Latest Research On Prong & Check Collars

[espinay2]

For Corvus:

"Resistance to Extinction Functions in the single organism" by Eliot Hearst

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1404040/pdf/jeabehav00197-0049.pdf"For both individual and grouped data, the results of the present experiment indicate a greater resistance to extinction for intermittently reinforced behaviour than for continuously reinforced behaviour. the relation obtained between variable ratio during conditioning and resistance to extinction thus confirms the major findings of Mowrer and Jones (1945) and Boren (1953 who used separate groups ....to obtain each extinction data point"

This is what is called the "Partial Reinforcement Extinction Effect".

Basically, continuous reinforcement works well while you continue to reinforce every single time the behaviour is performed (and if people want to do that then no problem). But if you remove the reward at all, the behaviour is less likely to continue and will not continue for as long as if a variable schedule of reward had been used.

[espinay2]

Oh good grief...no wonder people go it alone and use whatever works best for them

Fair point, but remember that some people enjoy these sorts of discussions. However, with clients I rarely discuss theory (unless they want to) and usually present some limited "rules" or "methods" that I think will work best for them. For example, I often use the "300 Peck" method for increasing criteria. I don't use differential reinforcement or even variable schedules because it's hard for people to know when to change schedules.

:laugh: Yes, I wouldn't go into this with clients! Generally using the 'poker machine analogy' is usually as far as I would ever go with discussing theory :laugh: I DO teach them how to vary rewards once a behaviour is taught. Most don't want to have to carry treats or other rewards around with them all the time to get the dog to respond. I also teach them (where appropriate) that aversives have to be meaningful to the dog or they are not aversives.

Discussions on the subject can be interesting though.

Edited November 11, 2012 by espinay2

[The Spotted Devil]

espinay2 - I agree that intermittent reinforcement reduces the likelihood of extinction but I don't believe that a continuous rate of reinforcement weakens the response. The definition of a reinforcer is that it increases the likelihood of the behaviour being performed - if it doesn't then it's simply not a reinforcer.

Edited November 11, 2012 by The Spotted Devil

[espinay2]

espinay2 - I agree that intermittent reinforcement reduces the likelihood of extinction but I don't believe that a continuous rate of reinforcement weakens the response. The definition of a reinforcer is that it increases the likelihood of the behaviour being performed - if it doesn't then it's simply not a reinforcer.

Continuous reinforcement does result in the most rapid learning during initial phases and in this respect it is important. But used long term relies on how much the reward is wanted. If it is known that the reward comes every time the behaviour is performed, and they don't really want the reward this time, they know it will be there the next time. A reward may have to be very high value for the behaviour to be consistent every time. And a reward may lose its value over time if it is freely available for every performance of the behaviour. With variable schedule, they are more likely to take the reward when they can, and try harder to make it happen. And this may actually help to increase the value of some rewards.

[Aidan3]

If you give a reward every time you will [probably] get more of that behaviour, it just might not be exactly what you want. I most often see this with poor recalls. Dog receives reward because owner is relieved and wants it to be a good experience, but fails to improve the recall.

Edited November 11, 2012 by Aidan2

[Simply Grand]

Espinay, you explain this stuff so well.

I studied psych at school and Uni and learnt the conditioning stuff but I can never quite get it straight in my own head, let alone to explain to others!

[espinay2]

If you give a reward every time you will [probably] get more of that behaviour, it just might not be exactly what you want. I most often see this with poor recalls. Dog receives reward because owner is relieved and wants it to be a good experience, but fails to improve the recall.

A recall is a very good example. If the reward is there every time the dog comes to you, there is no incentive to get there quickly - as the reward will still be there when they eventually do. They can therefore do what they want until they are ready to come and get the reward.

[The Spotted Devil]

espinay2 - I agree that intermittent reinforcement reduces the likelihood of extinction but I don't believe that a continuous rate of reinforcement weakens the response. The definition of a reinforcer is that it increases the likelihood of the behaviour being performed - if it doesn't then it's simply not a reinforcer.

