On The Effective Use Of Punishment

[Cosmolo]

I can't answer for Huski but for me- removal of reward = punisher, no further action from dog, new exercise commenced. Neutral no reward mark = asking dog to immediately try again without additional command/ new exercise.

[megan_]

6. The punishment must happen every time the behavior occurs. If punishment does not happen every time the behavior occurs, the behavior gets put on a variable schedule of reinforcement. Depending on the behavior and how often the punishment actually occurs, the animal could decide that performing the behavior was worth the risk of getting punished.

Point six is weird, if punishment has to happen every time, then clearly the punishment hasn't worked. I would be looking for a better training system, dogs never chose to be in trouble.

I don't think so - if the behaviour occurs twice you punish twice - a dog might not learn after only one correction?I think the point that they're trying to make is consistency - don't let a dog get away with bad behaviour sometimes and then wonder why they do it.

[Tilly]

6. The punishment must happen every time the behavior occurs. If punishment does not happen every time the behavior occurs, the behavior gets put on a variable schedule of reinforcement. Depending on the behavior and how often the punishment actually occurs, the animal could decide that performing the behavior was worth the risk of getting punished.

Point six is weird, if punishment has to happen every time, then clearly the punishment hasn't worked. I would be looking for a better training system, dogs never chose to be in trouble.

An example might be the dog getting up on the couch when he is not allowed. You "punish" the dog by removing him from the couch and putting him in the laundry. If you only remove him some of the time then the dog "could decide that performing the behavior was worth the risk of getting punished".

[PAX]

I don't think so - if the behaviour occurs twice you punish twice - a dog might not learn after only one correction?I think the point that they're trying to make is consistency - don't let a dog get away with bad behaviour sometimes and then wonder why they do it.

Well if it's inconsistency the handler should be punished.

If it is training to stay off the lounge, wouldn't it be easier to reward the dog for being in the right place, I am obviosly thinking about this too much.

I just don't ever want to correct something that I haven't trained and I think that is the most important point. If I have trained it correctly, the dog wont get it wrong.

[bedazzledx2]

Totally agree with you Pax I see punishment as largely unnecessary if you are prepared to teach and manage. Positive is not permissive!

I don't think so - if the behaviour occurs twice you punish twice - a dog might not learn after only one correction?I think the point that they're trying to make is consistency - don't let a dog get away with bad behaviour sometimes and then wonder why they do it.

Well if it's inconsistency the handler should be punished.

If it is training to stay off the lounge, wouldn't it be easier to reward the dog for being in the right place, I am obviosly thinking about this too much.

I just don't ever want to correct something that I haven't trained and I think that is the most important point. If I have trained it correctly, the dog wont get it wrong.

[huski]

I can't answer for Huski but for me- removal of reward = punisher, no further action from dog, new exercise commenced. Neutral no reward mark = asking dog to immediately try again without additional command/ new exercise.

Yep and I also find it easier to have a 'nope you got it wrong' word, I find it faster than just removing the reward, like I find it faster to mark 'yes' when I get a behaviour I want to see instead of just giving her the reward. As soon as I give Daisy a NRM, she tries something else straight away.

Edited January 3, 2010 by huski

[PAX]

I can't answer for Huski but for me- removal of reward = punisher, no further action from dog, new exercise commenced. Neutral no reward mark = asking dog to immediately try again without additional command/ new exercise.

That's true, but it is hardly a punisher in my dogs eyes = never been punished in training so if they don't get the reward they just try again even harder, hardly an adversive the way we train.

I would never reprimand my dog for trying???, that makes no sense to me.

This stuff is hard to express for me in text, much easier in person. If they make the wrong choice then clearly I haven't trained it, so bad me.

[Cosmolo]

Punishment IS necesary in training, even if its just the removal of a reward. If there was no punishment at all how would the dog progress. If i taught a dog to sit and asked them to sit, they didn't and i did not punish, that would mean i would give the reward anyway in which case the dog wouldn't need to sit. I agree that positive does not mean permissive but i don't think that using a punisher means you're not willing to teach and manage- these things aren't mutually exclusive

The area of debate is what kind of punishments are acceptable for what dogs and i completely accept that everyone has different opinions on this- and every dog needs a different approach.

