Positive vs. Negative; Reinforcement vs. Punishment

This forum is all about training and behavior. Everything from potty training to working titles!

Postby katiek0417 » January 23rd, 2007, 4:53 pm

Michelle asked me to make a post much like one I had made before on another board. As many of you know, there is much debate about terminology in the dog world. One of those biggest debates concerns the use of positive vs. negative and reinforcement vs. punishment. Often times I have seen these terms misused (according to their true definitions), and have posted.

So, you might ask, what makes me know these "true definitions" better than other people? For those of you who don't know, I have my Ph.D. in psychology, and am currently employed as a professor of cognitive psychology at a 4-year university in MD. My specific areas of research consist of cognitive deficits in neurological populations, with a sub-area of interest in learning and memory. As such, I have taken numerous classes on the principles of learning (both classical - think Pavlov's dogs; and instrumental/operant - where a behavior is followed by a consequence which either results in an increase or decrease in behavior). I also teach these principles to my students, and have written about them.

The specific terminology I am referring to in this post are those associated with instrumental/operant conditioning. These terms have been used for decades, and were originally coined between 1910-1920 by John Watson and BF Skinner.

According to behaviorists, you can have positive and negative. However, it is not positive in the sense that it is good nor is negative something bad. Instead, positive always means giving something. Think of positive as a "+" sign in math. It means the addition of. Negative always means taking away. If you go back to my math analogy, "-" means the subtraction of.

So, when people say they are "all positive," this literally translates as they are always "giving something."

What you give can then be either good or bad. More on that in a second.

Reinforcement means that there is a resulting increase in behavior. Punishment means that there is a resulting decrease in behavior. So you can add something to get an increase in behavior; take something away to get an increase in behavior; add something to get a decrease in behavior; or take something away to get a decrease in behavior.

Now we get to good and bad. Positive reinforcement is giving something to increase behavior; specifically, giving something GOOD to increase the likelihood of a behavior. You tell your dog to sit, he/she sits, you give him/her a treat. Next time you tell him/her to sit, he/she will most likely sit.

Negative reinforcement is taking something away to increase behavior. Specifically, you are taking something BAD away to increase the likelihood of a behavior. I hate this example, but I'll use it here (I don't do this with my dogs). Let's say you are teaching your dog to heel, and you use an e-collar with a continuous stimulation button. You might stimulate a dog at a high level when he is not in the "perfect heel" but decrease the level of stimulation when he moves into the right position. You are taking away the bad (high level of stimulation to lower level of stimulation) to increase the behavior (of walking at the perfect heel).

Positive punishment is giving something to decrease behavior. Specifically, you are giving something BAD to decrease the likelihood of a behavior. For example, you tell your dog to sit, he doesn't sit, you give a pop on the leash which is connected to the prong collar (which results in a pinch at the neck), which is something bad, so next time the dog won't be as likely to NOT sit.

Negative punishment is taking something away to decrease behavior. Specifically, you are taking away something GOOD to decrease the likelihood of a behavior. For example, your dog has a ton of freedom in your house, but he starts peeing everywhere, so you start crating him more (or keeping him tethered to you). You've taken away his freedom so that he stops peeing everywhere.

As I stated at the beginning of this post, it's a matter of semantics. True behaviorists, however, think of "purely positive" as simply meaning always "giving something." Please feel free to respond or PM me if you'd like more information.
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby a-bull » January 23rd, 2007, 6:24 pm

Great post. Thanks for all your effort.

Very interesting areas of research. I want to be you when I grow up. :wink:
DISCLAIMER:

My posts are my own opinions unless otherwise stated. They are not necessarily correct for all dogs or all owners.
a-bull
I live here
 
Posts: 2926

Postby DemoDick » January 23rd, 2007, 6:57 pm

What I find really interesting is that Konrad Most "discovered" the basic principles of OC and a lot of Pavlovian thought years before mainstream science. He didn't use the same labels but he definitely described primary and secondary reinforcers, etc. I'm sure you would find his writings interesting, especially given your vocation.

Demo Dick
User avatar
DemoDick
They Like to Fondle My Gun
 
Posts: 1910
Location: New York

Postby katiek0417 » January 23rd, 2007, 7:44 pm

On January 23 2007, 5:57 PM, DemoDick wrote:What I find really interesting is that Konrad Most "discovered" the basic principles of OC and a lot of Pavlovian thought years before mainstream science. He didn't use the same labels but he definitely described primary and secondary reinforcers, etc. I'm sure you would find his writings interesting, especially given your vocation.