Continuous reinforcement does result in the most rapid learning during initial phases and in this respect it is important. But used long term relies on how much the reward is wanted. If it is known that the reward comes every time the behaviour is performed, and they don't really want the reward this time, they know it will be there the next time. A reward may have to be very high value for the behaviour to be consistent every time. And a reward may lose its value over time if it is freely available for every performance of the behaviour. With variable schedule, they are more likely to take the reward when they can, and try harder to make it happen. And this may actually help to increase the value of some rewards.

I'm not arguing the complexities of reinforcement schedules, particularly in a practical sense. I'm saying that, by definition, continual reinforcement increases the likelihood of a behaviour being performed. Otherwise it's not a reinforcer. It's an argument of theory which is where we are at cross purposes I expect.

I guarantee that if my ESS finds game every single time she is asked to retrieve, that ain't going to diminish her response. HOW she retrieves and delivers that game is a whole other matter which is where criteria etc and my use of other rewards (yes food ) come in.

[The Spotted Devil]

If you give a reward every time you will [probably] get more of that behaviour, it just might not be exactly what you want. I most often see this with poor recalls. Dog receives reward because owner is relieved and wants it to be a good experience, but fails to improve the recall.

A recall is a very good example. If the reward is there every time the dog comes to you, there is no incentive to get there quickly - as the reward will still be there when they eventually do. They can therefore do what they want until they are ready to come and get the reward.

No that is about criteria. If you only reward fast recalls you will get fast recalls. If you reward anything as long as the dog comes back you will define "come" as come whenever the heck you feel like it.

Edited November 11, 2012 by The Spotted Devil

[espinay2]

If you give a reward every time you will [probably] get more of that behaviour, it just might not be exactly what you want. I most often see this with poor recalls. Dog receives reward because owner is relieved and wants it to be a good experience, but fails to improve the recall.

A recall is a very good example. If the reward is there every time the dog comes to you, there is no incentive to get there quickly - as the reward will still be there when they eventually do. They can therefore do what they want until they are ready to come and get the reward.

No that is about criteria. If you only reward fast recalls you will get fast recalls. If you reward anything as long as the dog comes back you will define "come" as come whenever the heck you feel like it.

But that is a variable schedule - as you are not rewarding the dog for every time it comes on command, just when it comes fast on command. The variable schedule is used to 'shape' the behaviour.

[Kavik]

If you give a reward every time you will [probably] get more of that behaviour, it just might not be exactly what you want. I most often see this with poor recalls. Dog receives reward because owner is relieved and wants it to be a good experience, but fails to improve the recall.

A recall is a very good example. If the reward is there every time the dog comes to you, there is no incentive to get there quickly - as the reward will still be there when they eventually do. They can therefore do what they want until they are ready to come and get the reward.

No that is about criteria. If you only reward fast recalls you will get fast recalls. If you reward anything as long as the dog comes back you will define "come" as come whenever the heck you feel like it.

But that is a variable schedule - as you are not rewarding the dog for every time it comes on command, just when it comes fast on command. The variable schedule is used to 'shape' the behaviour.

My understanding is that a variable schedule is just that - variable. So you would reward maybe the 1st time, then the 3rd time, then the 2nd time then the 5th and bounce it around. Increasing the difficulty of criteria and only rewarding those that meet or exceed criteria is not variable.

Edited November 11, 2012 by Kavik

[Aidan3]

If you give a reward every time you will [probably] get more of that behaviour, it just might not be exactly what you want. I most often see this with poor recalls. Dog receives reward because owner is relieved and wants it to be a good experience, but fails to improve the recall.

A recall is a very good example. If the reward is there every time the dog comes to you, there is no incentive to get there quickly - as the reward will still be there when they eventually do. They can therefore do what they want until they are ready to come and get the reward.

No that is about criteria. If you only reward fast recalls you will get fast recalls. If you reward anything as long as the dog comes back you will define "come" as come whenever the heck you feel like it.

But that is a variable schedule - as you are not rewarding the dog for every time it comes on command, just when it comes fast on command. The variable schedule is used to 'shape' the behaviour.

It's continuous differential reinforcement.

