Extinction Vs -p

[Aidan]

Re-read the link and you will see that she mentions Extinction as a process, not a reinforcement quadrant, so I am really not understanding your argument Aiden.

Clearly you are misunderstanding my argument. I have never said that extinction is a reinforcement quadrant! I have been very clear throughout the entire discussion that extinction is the failure to reinforce a previously reinforced behaviour (if you like, a "failure to use an expected reinforcement quadrant").

The "quadrants" are all consequences, the environment provides a consequence to the organism. Extinction is a failure to provide an expected, reinforcing consequence (in broad, layman's terms). This is why it is not a "quadrant".

In order for the behaviour to become extinct or lessen you can use the opposite of R+ which is P-.

In order to diminish behaviour you can use either +P, -P or Extinction. Are you understanding that Extinction is not the result of -P? I take this quote from the article I linked to earlier (http://www.animalbehavioranswers.com/id117.html)

"But there’s another process that sounds a lot like extinction. Negative punishment sounds, at first blush, so much like extinction that some people have decided that extinction is really the same thing. But they dismiss it too soon. "

In the example used in the original discussion you were making this same mistake, throwing an extinction procedure into the -P quadrant by referring to it as -P when you had not taken away a stimulus contingent upon the unwanted response (but instead had withheld the reinforcer that was maintaining the unwanted response).

[Aidan]

I think I can see why you think I refer to extinction as a "reinforcement quadrant". In the earlier discussion you said that "withholding a treat until the correct response is given" was -R (your typo) and I said that it was positive reinforcement.

OK, so when a dog gives you a response and you give it a treat that is +R, right? (Assuming the response increases, of course)

Whenever you use +R you are always withholding the treat until the correct response is given. How else can you do it?

Whether that "withholding the treat" is extinction or not depends on whether the reinforcer you withhold is maintaining whatever behaviour is happening until the reinforcer is given to the dog. Unless we can find something else that is maintaining those behaviours, we tend to assume that it is. It is definitely not -P, not by any accepted definition.

Edited September 7, 2009 by Aidan

[corvus]

No really, I'm glad you started this topic, Aidan. It helps me understand what we're doing when we use different training methods and why it works. Which is so useful in understanding why something might be not working, or working slowly.

[Are You Serious Jo]

One tip, make sure you are clear between the specific meaning of a positive reinforcement event and the concept of training using only positive reinforcements. What happens within a training session may include a few different types of reinforcement, don't think of it purely as only one of the other. One of the reasons I don't like the term purely positive is because it technically can't be always positive.

[Aidan]

One tip, make sure you are clear between the specific meaning of a positive reinforcement event and the concept of training using only positive reinforcements. What happens within a training session may include a few different types of reinforcement, don't think of it purely as only one of the other. One of the reasons I don't like the term purely positive is because it technically can't be always positive.

I agree, many clicker trainers tend to miss the negative reinforcement procedures that are happening regardless of their intentions, particularly with reactive dogs (my specialty).

Although I have to ask, are you referring to anything in particular about either of those posts? I have used +R purely in the technical sense, and when referring to the definition of extinction, I have deliberately not made the distinction, except where appropriate, because extinction (operant extinction) can apply to either positive or negative reinforcement. Do you think others might erroneously assume that I am using it [+R] as a popular term to describe a training style?

I have learned to be particularly cautious even when writing for a popular audience because most of my writing is peer-reviewed. You scientists are a pedantic lot and over time I have learned that it is for a good reason. You don't maintain these standards simply because you enjoy it!

[Are You Serious Jo]

Do you think others might erroneously assume that I am using it [+R] as a popular term to describe a training style?

I can't comment on all your posts, but I think it is a common problem across the board. I do think I have seen cases where one person is referring to the concept and another to a +R event and they are confusing each other and everyone else. Since positive training has become popular with a lot of people the jargon is getting misused (I think). It is a complicated area, not as black and white as some people new to the positive training camp think it is.

