Positive Re-enforcement Only Techniques

[pgm]

Sidoney: I don’t see how one can say you can have theory or method without each other. At its most simple, a theory is an organised set of ideas that attempts to explain or predict something. Method, at its simplest, is an ordered way of doing something.

PGM: No child requires a theory in order to master langauge. One can speak and write like a master (Shakespeare) without the slighest knowledge of theory. Knowing how language works comes from experience, not theory. Experience is enough to know how to go on in unexpected contexts – children are some of the most inventive and creative speakers. No theory is required for this.

Sidoney: For some reason, pgm seems to have a particular aversion to behaviourism and/or behaviourists. Thankfully others are more open minded.

PGM: Others have a more open mind? Certainly not behaviorists themselves, who have consistently ignored the results and work of actual animal trainers whose knowledge and experience have been passed down through generations. Behaviorism was developed in the laboratory in isolation from actual working trainers who were basically ignored and/or dismissed.

Sidoney: To throw out behaviourism in total is to throw out the baby with the bathwater.

PGM: To throw out the knowledge and experience gained over generations by working trainers is precisely the point where behaviorism began.

And besides, the baby is the dog, not the theory.

Sidoney: I categorically reject any idea that one can “know” what goes on in any other creature’s mind, and suggest that one can only guess, and that with dogs, the guesses are less accurate than with people. Even with language, the most complex communication tool, what one person says may be understood in a different way by the person who hears – due to many different factors, including characteristics of language itself.

PGM: This is a skeptics argument, which I know in advance cannot be defeated. I cannot prove the skeptic wrong. All I will say, is one: because we can be wrong about something doesn’t mean that everything is just ‘guessing’. Secondly, one lives in a rather desperate world if all one can do is guess what another person is thinking or saying.

Sidoney: While it is common for people to infer moral behaviour in dogs (“he knows he did the wrong thing/he feels guilty, just look at him”), this sets up a dangerous situation for the dogs themselves.

PGM: To give a child language is a dangerous gift, for it can cause grief and untold damage. There is a story of an autistic child who was taught to speak, this child had up until this point never shown any emotional attachment to any other human being. The child was taught the meaning of ‘hug me’ - the child then sought out his brother, also an autistic child, and said ‘hug me’ – his brother ignored him, and he broke broke down and wept for the first time in his life.

Some people would see this as a reason not to teach such children – they are after all not unhappy in their autism – I am not one of those people. Striving for excellence in any field of endeavour brings with it the knowledge of terror and grief. You cannot truly possess the knowledge of beauty without also the knowledge of terror – you cannot have one without the other.

Sidoney: They become subjected to similar kinds of (ineffective) procedures of justice to those that are inflicted upon humans. Humans are supposed to know “right” from “wrong” and are punished for doing “wrong”. It is no secret that this is ineffective.

PGM: It depends where you look for your evidence. If you are looking at the people whose upbringing (for whatever reasons) have failed them, then you will find plently of evidence for your thesis. The vast majority of people however grow up to be law abiding citizens, who have strong sense of right and wrong and do not need to be punished or reminded of their responsibilities.

Furthermore, I do not, and have not, ever punished my dog. This is where the vocabulary of behaviorism is both ugly and wholly inadequate to describe what is really going on.

Sidoney: I should mention, this kind of high level of motivation is needed for any dog that is asked to do anything that requires particular intensity or duration – see Sam’s post above re the scent dogs – these dogs are required to work for an extended time and may go some long time without finding the scent that brings their reward – they need to be very motivated to sustain that.

PGM: Actually I have come across a number of sniffer dogs (beagles) at airports. What can I say? These dogs looked no more mature than the dogs I see at agility. Not suprisingly they are constantly being fed treats by their handlers for almost every command. I have my doubts that these dogs would work for more than 10mins without a treat. They do not impress me as an example of a trained dog.

I have heard stories of tracking dogs who will follow a scent with their poor handler in tow for twenty hours or more. They don’t require anything more than the work itself to motivate. Bordercollies that are trained for sheepwork are not trained with the use of treats or play toys in order to motivate them. The work itself is more than enough. A well trained sheepdog can be sent out on its own to bring back the sheep to the handlers feet. They do not require even the handlers presence to do this, let alone to be treated every couple of mnutes.