Demo Dick


What's funny is that everyone credits Pavlov with his findings on classical conditioning, where actually it should be Edwin B. Twitmeyer (no joke) getting the credit for a dissertation written at UPenn.

And primary and secondary reinforcers actually refer to something else.
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby mnp13 » January 23rd, 2007, 7:55 pm

Nice work Dr. K
Michelle

Inside me is a thin woman trying to get out. I usually shut the bitch up with a martini.
User avatar
mnp13
Evil Overlord
 
Posts: 17232
Location: Rochester, NY

Postby Malli » January 23rd, 2007, 10:38 pm

So... In my training I start by teaching something by rewarding my dog everytime he does an action that I want him to do (guided or not), once I am sure he has learned what a gesture or noise (verbal command) means, and he chooses not to respond how I have taught him, I give him a verbal correction and/or a physical correction (leash pop with a training collar, most recently a prong, though I get that its not what you use, but how you use it)

From what I understand I use Positive reinforcement and positive punishment ? (correct me if I'm wrong)

All this time I thought it was positive and negative reinforcement

I guess, it doesn't really matter what I call it, as long as I had the ideas/concepts and practices right, though it is nice to know so I can intelligently become involved in a conversation about dog training... :)

very informative!
I can only please one person per day. Today is not your day, tomorrow doesn't look good either.
_______________________________________
"You didn't know of the magical powers of the break stick? It's up there with genies and Harry Potter as far as magic levels go." SisMorphine 01/07/07
User avatar
Malli
E-I-E-I-O!
 
Posts: 6341
Location: CANADA EH?

Postby a-bull » January 24th, 2007, 12:25 am

On January 23 2007, Malli wrote:So... In my training I start by teaching something by rewarding my dog everytime he does an action that I want him to do (guided or not), once I am sure he has learned what a gesture or noise (verbal command) means, and he chooses not to respond how I have taught him, I give him a verbal correction and/or a physical correction (leash pop with a training collar, most recently a prong, though I get that its not what you use, but how you use it)

From what I understand I use Positive reinforcement and positive punishment ? (correct me if I'm wrong)

All this time I thought it was positive and negative reinforcement

I guess, it doesn't really matter what I call it, as long as I had the ideas/concepts and practices right, though it is nice to know so I can intelligently become involved in a conversation about dog training... :)

very informative!


If you're using positive reinforcement to train your dog, you reward the dog the minute he performs the task you requested. You do this every time he performs the task. Once he does it every time without fail (over time), you then drop back to rewarding randomly, (intermittent reinforcement).

Eventually you'll find you can just request a certain task and the dog will simply perform it.

I have heard trainers say some dogs will then stop performing the task somewhere down the line when they figure out they may no longer get a reward, but I have never experienced that. :|
DISCLAIMER:

My posts are my own opinions unless otherwise stated. They are not necessarily correct for all dogs or all owners.
a-bull
I live here
 
Posts: 2926

Postby Malli » January 24th, 2007, 5:30 am

honestly, my dog is not that "eager to please" type. My opinion of him is not the end all and the be all in his world.
On January 23 2007, a-bull wrote:
On January 23 2007, Malli wrote:So... In my training I start by teaching something by rewarding my dog everytime he does an action that I want him to do (guided or not), once I am sure he has learned what a gesture or noise (verbal command) means, and he chooses not to respond how I have taught him, I give him a verbal correction and/or a physical correction (leash pop with a training collar, most recently a prong, though I get that its not what you use, but how you use it)

From what I understand I use Positive reinforcement and positive punishment ? (correct me if I'm wrong)

All this time I thought it was positive and negative reinforcement

I guess, it doesn't really matter what I call it, as long as I had the ideas/concepts and practices right, though it is nice to know so I can intelligently become involved in a conversation about dog training... :)

very informative!


If you're using positive reinforcement to train your dog, you reward the dog the minute he performs the task you requested. You do this every time he performs the task. Once he does it every time without fail (over time), you then drop back to rewarding randomly, (intermittent reinforcement).
Yes, thats how I would teach a command, say, if I was teaching sit to a dog that had never learned. I become certain that the dog understands the task when he does it every time and is still getting the reward. I have dropped back the reward (first when Oscar was a puppy it was treats, then we moved on to his toy, and then I dropped the reward back) and praise of course is after almost every task(every task when learning) I ask of him.