ETA I agree PAX that a NRM is not a punisher- i was just explaining why i'd use it instead of just removing a reward. : ) And i agree its hard to explain in text!

ETA2 I just re read your post PAX and i'm not sure what you think i mean? I am not talking about punishing for trying??

Edited January 3, 2010 by Cosmolo

[PAX]

Well here goes...

Even the way you say removal of reward eerks me, my dogs never know when they will be rewarded and they are never bribed, I never lie to them. If they do as I want they get what they want.

Is a NRM good for you? Or good for the dog? Have you tried reward training without it, and noticed a difference in learning speed?

I believe dogs are so tuned into body langauge that I struggle to fake that I'm happy, not prove that I'm displeased?

I totally get that this is all in the print and we would most likely agree in person... maybe.

The way I use punishment and yes I do use it, I wont open the gate until you sit, if you don't sit that is totally ok, the gate just doesn't open (of course I know the dog knows sit), I am consistant, the rules are the same every day, and I offer lots of great things for complience.

I board a few dogs and I never have a problem with rudeness after a day, the rules are black and white.

I just worry with all the talk of punishment because I see how most people interpret it, and that's smack the dog, which the dog learns nothing except that people are mental.

For example, I have a dog here boarding that the owners adore, but they scream at him to sit or whatever, I have never had to do this. If he doesn't obey the gates don't open, he is so obedient here with me.

I am always amazed at how much dogs learn in a world of ignorant dog owners. We are the supposedly superior race but we expect our dogs to understand English.

[bedazzledx2]

Great post Pax

Well here goes...

Even the way you say removal of reward eerks me, my dogs never know when they will be rewarded and they are never bribed, I never lie to them. If they do as I want they get what they want.

Is a NRM good for you? Or good for the dog? Have you tried reward training without it, and noticed a difference in learning speed?

I believe dogs are so tuned into body langauge that I struggle to fake that I'm happy, not prove that I'm displeased?

I totally get that this is all in the print and we would most likely agree in person... maybe.

The way I use punishment and yes I do use it, I wont open the gate until you sit, if you don't sit that is totally ok, the gate just doesn't open (of course I know the dog knows sit), I am consistant, the rules are the same every day, and I offer lots of great things for complience.

I board a few dogs and I never have a problem with rudeness after a day, the rules are black and white.

I just worry with all the talk of punishment because I see how most people interpret it, and that's smack the dog, which the dog learns nothing except that people are mental.

For example, I have a dog here boarding that the owners adore, but they scream at him to sit or whatever, I have never had to do this. If he doesn't obey the gates don't open, he is so obedient here with me.

I am always amazed at how much dogs learn in a world of ignorant dog owners. We are the supposedly superior race but we expect our dogs to understand English.

[SkySoaringMagpie]

But what kind of punishment are we talking about PF. I disagree, i don't think punishment has a limited place in training more than any other technique- suitable for some, not for others- i just think it needs to be taught properly!

I think it has a limited place in training mainly because I think people don't understand its limitations and until they do I'm disinclined to recommend it given the ways in which it can go wrong (for the deliverer of the punishment, not just the recipient).

Perhaps I'm agreeing with both of you, I don't know.

Fairly early on in my dog career someone put me through that teaching exercise where they loop a slip leash around your wrist and try to get you to follow what is required without using any verbal cues. I had a bit of a jump start on it as I'd already done a fair amount of push hands in tai chi class, but still, with my bigger brain, I didn't get it all the time and the person was an experienced dog person so gave much clearer physical cues than your average member of the GP. It's a humbling experience, and I recommend it for a dog's eye view of the experience.

PF notes the importance of not punishing when it's unclear what is required. Most of the time what the GP gives a dog is a mish mash of contradictory information - how many of us would be prepared to have a choke chain around our necks administered by one of our entry level students? Not me, that's for sure. I agree with Pax that I think it's astonishing that some dogs learn at all when you look at how their owners behave.

[Cosmolo]

I think we all agree on most points- i certainly agree that my entry level clients would never be best placed to use a correction collar. However, this is again where we need to look at dog and owner. I have some clients who pick up any training skills so quickly that i joke with them that they could train to be dog trainers if they ever want a career change I have others that have to work so hard for the basics as it does not come naturally.

I don't understand why saying removal of reward irks you though PAX?