[poodlefan]

PQM:

I was watching people train agility the other day, and whilst I would say that the dogs were quite responsive to their handlers (they were of course using treats constantly) and looked quite 'happy', their happiness to my eyes looked frivilous and immature.

why using treats constantly "of course"?

Many of the agility dogs trained at our club aren't trained on treats AT ALL. Other motivators are used eg. balls, toys, praise.

My guess would be that you were watching a foundation level class. When the skill is being learned, more positive reinforcement is used "of course".

As the dog's skills and enthusiasm increase, agility becomes its own reward. ANKC competition prohibits the carrying or use of any motivators in the ring. If you can't motivate your dog to perform without constant 'bribing' with treats, there's no way you can compete. A successful agility team has a strong and constructive relationship between dog and handler that is based on considerably more than food. Perhaps you might have gained a better appreciation of that by observing a competition rather than a training class.

I'd be interested your views on how a dog performing an agility obstacles happiness differs substantially from that of a well trained gun dog returning with a bird or a protection dog bringing down a person.

Its all training PQM - its merely the objective that might appear frivolous or immature in the eyes of the beholder.

Edited February 14, 2005 by poodlefan

[pgm]

Poodlefan: Its all training PQM - its merely the objective that might appear frivolous or immature in the eyes of the beholder.

PGM: sniffer dogs working at airports are not to my mind engaged in frivilous work. It is their manner and behavior and the constant reinforcemnt by way of treats that strikes me as being immature. I thought the objective of training was to train the dog to work independently of external rewards? It does not appear to be the case with these dogs however.

[Guest Tess32]

I don't think that's a universal objective. I don't expect my dog to raise his paw for eternity with no reward ever again. Some behaviours are not intrinsically rewarding for the dog and some reinforcement every now and then is not something I want to phase out.

Nat

[poodlefan]

PQM: So an explosives detection dog that works its butt off for a tug game with a towel or a throw tennis ball is somehow frivolously trained too?

Surely knowing what motivates a dog and using that motivation can never be frivolous. For what its worth, I have never observed an AQIS beagle being fed whilst working at an airport. Perhaps you observed a trainee?

All dogs work for reward, the better trained the dog, the less is the need to constantly reinforce with rewards - but to expect the dog to work with NO prospect of reward is to ask something of our dog that we would not do ourselves IMHO. Lessening the frequency of rewards/improving the standard of response that generates a reward is just another phase of the training.

I too, don't agree with using 'bribery' constantly to obtain desired responses from dogs. However, be it food, toys, praise or what ever, I'll use what ever works best.

[Kavik]

Dog always work for external rewards! As they get better, the rewards just take longer getting there, or secondary rewards are used more often.

I have done a scent detection course - really good dogs are amazing to watch! Super motivated, and the active reward dogs super keen on their toy (and will do anything and destroy anything to get to it LOL!) I love to see dogs which are happy while they are working.

I also happen to have a dog that sees agility as its own reward (although we no longer go due to dog aggression issues). She waits in a sit for my signal, you can see how eager she is to do the course in the way she is waiting. We do use food, mainly for new exercises and to teach her to slow down for contacts. Interestingly, while I though calling her off an obstacle (say the 3 jumps mentioned before) would be difficult, she responds very well and fast to my voice commands, and does not make up her own course (have seen many that do LOL!)

[pgm]

PoodleFan: So an explosives detection dog that works its butt off for a tug game with a towel or a throw tennis ball is somehow frivolously trained too?

PGM: in short, yes. A well trained bordercollie is not working for a treat, or a game of tug - the work itself is the reward.

We are not talking about dogs that are not bred for this kind of work. We are talking about dogs that are specifically bred for this type of work, just as bordercollies are bred for sheepwork. If you cannot train a dog that has been specifically bred for a specific job to work without external rewards, then yes, I don't think much of the training methods.

[aatainc]

in the case of sniffer dogs, the objective of the training is NOT to train the dog to work independently of external rewards. The objective of the training is to get the dogs to sniff out illegal quarantine material as reliably as possible, period. - in other words, the objective is the BEHAVIOUR not the moral high ground of having a dog that works without food. If AQIS have found this reliability to be better with food than without "External rewards" then who cares? I mean really, when the dogs are extremely reliable, I think it takes a particular kind of arrogance to say that the dogs are immature and not well trained enough simply because they continue to use food. Go get a beagle, train it for quarantine, and get it to work for 4 hours a day for 10 years - and don't give it any external rewards. When it is as consistent, contented and reliable as an aqis beagle, then you have a point.