[corvus]

espinay2 - I agree that intermittent reinforcement reduces the likelihood of extinction but I don't believe that a continuous rate of reinforcement weakens the response. The definition of a reinforcer is that it increases the likelihood of the behaviour being performed - if it doesn't then it's simply not a reinforcer.

Continuous reinforcement does result in the most rapid learning during initial phases and in this respect it is important. But used long term relies on how much the reward is wanted. If it is known that the reward comes every time the behaviour is performed, and they don't really want the reward this time, they know it will be there the next time. A reward may have to be very high value for the behaviour to be consistent every time. And a reward may lose its value over time if it is freely available for every performance of the behaviour. With variable schedule, they are more likely to take the reward when they can, and try harder to make it happen. And this may actually help to increase the value of some rewards.

Thanks for the reference espinay. :)

I am not really sure what to make of this stuff sometimes. Today I had my dogs working for slivers of tinned fruit (don't ask). They were every bit as enthusiastic as they are when I'm using dog food rolls as a reward. They are the same for kibble, cooked meat, dried Schmacko style treats, liver treats, and raw vegetables like carrot and zucchini. My dogs didn't used to work with equal enthusiasm for all those types of food. They had clear preferences. Over time they just seemed to care less and less what they were getting. Pavlov is always on your shoulder, I guess. A few weeks ago one of my dogs was apparently feeling nauseous or something. I didn't know this until I tried to reward him with a treat. He spat it out. Thinking he may have dropped it I tried a few more times with other behaviours and he just didn't seem to want the treats, although he continued to try to earn them and his enthusiasm didn't diminish at all. I found it quite fascinating. There's more at play there than just reinforcement schedules. Then again, I mix up reinforcement types as well. My dogs don't exactly know what reward they are going to get. It seems likely there is an element of gambling to that. Or maybe it's just a generalised positive association with training.

TSD seems to be at the same place I am. How do you define a behaviour? For the purposes of training, I just kinda define it by criteria. Which means to me that if I don't reward it it probably wasn't the target behaviour. I might define a sit as bum on ground immediately wherever you are and whatever direction you are facing and only reward when I see that exact behaviour, but reward every time I see that exact behaviour. But if I defined a sit as bum on ground within 4 seconds of sit cue, there will be a range of sit variations that fall within that definition. If I reward all of them, I will get all of them. If I want to improve speed, I reward the faster ones, which would be considered differential reinforcement. But in my head, if I want to improve speed, suddenly a sit is bum on ground within 2 seconds of sit cue and I reward every incidence of that behaviour. It's the exact same process. I've just defined the behaviour differently. Are reinforcement schedule definitions dependent on definitions of the target behaviour?

[corvus]

Are continuous reinforcement and differential reinforcement mutually exclusive? Only place I can remember seeing or hearing continuous differential reinforcement is from Bob Bailey. Who is obviously a demi-god of learning theory, but is he the only one that uses that concept?

This is a nice article from Clicker Solutions written by Melissa Alexander about it: http://www.clickersolutions.com/articles/2001b/continuous.htm

Variable reinforcement builds strong behaviors. Bob has no problems with beginners using it because they don't tend to train strong enough behaviors. Therefore, without a variable schedule, the behaviors would be too weak and perhaps extinguish, and the person would get frustrated and quit. But competition trainers aren't beginners. They train (or should train) strong behaviors at each step before graduating to the next step.

So continuous (differential) reinforcement enables the competition trainer to build strong, excruciatingly precise behaviors. A variable schedule builds strong behavior, but it also add VARIABILTIY, which is the bane of a competition trainer.

Bob likes continuous. Continuous is simple and continuous works. Why make things unnecessarily complex?

[Bartok]

As far as I new aren't check chains and prong collars banned by use from RSPCA, AWL & Guide dogs Australia.

As well as quite a few training clubs?

[Kavik]

ANKC affiliated clubs do not allow prong collars at training or trials.

But they are not banned as in illegal anywhere except VIC.

Edited November 12, 2012 by Kavik

[tdierikx]

Check chains can be bought in any pet supply store - they are not illegal. I think you need to find specialist providers to get prong collars.

Check chains are only "banned" at training clubs due to pressure from groups who advocate that they are "cruel".

T.

