wagged just as fast as when I was teaching them how to down on articles on the track.  I would place articles along the track and whenever they came upon on them I would command down and then offer praise.  They loved finding articles and we had great success over our careers finding guns, knives, etc discarded by the criminals.

As you stated, praise is a conditioned stimulus for the dog.   Just like food, it can be used to reinforce behaviors that we wish to strengthen, like downing on articles while tracking.   The operant model being based upon the principles of Thorndike's Law of Effect, has components to help strengthen or weaken a behavior.   I always think of operant conditioning of behaviors having consequences, and it's what we as trainers do to manipulate the consequences that will effect future behaviors.  If I understand it correctly, anytime we want to increase the likely hood of a behavior happening again, we should reinforce that behavior, be it either positive reinforcement or negative reinforcement.  Under positive reinforcement, the behavior, something the dog did, the consequence was something good happened.  With negative reinforcement, the behavior, something the dog did, the consequence was something bad stopped happening….the environment subjected the dog to something bad (an aversive) and he did something to make it go away.  

You mentioned avoidance earlier.  We can use avoidance as the end product of negative reinforcement to establish a behavior that we can later reinforce in a positive manner, like praise.  Lets take our little puppy and decide we wish to teach him to sit using negative reinforcement.   So we hold our puppy by the collar and gently push down on his rear.  As soon as that butt hits the floor we stop that pushing.  Under the operant model, the behavior of the puppy sitting, the consequence was something bad stopped happening.  We stopped pushing.   Thanks to the wonders of escape and avoidance, we wont always have to push on that puppy.  Soon, as we start pushing, the puppy will go ahead and sit to escape from that pushing.  Lets throw in a command and soon that puppy will sit to avoid being pushed on at all.  So now we have the operant model strengthen the behavior through negative reinforcement.  Lets strengthen it even more by using some of that wondrous praise when he sits.   Of course we could have used positive reinforcement to teach that sit.  I just wanted to use the negative reinforcement example to dispel some of the concerns about it being all ecollars and prongs. 

We used our voice to ‘praise’ and stregthen a behavior, why not use our voice to weaken a behavior.   Looking at the operant model, if we wish to weaken a behavior or decrease its frequency, we need to punish it.  It too comes in positive and negative and influences behavior based upon the consequence.   Back to the behavior and consequence thing.   With positive punishment, the behavior, something the dog did, the consequence was something bad happened.  I'll use an extreme example, but I am gonna pair a neutral stimulus so that later I can elicit an unconditioned response.  Pavlov would be happy.  Lets take advantage of the dog’s sensitivity to having something suddenly clamp down on their neck.  They feel vulnerable around the neck and even have thicker hair to help protect it.  Good trainers know that sensitive area and sometimes call it a startle reflex.  

It’s unfortunate that I ever have to utilize positive punishment, and I tried to minimize it as much as possible.  By far, bitework is where I have had to resort to the operant component of positive punishment the most.  Teaching a dog who has great instinct to perform a behavior, to abandoned that behavior can be a feat.  I didn’t have much luck with extinction or negative punishment (cover that later).  I don’t teach the out, as in hold and bark, but I teach a return to the heel position on my command to get the dog to release the decoy and later the suspect.  I begin to shape this behavior away from the decoy.  I explain to the dog that coming to my side, brings great reward.  There’s that behavior and consequence thing of operant conditioning.  Coming to my side makes good things happen, a toy, a pat on the head, praise, maybe even a French fry if I have any leftover.  And they do a good job coming to heel for all that cool stuff, away from the decoy that is.   That old French fry just doest have the same appeal as Mr. Decoy.  Fortunately, operate conditioning provides for another componet that helps me punish the dog for not coming back.  His behavior of not coming back results in a consequence of something bad happening.  Here comes the startle reflex.  Should the dog disobey my command, he will be punished swiftly and effectively by a strong yank on his collar.  At the same time, I will invoke the use of a neutral stimulus, using a loud AAAYYYYY!  I try to make it a very significant event to minimize the repetitions needed.  This sometimes has to be repeated several times.  Thanks to classical conditioning, I am soon able to elicit the same response without having to use the physical manipulation.  If the dog, fails to comply I can now yell aayyyyy creating the same response without the collar. 

I have also made use of negative punishment.   Mainly on my recall training.  This is where I send the dog after the criminal and then call him back to me.  I only used it once on a bad guy, and then, instead of giving the recall command I panicked and blurted out the down command.  The dog listened, which was a good thing cause the criminal he was after just dove off a cliff.  I did use the recall command to save a couple officers.  A couple others times, I didn’t get the command out fast enough.  It happens. 

Negative punishment.  The behavior, something the dog did, the consequence was something good stopped happening.  When teaching the recall we apply the four common principles governing proximity and distance.  I think everyone knows these, but I’ll describe it anyway.  Upon the command - 1) the closer the dog is to the handler, the more likely he is going to recall.  2)  the farther away the dog is from the handler the less likely he will recall. 3)  The closer the dog is to the decoy, the less likely he will recall.  4)  The farther away from the decoy, the more likely he will recall.  For years I used positive punishment for teaching the recall.  Behavior, the dog not coming back.  Consequence, something bad happened to him.  This was mainly with the use of a long line.  Later, I started using negative punishment.  The decoy was always in a position where he could prevent the dog from getting the bite, ie closing a gate or door.  This is important.  Should the dog not recall, decoy closed gate.  I go get dog and put him up.   Behavior, not recalling.  Consequence, something good stopped.  He goes back to patrol car and gets to watch all the other dogs have fun with the decoy.  Bandura would have been proud.  On a side note, if the dog did recall, his behavior was reinforced by allowing him to go back and bite the decoy.  

Almost forgot extinction.  I use extinction when a dog alerts on a non targeted odor during narc or explosive detection.  Instead of punishing (positive punishment) the dog, I do nothing but wait.  Punishing only weakens the behavior, in it’s purity, extinction is what eliminates a behavior.   There’s also problems with some dogs looking towards the punishing agent (me the handler) for guidance. 

There’s a few examples of classical conditioning being seated in the operant model.   And don’t forget that bridging stimulus thing.   Gotta go through Pavlov if we want that marker to work. 

Most trainers make great use of operant conditioning and have been for decades.  I agree that many people don’t fully understand it, as it’s easy to get hooked up on the commercialization of the bridging stimulus like the clicker and think that operant conditioning is all about the click and treat.   That’s just a small part of the model.  It has its place just like reinforcement schedules, chaining, shaping, reinforcement intervals, spontanous recovery, etc.

emory, you make a good point. but drawing from my youth and growing up in a very rural area with much hunting, rabbits, birds, and racoons,

the most powerful tool for dogs successful in these areas was what they were born with.

beagles ran rabbits or they didn't and they did not get far to show it. same for bird dogs and night hunting dogs.

have no experience on LE dogs, so no comment.

just that these dogs I am talking about were owned by farmers/factory workers with little patience for error. and were bred from dogs very strong for the task and the good ones showed it from a very young age.