Edited February 14, 2005 by aatainc

[pgm]

aatainic: I mean really, when the dogs are extremely reliable

PGM: you have a very different idea of reliability than I do.

[poodlefan]

PQM:

We are not talking about dogs that are not bred for this kind of work. We are talking about dogs that are specifically bred for this type of work, just as bordercollies are bred for sheepwork. If you cannot train a dog that has been specifically bred for a specific job to work without external rewards, then yes, I don't think much of the training methods.

PQM you have narrowed the context of training considerably here. Essentially you've limited discussion to the type of 'work' which is essentially instinctive. Herding is, after all, a modified and controlled form of hunting. Let most sheep trained BC run loose long enough and you'll witness this.

I would argue to train a dog to do anything that is not within its natural instincts will require some kind of reward (edited to add: or aversive)

I didn't need to train my dogs to chase a lure - they do it natually and the chase is its own reward. To require a dog to do other than obey its instincts is going to require some kind of motivation - either positive or negative.

Agility, gun dog work, scent work, detection work all utilize natural instincts but most are not (at least initially) self rewarding behaviours. Herding is!

Edited February 14, 2005 by poodlefan

[aatainc]

would you prefer accuracy of detection?

I don't know the stats, I"m only presuming, but when a beagle comes up to me and sits because I HAD a banana peel in that bag a day ago - and he sniffed it out - that's pretty good to me. (And don't go saying that he shouldn't have alerted because it wasn't there any more. The dogs have to be trained to alert ot even the tiniest amount of scent.)

[pgm]

Poodlefan: Herding is, after all, a modified and controlled form of hunting.

PGM: what, you don't think sniffing for signs of prey is part of the hunting instinct?

Poodlefan: Agility, gun dog work, scent work, detection work all utilize natural instincts but most are not (at least initially) self rewarding behaviours.

PGM: I didn't claim that there were 'initially' self-rewarding. I just thought it was part of good training to develop the self-rewarding aspect of work. But apparently I am wrong.

[poodlefan]

You may not be wrong PQM but personally I don't agree with you. Furthermore, I can think of no work activity for which a dog receives no natural or handler provided reward. Perhaps livestock guarding (however the dog is rewarded in the successful protection of its "pack")? Even guide dogs get rewards.

Sniffing for signs of prey is considerably different to tracking a SPECIFIC handler selected target or to identifying a scent that is not part of a dog's natural diet and won't be killed and consumed.

Wanna see a Beagle work its butt off with no external reward? Put it in a paddock with rabbit scent and it will take off and follow that scent for HOURS... ask it to find a banana peel and its motivation is not the same.

Remove the chance of the kill and you have to provide an incentive to work...

Not natural behaviour? I think you need a reward or aversive IMHO.

Edited February 14, 2005 by poodlefan

[pgm]

PoodleFan: I think you need a reward or aversive IMHO.

PGM: to train yes. But if I was to claim that my dog was trained as opposed to being in the process of training, and you then witnessed me giving the dog corrections on a regular basis, I suspect that you might dispute my claim that the dog was trained. The same goes for external rewards.

Poodlefan: Remove the chance of the kill and you have to provide an incentive to work...

PGM: the trainers job, not to mention skill, is in being able to show the dog the reward that is internal to the work itself. Constant reliance on external reward (such as treats and games of tug) prevents the dog from coming to this awareness.

Personally Poodlefan I don't really know what you are arguing about, given that you stated in your first post:

"My guess would be that you were watching a foundation level class. When the skill is being learned, more positive reinforcement is used "of course".

As the dog's skills and enthusiasm increase, agility becomes its own reward."

So this applies to agility, but not scent work?

[sidoney]

pgm, I can't be bothered going point by point through what you are saying. I don't find your arguments very relevant.

PGM: the trainers job, not to mention skill, is in being able to show the dog the reward that is internal to the work itself. Constant reliance on external reward (such as treats and games of tug) prevents the dog from coming to this awareness.

Since you seem to know so much about it, certainly more than professional scent dog trainers, how about share your wisdom and let us how you are going to do this in the context of a difficult and sustained task that is not tapping into instinctive drives of the dog? Instead of just saying, "it should be so", tell us how it can be so.