Eventually you'll find you can just request a certain task and the dog will simply perform it.
Nope! Not my dog. Eventually I notice that he becomes slower, turns his head the other way (because he thinks even if the task is asked verbally that he won't "hear" it if he's facing the other way-I do acknowledge that he seems more "visual" then audio), or looks over looks at me, "thinks" about it for a few seconds, " 1/2 does the task requested and then f*cks off)

I have heard trainers say some dogs will then stop performing the task somewhere down the line when they figure out they may no longer get a reward, but I have never experienced that. :|

Oscar would be one of those dogs. I would have stuck with all positive if it seemed to stick with him(we do tricks that way, always with reward, because I intend them to be totally "fun").
I can only please one person per day. Today is not your day, tomorrow doesn't look good either.
_______________________________________
"You didn't know of the magical powers of the break stick? It's up there with genies and Harry Potter as far as magic levels go." SisMorphine 01/07/07
User avatar
Malli
E-I-E-I-O!
 
Posts: 6341
Location: CANADA EH?

Postby a-bull » January 24th, 2007, 8:10 am

Lol, well here we have it! :D Like I said, I've "heard" this can happen, but I've never experienced it myself. He sounds like a pip. :)

I guess you could combine the positive and the negative as you mentioned above, and maybe other people have had good luck with that method. I have watched trainers who try to use both, and frankly, to me, the dogs always looked kind of confused and like that don't get what's being expected of them. :|

How about if you go back to an intermittent reward? Even to this day, I will give my dogs the occasional treat for performing tasks that are basically sure fire, but that spontaneous reward will keep them performing, because they NEVER know when they might get one, so they never stop trying! The spontaneous treat is an intermittent reinforcer, which is a great motivator. :)
DISCLAIMER:

My posts are my own opinions unless otherwise stated. They are not necessarily correct for all dogs or all owners.
a-bull
I live here
 
Posts: 2926

Postby katiek0417 » January 24th, 2007, 3:01 pm

Malli,

What you describe is indeed positive reinforcement and positive punishment. In both cases you are giving something. But with reinforcement you are giving something good (a treat) to increase the likelihood of a behavior, and in positive punishment, you are giving something bad (a leash pop) to decrease the likelihood that he won't listen to you.

As a-bull suggests, you can go to intermittent (also called variable) reinforcement, but you need to decide if you want to do a variable ratio or interval. Variable ratio is a reward after an "average" number of behaviors. So, you could reward an average of every 5 times the dog sits. However, you wouldn't do every 5 times, you are looking for an average, so you might do after 4 sits, after 6 sits, after 5 sits...4+6+5=15, 15/3=5. Same concept for interval, except you are going for average time (like, on average, every 2 minutes, etc).

Also, a-bull, what you are talking about with the ceasing of a behavior is called extinction. Extinction is much more likely with classical conditioning, but can also occur with operant. And even though there can be extinction, it is never fully gone, as it is relearned rather easily.
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby katiek0417 » January 24th, 2007, 3:04 pm



Awww...shucks...Thank you, Michelle :mrgreen:
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby mnp13 » January 24th, 2007, 3:13 pm

ok Katrina, I'm thinking that you are approaching nerd status, but we still love ya!
Michelle

Inside me is a thin woman trying to get out. I usually shut the bitch up with a martini.
User avatar
mnp13
Evil Overlord
 
Posts: 17232
Location: Rochester, NY

Postby a-bull » January 24th, 2007, 3:25 pm

On January 24 2007, katiek0417 wrote:Malli,

What you describe is indeed positive reinforcement and positive punishment. In both cases you are giving something. But with reinforcement you are giving something good (a treat) to increase the likelihood of a behavior, and in positive punishment, you are giving something bad (a leash pop) to decrease the likelihood that he won't listen to you.

As a-bull suggests, you can go to intermittent (also called variable) reinforcement, but you need to decide if you want to do a variable ratio or interval. Variable ratio is a reward after an "average" number of behaviors. So, you could reward an average of every 5 times the dog sits. However, you wouldn't do every 5 times, you are looking for an average, so you might do after 4 sits, after 6 sits, after 5 sits...4+6+5=15, 15/3=5. Same concept for interval, except you are going for average time (like, on average, every 2 minutes, etc).

Also, a-bull, what you are talking about with the ceasing of a behavior is called extinction. Extinction is much more likely with classical conditioning, but can also occur with operant. And even though there can be extinction, it is never fully gone, as it is relearned rather easily.


How does the interval work out for dogs in general? Seems like an 'iffy' way to train a dog to me, but I can't pinpoint why.

I would think if these are tasks her dog already knows, but is refusing to perform, (for whatever reason), the intermittent reward would be the best approach, yes?