[corvus]

The jargon isn't exactly clear anywhere if you ask me (which no one did). Aidan has pointed out to me that a correction can be positive or negative if it fixes a problem, but lots of people use it to mean the same thing as a punishment, leading to people like me getting irritated about using the word at all. I guess the same reason why I get irritated with words like dominance (even though it's often accurate) and leadership when used to describe how we relate to our dogs. And "negative reinforcement" is often used interchangeably with "punishment" even though they aren't the same thing. Some terms aren't particularly intuitive and some are misused or have come to carry connotations that change how the word might be interpreted.

[Aidan]

No really, I'm glad you started this topic, Aidan. It helps me understand what we're doing when we use different training methods and why it works. Which is so useful in understanding why something might be not working, or working slowly.

Thanks Corvus. There will always be some who are simply not interested, and others who want to keep "sharpening the saw".

[Aidan]

The jargon isn't exactly clear anywhere if you ask me (which no one did). Aidan has pointed out to me that a correction can be positive or negative if it fixes a problem

Just to be clear, I'm pretty sure I didn't use the terms "positive" or "negative"

Some terms aren't particularly intuitive and some are misused or have come to carry connotations that change how the word might be interpreted.

This is true in many fields. We just have to do the best we can when it matters.

[corvus]

Whether that "withholding the treat" is extinction or not depends on whether the reinforcer you withhold is maintaining whatever behaviour is happening until the reinforcer is given to the dog. Unless we can find something else that is maintaining those behaviours, we tend to assume that it is. It is definitely not -P, not by any accepted definition.

Okay, I have a question about this.

Could you offer an example?

[Kelpie-i]

Aiden I've had a chance to re-read the link with a fresher mind and I understand what you are saying.

After doing some deeper research on the various definitions of extinction v -P, I can see how easily it can be misinterpreted which I obviously had. Fancy thinking that -P was the removal of a reinforcer, well at least this was what I was taught (or remember learning) so many years ago, when in actual fact it is the removal of something else that the dog may see as valuable but not the actual reinforcer itself.

I never really bothered to research -P any further, hence this was my understanding of it. Having said that, I did initially think that you were putting extinction in the same boat as the quadrants.

(goes away with tail tucked)

Question: When teaching a new behaviour by successive approximation ie. whilst luring you give the dog bits of the treat for each increment that draws closer to the desired behaviour and then the dog suddenly gets up and you withdraw the treat, what would you call that process? There was no associated reinforcer to the behaviour, therefore it is not extinction. Do you still see this as R+?

[Aidan]

Thank you for taking a deeper look at this.

Fancy thinking that -P was the removal of a reinforcer, well at least this was what I was taught (or remember learning) so many years ago, when in actual fact it is the removal of something else that the dog may see as valuable but not the actual reinforcer itself.

-P can involve the reinforcer that is maintaining the target behaviour, but there are two important points:

1. it doesn't have to be that reinforcer, just a stimulus that is removed that results in punishment

2. the stimulus is taken away as a consequence of the unwanted response

(goes away with tail tucked)

Haha, don't be silly!

Question: When teaching a new behaviour by successive approximation ie. whilst luring you give the dog bits of the treat for each increment that draws closer to the desired behaviour and then the dog suddenly gets up and you withdraw the treat, what would you call that process? There was no associated reinforcer to the behaviour, therefore it is not extinction. Do you still see this as R+?

Good question. Sometimes we can't answer this sort of hypothetical question without making some assumptions. I'll fill in some extra details.

Let's say we're working on 'roll over' (it fits with the info given), we've been doing this for a while when suddenly the dog gets up and walks off. Trainer puts food back in the fridge and starts washing the dishes. These are a few of the possibilities:

1. the dog was satiated - in which case withdrawing the food is meaningless

2. the dog was distracted by something going on outside and goes to check it out - in which case withdrawing food is probably meaningless

3. the dog was distracted by a noise, but realising that the food is no longer on offer, returns to work and does not get up and leave when working on this behaviour in future sessions - in which case withdrawing the food was negative punishment.

4. the dog pops up into a sit (previously reinforced with food) half a dozen times while working on 'roll over' - in which case you are probably looking at an extinction procedure

I'm not sure if this answers your question or not?

It may also answer Corvus' question?