[corvus]

Check chains are not banned at my club, but I could wish they were. I can't remember ever seeing them used well. I've seen them used ineffectively, and I've seen them used as a crutch. When dogs come to clicker class at our club they usually ditch the check chain sooner or later. Our clicker class suffers the same problems of lumping criteria, asking too much too soon, and phasing out treats too early that the other classes have in abundance, but the presence or absence of a check chain seems to relate poorly to any one person's ability to control or train their dog. I'd sooner they used food as a crutch. Maybe then they wouldn't fade it out way too soon and end up with a dog who has found half a dozen other things that are more interesting than training. Check, check, check, check, check. By a process of elimination the poor dog usually learns what they are meant to learn eventually. Like, in a year or two.

[Salukifan]

Check chains are not banned at my club, but I could wish they were. I can't remember ever seeing them used well. I've seen them used ineffectively, and I've seen them used as a crutch.

So what you're saying is that the validity of the tool should be judged by how effectively your average pet person can use it?

When someone arrives at a dog training club with a 65kg Bullmastiff that pulls like a train, then ask yourself what tool you can suggest to that person to help them physically control their dog until such time as they get its attention.

I dont' think check chains have any place in your average pet person's hands but I'd argue the same about haltis. Neither is particularly benign and neither is suitable for use by a person who is not instructed carefully and consistently about how to employ one. Both can be found in your average pet shop though.

Treat bags make good crutches too. Plenty of dogs that wont' do squat for their owners unless there is food around.

Edited November 12, 2012 by Haredown Whippets

[Mrs Rusty Bucket]

I have never corrected my dog

withholding a treat is a correction.

m-sass

Anecdote - I saw an owner trying to catch his dog a couple of weeks ago. The dog was clearly terrified of him. I stupidly helped him catch the dog, and before I handed it over - I told him not to punish it. But he immediately shoved a choke collar over its head and yanked it around - told the dog "not to play games", and then he pushed the dog on the ground and hit it.

In my best growly voice - I said "I will NEVER help you catch that dog again". And he grovelled - smiled and waved at me and left. Next time I see that dog - I will get its id info off its collar and report him for abuse. Some people should not be allowed to own dogs. But this is classic fallout. Ie the punishment is getting the opposite of what the owner intended in a way that treats never would used in the same situation.

I accidentally trained my dog to go off big time at the lawnmower man. But so many (badly timed) treats were involved that she is always really pleased to see him and licks him to death if she gets close enough. I've stopped using treats and started using blocking with more success. Sigh. But at least I didn't use a technique that would likely cause her to blame the lawnmower man for her pain and lead to her attacking him.

So the science - it's already been proven somewhere that reward based training is more effective, faster and lasts longer than aversive based... I think it would be straight forward to split new comers in dog clubs into two streams - old school yank and crank and new school reward based, and see which group graduates more dogs faster - with the same criteria and assessors who don't know which group is which or even that there is a split in the training techniques - though it might be obvious to them if they're doing an assessment.

Another anecdote - watched a GSD club training. Watched tonnes of aversives - scoldings yanking, shaking, alpha rolls!!! being dished out, and no rewards, no pats, no praise, no fun, no treats. All the dogs looked sad. Heads were low, they were making calming signals at their owners and the instructor. The instructor stopped during class to punish his dog for failing to hold a down stay. My dog holds down stays for longer than his could, and my dog wags her tail the entire time.

Mrs Rusty Bucket, these are not "methods" you have described, just people abusing dogs which is not even close to using an aversive where required in a balanced training system??

I would agree that these are abuse, but when we have dog club instructors advocating - how are people supposed to know what a "balanced training system" is? When the majority are getting it wrong and abusing their dogs - how do we fix that. The easiest way is to eliminate the possiblity of abuse - by focusing on the reward based training and avoiding the aversive stuff. It makes dog training for beginnners way more fun too - as they'd much rather reward their dog than punish it. It's also hard for the instructors - if they ask the people to use aversives and then constantly have to correct mistimed application (scold the beginners). But more often - they do the opposite. And this takes the fun out of dog club for the beginners too. We have 700 join every year but only about 100 competition level trainers if that - rejoin.