I agree that extinction is more probable with classical conditioning and will be better retained; however, with naughty behaviors I find it hard to figure out how to "reward" away a naughty behavior. I always worry I could be accidentally reinforcing it instead; thus, I like the idea of a consequence, but only to the extent I mentioned in another thread, (wherein I gave the example of the bitter apple on socks. My female sock biter has never gone after a socked foot again). :|

I feel like I have experienced the reoccurrence of negative behaviors using classical conditioning more than I have using operant. Any insight?? Am I hallucinating or have I just been lucky, (or unlucky, as the case may be). :D

Thank you for your posts~They're very interesting, (atleast to me, lol).
DISCLAIMER:

My posts are my own opinions unless otherwise stated. They are not necessarily correct for all dogs or all owners.
a-bull
I live here
 
Posts: 2926

Postby a-bull » January 24th, 2007, 3:34 pm

Hey, and as a result of this thread, I think I answered my own question from another thread about why my dog insists on pulling apart anything with stuffing or foam in it---Is it because she perceives the "innards" as a reward of sorts, which occurs almost every time she mauls a stuffie? :lol3: (Or am I being too deep and she's just a goober)?? :P
DISCLAIMER:

My posts are my own opinions unless otherwise stated. They are not necessarily correct for all dogs or all owners.
a-bull
I live here
 
Posts: 2926

Postby katiek0417 » January 24th, 2007, 9:24 pm

On January 24 2007, 2:25 PM, a-bull wrote:
On January 24 2007, katiek0417 wrote:Malli,

What you describe is indeed positive reinforcement and positive punishment. In both cases you are giving something. But with reinforcement you are giving something good (a treat) to increase the likelihood of a behavior, and in positive punishment, you are giving something bad (a leash pop) to decrease the likelihood that he won't listen to you.

As a-bull suggests, you can go to intermittent (also called variable) reinforcement, but you need to decide if you want to do a variable ratio or interval. Variable ratio is a reward after an "average" number of behaviors. So, you could reward an average of every 5 times the dog sits. However, you wouldn't do every 5 times, you are looking for an average, so you might do after 4 sits, after 6 sits, after 5 sits...4+6+5=15, 15/3=5. Same concept for interval, except you are going for average time (like, on average, every 2 minutes, etc).

Also, a-bull, what you are talking about with the ceasing of a behavior is called extinction. Extinction is much more likely with classical conditioning, but can also occur with operant. And even though there can be extinction, it is never fully gone, as it is relearned rather easily.


How does the interval work out for dogs in general? Seems like an 'iffy' way to train a dog to me, but I can't pinpoint why.

I would think if these are tasks her dog already knows, but is refusing to perform, (for whatever reason), the intermittent reward would be the best approach, yes?

I agree that extinction is more probable with classical conditioning and will be better retained; however, with naughty behaviors I find it hard to figure out how to "reward" away a naughty behavior. I always worry I could be accidentally reinforcing it instead; thus, I like the idea of a consequence, but only to the extent I mentioned in another thread, (wherein I gave the example of the bitter apple on socks. My female sock biter has never gone after a socked foot again). :|

I feel like I have experienced the reoccurrence of negative behaviors using classical conditioning more than I have using operant. Any insight?? Am I hallucinating or have I just been lucky, (or unlucky, as the case may be). :D

Thank you for your posts~They're very interesting, (atleast to me, lol).


Hi there,

First off, you're welcome!!!

As far as interval, it actually works much like ratio. Again, it works because your dog doesn't know when the reward will come. As far as a dog refusing to do something. For me, I never give a choice. If my dogs do what I want, I reward (whether it be with food, praise, tug, loves, a bite)...if my dogs don't do what I want, they get corrected...however, if you think of interval in terms of heeling or stays. My dogs never know how long they'll have to hold a stay...but they always get rewarded after. My dogs never know how many steps they'll have to take...I usually reward by time (2 minutes, 1 minute, 30 seconds)...

As far as "naughty" behaviors, this is what I do. An example, when I'm getting their food ready, Sacha incessantly whines (as if she's never eaten before in her life)...I yell at her (which you would think would be reinforcing as it's attention)...however, attention isn't what she wants...so, I yell...AND I feed everyone else. When she's been quiet for a period of time, she gets fed. Counter surfing, another pet peeve of mine...so, I set up my dog, I put a jalapeno covered with cream cheese on the counter. Reward? They didn't think so. At the same time they ate the jalapeno, I said "NO!"

As far as stuffed animals, etc. Well, I have the same problem...I never "fixed" it...I just don't give her those types of toys anymore...I think she likes the sqeakers...
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby Malli » January 24th, 2007, 10:15 pm

ok, I had a second post in response to a-bul's :|

Its disappeared, I guess.