I think training principles are universally simple- Dog does good thing, dog gets good thing and dog does not do good thing, so dog does not get good thing. Its once we start exploring those concepts further that i think the written word makes it tricky. Like with the gate example PAX gave- i would talk about the gate opening as a reward and not opening as removal of the reward (negative punishment) but that doesn't change anything, its still the same exercise with the same meaning for the dog- just different words to explain it.

I have trained with and without NRMs (and still do- it depends on the dogs) and the dogs i do use NRM's for work MUCH better with than without. Doesn't mean i use them for all dogs though. I like NRM's for advanced dogs and advanced exercises.

[Diva]

To me, the eight rules in the OP seem to be speaking of 'positive punishment' - ie actively applying a bad consequence to decrease a behaviour. And in that context I don't have much problem with them.

What you are talking abut Pax - and incidently it's a technique I use a lot - I'd usually think of as the dog simply not meeting the criteria for positive reinforcement (eg door opening) rather than as punishment.

I can see how it could be thought of differently though if it is the action of 'not sitting' that you are punishing, rather than the act of sitting that you are reinforcing, in which case it would to me be 'negative punishment' - ie the removal of a good consequence to decrease the behaviour of 'not sitting'.

I don't think the eight rules are really designed to apply to negative punishment even though they don't make that distinction. I still think in the four quadrants.

[PAX]

I guess it's really semantics, I always worry what the general DOL readers think is a fair punishment for a dog, most people reading this don't have the experience with dogs that those of us argueing do.

I have never met a dog that wants to be wrong and in trouble... ever!

[Cosmolo]

Me neither PAX And yet some people seem to think the dog is doing XYZ deliberately

[Vickie]

Punishment has a very limited place in training. I see too many handlers use 'punishment' for non compliance when they haven't clearly indicated what they want or the dog doesn't know what is expected.

I totally agree especially as it relates to agility training. For some things I punish, but for most I don't. When I give lessons I don't allow NRM or any form of punishment for things like dogs missing jumps (and it's amazing how many people want to punish their dogs for things like this) All it ends up doing is confusing the dog & they either learn to ignore the handler & run out of control or shut down. We break it down & reward for understanding & compliance. It seems to work a lot better.

Punishment IS necesary in training, even if its just the removal of a reward. If there was no punishment at all how would the dog progress. If i taught a dog to sit and asked them to sit, they didn't and i did not punish, that would mean i would give the reward anyway in which case the dog wouldn't need to sit.

FRom the dog's perspective, what it the difference between removal of reward & not delivering a reward they are expecting?

I find this hard to understand from the dogs POV. How do they know the difference between removal/withdrawal & delay of reward.

Let's say I ask the dog to sit & they know I have food on me. If they don't sit, they don't get the food.

What if I am asking them to sit & was planning to use the food to reward for a wave rather than the sit. Between the sit & the wave, do they think I am removing/withdrawing the reward for the sit?

I find my dogs are very responsive to the tone of my commands. I tend to say "lie down" for instance in a very specific tone. Quite often someone else will ask my dogs to lie down & they look like they have never heard the command before. I don't think they even recognise it as the same command. Usually as soon as I tell the person how to say it, the dogs will comply as quickly as they will for me. Is this a punishment if the dogs are expecting a reward?

For that matter, how do we even know when & if the dogs are expecting a reward. I often don't train with a reward on me and will send them to a reward. Sometimes I have a reward in my pocket which I am not intending to give (but still expect compliance). Sometimes I will have 2 different rewards on me. Is it a punishment if I give them the wrong reward? I can't always pick which they will prefer.

Even the way you say removal of reward eerks me, my dogs never know when they will be rewarded and they are never bribed, I never lie to them. If they do as I want they get what they want.

Is a NRM good for you? Or good for the dog? Have you tried reward training without it, and noticed a difference in learning speed?

I believe dogs are so tuned into body langauge that I struggle to fake that I'm happy, not prove that I'm displeased?

Totally agree, especially about the body language bit.

I guess it's really semantics, I always worry what the general DOL readers think is a fair punishment for a dog, most people reading this don't have the experience with dogs that those of us argueing do.

I have never met a dog that wants to be wrong and in trouble... ever!

Again, totally agree. These discussions always go like this. There is the laymans use of the word & the academic use. I find it scary that the general public would think punishment should be something to be embraced.