As the dog's skills and enthusiasm increase, agility becomes its own reward

BTW this is not the case for every dog, and as I said before, to get that extra effort, you have to (in most cases anyway) give them something back for it. My dogs will look for game for hours if they get the chance, but they don't have that same desire for agility, even though they are fast, accurate, and enthusiastic, oh and win a fair bit too, even though I don't trial much.

Edited February 14, 2005 by sidoney

[pgm]

Sidoney: let us how you are going to do this in the context of a difficult and sustained task that is not tapping into instinctive drives of the dog?

PGM: so you are now claiming that distinguishing scent is not instinctive for dogs?

Sidoney: As the dog's skills and enthusiasm increase, agility becomes its own reward - BTW this is not the case for every dog, and as I said before, to get that extra effort, you have to (in most cases anyway) give them something back for it.

PGM: It wasn't I who made that statement.

Sidoney: pgm, I can't be bothered going point by point through what you are saying. I don't find your arguments very relevant.

PGM: I realise that.

[sidoney]

It wasn't I who made that statement.

No you didn't, I am aware of that, that part wasn't meant to look as if it was a response to you - my error.

so you are now claiming that distinguishing scent is not instinctive for dogs?

I've not met a dog yet that has an instinctive drive to distinguish between bananas and bras - especially not for periods of time.

Edited February 14, 2005 by sidoney

[poodlefan]

PQM:

As the dog's skills and enthusiasm increase, agility becomes its own reward."

So this applies to agility, but not scent work?

PQM, I suspect it does where the dog is not choosing the scent it follows. Sidoney is also right - some dogs enjoy agility more than others. The dogs who enjoy it the most are those ones who will run for anyone - some dogs only run for their own handlers. The self rewarding dogs are also sometimes the dogs who can be the most difficult to control...

PQM can you ponder this for me? If my dog runs an agility course enthusiastically and accurately, and then at the end of the course, looks to me for a reaction should I NOT reward it because "it should just enjoy the run"? EVERY time a dog performs a task it may learn from it. Training is for life - reinforcing can be done during the performance of the task but the dog is still learning things even if all it is learning is that perfomance must be better or more sustained to earn rewards.

I don't know of any skill that isn't practiced to improve and sustain perfomance. To not encourage sustained performance is to risk a lowering of how a task is performed.

What is the training issue with rewarding a dog for a job well done? Would you expect a person to give you of their best because "they should just enjoy their work"? I'm not just interested in performance but in optimum performance and for that you have to give feedback.

Any sentient being will work harder at a task if they believe they will gain a benefit from it. Dogs are no different IMHO. When you talk about rewards, you appear to focus only on food or toys. Isn't praise a reward too? Can't speak for you PQM but I get a buzz when I'm praised for a job well done and it certainly motivates me to maintain or even improve my performance. Isn't that every leader's job- to find out what motivates and to encourage and reward optimum performance. Hey, even a "good dog Fido" is a reward.

I've seen agility and obedience dogs whose handlers adopt the approach you appear to be suggesting - suffice it to say, these are not the dogs who give of their best. Even worse are the dogs who get blamed for handling errors - they don't tend to be winners.

Edited February 14, 2005 by poodlefan

[aatainc]

I think pgm's point is that there is something instinctively superior to a dog that doesn't get external rewards. I used to think this too. Then I started training dogs for real life situations.

[pgm]

Poodlefan: What is the training issue with rewarding a dog for a job well done?

PGM: there is no training issue. I have not at any stage suggested that one should not reward their dog in training. But I do not see the point in rewarding a dog in training as a matter of motivating the dog - I see it as a matter of developing the dogs understanding. As the understanding increases, the rewards (in terms of external rewards) decrease and become less and less necessary.

Poodlefan: If my dog runs an agility course enthusiastically and accurately, and then at the end of the course, looks to me for a reaction should I NOT reward it because "it should just enjoy the run"?

PGM: I don't regard my praise for a job well done as a reward. I regard it as an acknowledgement. I know that you and others can read it as a reward - I don't agree with that reading on what is going on. There is an element of reward involved, I agree, I just don't believe that it is the whole story. When one treats it as the whole story, I am inclined to believe that one ends up with a dog constantly in need of treats.

Acknowledgement is something that can only occur within a social relationship. Rewards as understood within behaviorism (the science of behaviorism that is)give no consideration to that relationship.