[corvus]

It may also answer Corvus' question?

Yes. I started writing out a very similar example and then realised that I didn't really know where I was going with it.

So, say you are preparing a meal for your dog and they jump up and down on the spot. You ignore it and the food stays on the counter until there is a lull in the jumping. Over time the jumping fades away. Now say you ignore it, but when you go to give your dog the food they start really jumping, so you pull the food away from them and only put it down when they have stopped jumping and over time the jumping stops. Would it be correct to say that in the first scenario it's extinction but in the second scenario it's P-? Or is that a false distinction?

[Aidan]

It may also answer Corvus' question?

Yes. I started writing out a very similar example and then realised that I didn't really know where I was going with it.

So, say you are preparing a meal for your dog and they jump up and down on the spot. You ignore it and the food stays on the counter until there is a lull in the jumping. Over time the jumping fades away. Now say you ignore it, but when you go to give your dog the food they start really jumping, so you pull the food away from them and only put it down when they have stopped jumping and over time the jumping stops. Would it be correct to say that in the first scenario it's extinction but in the second scenario it's P-? Or is that a false distinction?

We can probably assume that the jumping was reinforced by food at some point, so in the first example we have qualified that it is an extinction procedure.

The second example doesn't appear to be any different, because at no point did the dog have the food. He is still jumping in expectation of a reinforcer, and that reinforcer is not forthcoming.

[corvus]

Hmm. I was just reading a book where the writer explains P- with the example of a dog that would scratch the door of the crate when the trainer was approaching with a meal, so the trainer would turn around and walk away when the dog scratched. Is that different to going to put a bowl down for the dog, but then pulling it away before doing so? This is why I don't like the quadrants very much. But I do think that it's important stuff and the better I understand it the better I will understand my animals and my training methods.

ETA, could it be the presence of the trainer with the food was a stimulus for the scratching and not the food itself? In which case, if the trainer removes themselves and the food, could it be P-?

Edited September 9, 2009 by corvus

[Rusky]

We can probably assume that the jumping was reinforced by food at some point, so in the first example we have qualified that it is an extinction procedure.

or reinforced some other way ? like attention ? or the dog jumping at mealtimes and then always getting the food (dinner) mean that the reinforcer is the food?

Does it matter how the behaviour is reinforced to qualify as an extinction burst ?

[Aidan]

Does it matter how the behaviour is reinforced to qualify as an extinction burst ?

In this example at a practical level, probably not, but the definition of an extinction procedure does require you to remove the reinforcer that is maintaining the targeted behaviour (otherwise it would be -P).

It becomes more important when you are doing functional analysis, or attempting a Differential Reinforcement procedure.

[Aidan]

Hmm. I was just reading a book where the writer explains P- with the example of a dog that would scratch the door of the crate when the trainer was approaching with a meal, so the trainer would turn around and walk away when the dog scratched. Is that different to going to put a bowl down for the dog, but then pulling it away before doing so?

Not that I can see. If we can make the assumption that scratching at the crate door was being maintained by the approach of the person or the food then I would say that was an extinction procedure.

[corvus]

Hmm. I was just reading a book where the writer explains P- with the example of a dog that would scratch the door of the crate when the trainer was approaching with a meal, so the trainer would turn around and walk away when the dog scratched. Is that different to going to put a bowl down for the dog, but then pulling it away before doing so?

Not that I can see. If we can make the assumption that scratching at the crate door was being maintained by the approach of the person or the food then I would say that was an extinction procedure.

No wonder it's confusing. Oh well, I think I've got it, now, and next time someone tries to confuse me with another example of extinction instead of P- I won't be so confused.

[Kelpie-i]

I read in one of Lindsay's books that one way to avoid an extinction burst is to put the unwanted behaviour on a continuous schedule of reinforcement first (yes encourage it to happen!!) and then totally remove all reinforcement (extinction). If the behaviour is rewarded variably it is more resistant to extinction, therefore placing it on a continuous schedule can lessen the chance of any extinction bursts. I found this interesting and have never tried it. I guess it would need to be planned correctly and other environmental factors taken into account.

Has anyone tried this?