Edited January 3, 2010 by Vickie

[Zug Zug]

I think the list illustrates perfectly why the use of punishment is so difficult - and so hard to teach.

In particular, I think the combination of point 5 (punishment should not be associated with the handler) and point 6 (it should happen every single time) really makes this a difficult thing to do properly.

I've moved away from punishments. Just too hard for me to 'get' - whereas the positive/reward-based stuff is simple and easy to follow, and you can make mistakes you just fix them and you haven't unduly distressed your dog.

I do use an 'ah-ah' marker when she/we get things wrong. And I will remove her from a situation (e.g. a nice-smelling tree) if need be. But I have to say I am relieved not to be expected to use check chains anymore. I have seen them used quite violently by quite a few people, and it is a slippery slope IMO. It never helped me much as a learner trainer - just made me feel mean and spoilt the fun for both of us.

But I am a learner. If I knew more about the correct (non-violent!) use of punishment then I may feel more confident with it. But I would still hesitate to teach it to newbies in class. Too hard to follow the rules in some cases - especially when so many first-time trainers are dealing with their own frustration while they (and their dogs) learn the ropes.

Edited January 3, 2010 by Zug Zug

[Cosmolo]

To define punishment depends on whether or not the behaviour given is less likely to occur again- so alot will depend on what training you have done with the dog previously, what reward schedule they are used to etc.

I don't mean to imply that the general public should embrace punishment BUT i find it sad that the word punishment has now become loaded with such negative connatations- its fast becoming a 'dirty word'.

Just for the record, i don't mean to confuse anyone through the language i use- its just the way i explain myself. I think we probably are all talking about the same things though.

ETA I think one of the issues with punishers is when they get used by extremists- a sure fire way to colour everyone's view of the word punishment and the use of it in training.

Edited January 3, 2010 by Cosmolo

[corvus]

I think it's also important to note that what constitutes a correction for one dog may not constitute a correction for another.

I think that is what rule 3 is all about. At least, that's how I interpreted it. And rule 1. And rule 2, for that matter.

And the same goes for rewards, obviously. Hardly a thread goes by without someone noting that what works for one dog doesn't work for every dog. Yet that's the very basis of Operant Conditioning. It's just what the dog finds rewarding and punishing that changes from dog to dog. And from moment to moment, IME.

I think it's impossible to completely train a dog with only positive reinforcement, as at some point, you will do something to the dog that they find unpleasant.

I don't think that is the point. The point is if we do choose to use punishments in training, ideally we follow each of those eight rules. No one is perfect and even those of us who try very hard never to punish our dogs invariably do one way or another. Although mucking up once and punishing a dog does not necessarily lead to that dog learning that lesson forever. Unintentional use of punishment is not, one would hope, a part of anyone's training regime.

I would also like to point out that I believe the eight rules relate to P+ specifically, not witholding rewards. I am honestly a little bit dubious that witholding a reward is P-. We had this discussion a little while ago about P- versus extinction, and I think it is a fine line between the two sometimes. Erik is a very persistent little dog and comes back harder and more determined if he is presented with a little frustration. Kivi, we just don't frustrate him much. Baby steps with lots of rewards so he knows he's on the right track. Either way, I don't think it is the kind of thing those rules are aimed at.

Having said that, I think it is entirely possible to train a dog using only positive reinforcement if you're clever enough. They do it with exotic animals all the time. They don't withold rewards from them, but pause for a few seconds before delivering if the behaviour is not up to scratch. They call it a Least Rewarding Scenario. They do a lot of baby steps so their animals don't often make mistakes in the learning phase. I reward my hare every single time he chooses to interact with me, whether he does what I want him to do or not, and he still manages to progress. It's slow, but more because the opportunities to deliver rewards are fewer as you can't rapid-fire rewards to a hare who is afraid of quick movements! And because I don't have a lot of time to spare him, poor boy.

Personally, I'm not clever or forsightful enough to do it without the odd P-, but maybe one day. Hares don't bite your ankles if they get over-excited.

Also, that was a great post by Pax.

[poodlefan]

No doubt how we define "punishment" also plays a part in our views on its use.

I see "punishment" simply as a penalty for not doing the right thing. I don't see it as instructive.

If the aversive also gives the dog some indication of what is required, then I see it as a "correction".