I think I already kind of do the variable; sometimes I'll reward every second task completed, sometimes every 6th or so (It seems after that point that the training looses its "fun" and the dog in turn looses his attention span); this is all of course before the proofing stage and before I am sure the dog knows the task requested.

so what does everyone else use for training? :|

anyone else have a dog like me?
I can only please one person per day. Today is not your day, tomorrow doesn't look good either.
_______________________________________
"You didn't know of the magical powers of the break stick? It's up there with genies and Harry Potter as far as magic levels go." SisMorphine 01/07/07
User avatar
Malli
E-I-E-I-O!
 
Posts: 6341
Location: CANADA EH?

Postby a-bull » January 24th, 2007, 10:15 pm

I don't think you'd want to use the interval on a dog that doesn't already know specific commands, though, would you? For example, you wouldn't want to intermittently reward a dog just learning a command or it may not be learned/reinforced, agree?

My female whines while I'm preparing her food, too. I just tell her "quiet." Less taxing than yelling, lol. :wink:

Question though~Do you think the yell is accomplishing the "quiet," or is the feeding her only after she has ceased whining accomplishing the "quiet"??

Your jalapeno example sounds like reverse clicker training, lol.
DISCLAIMER:

My posts are my own opinions unless otherwise stated. They are not necessarily correct for all dogs or all owners.
a-bull
I live here
 
Posts: 2926

Postby a-bull » January 24th, 2007, 10:19 pm

On January 24 2007, Malli wrote:ok, I had a second post in response to a-bul's :|

Its disappeared, I guess.

I think I already kind of do the variable; sometimes I'll reward every second task completed, sometimes every 6th or so (It seems after that point that the training looses its "fun" and the dog in turn looses his attention span); this is all of course before the proofing stage and before I am sure the dog knows the task requested.

so what does everyone else use for training? :|

anyone else have a dog like me?


You can't drop back to rewarding intermittently unless the command is responded to every time without fail. The idea is to reinforce an already learned, fool-proof command.

If you reward intermittently before the command is fool-proof, in a sense, you are teaching the dog to decide when to cooperate or not.

My female was, and still is, extremely smart and strong willed when it comes to training. She has taught me soooo much.
DISCLAIMER:

My posts are my own opinions unless otherwise stated. They are not necessarily correct for all dogs or all owners.
a-bull
I live here
 
Posts: 2926

Postby katiek0417 » January 24th, 2007, 10:21 pm

On January 24 2007, 9:15 PM, a-bull wrote:I don't think you'd want to use the interval on a dog that doesn't already know specific commands, though, would you? For example, you wouldn't want to intermittently reward a dog just learning a command or it may not be learned/reinforced, agree?

My female whines while I'm preparing her food, too. I just tell her "quiet." Less taxing than yelling, lol. :wink:

Question though~Do you think the yell is accomplishing the "quiet," or is the feeding her only after she has ceased whining accomplishing the "quiet"??

Your jalapeno example sounds like reverse clicker training, lol.


No, when they are first learning, I use all positive, and EVERY time.

It doesn't matter how many times I tell Sacha "quiet" nicely. I've done it with e-collar, etc...she doesn't care....we've thrown things at her, etc...So, what we've found is that when she realizes that it will take longer to get her food, she quiets down quicker...

I've now used the jalapeno thing with 3 different dogs, and it's worked every time!!! I love it...
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby a-bull » January 24th, 2007, 10:29 pm

On January 24 2007, katiek0417 wrote:
On January 24 2007, 9:15 PM, a-bull wrote:I don't think you'd want to use the interval on a dog that doesn't already know specific commands, though, would you? For example, you wouldn't want to intermittently reward a dog just learning a command or it may not be learned/reinforced, agree?

My female whines while I'm preparing her food, too. I just tell her "quiet." Less taxing than yelling, lol. :wink:

Question though~Do you think the yell is accomplishing the "quiet," or is the feeding her only after she has ceased whining accomplishing the "quiet"??

Your jalapeno example sounds like reverse clicker training, lol.


No, when they are first learning, I use all positive, and EVERY time.

It doesn't matter how many times I tell Sacha "quiet" nicely. I've done it with e-collar, etc...she doesn't care....we've thrown things at her, etc...So, what we've found is that when she realizes that it will take longer to get her food, she quiets down quicker...

I've now used the jalapeno thing with 3 different dogs, and it's worked every time!!! I love it...


Funny story . . . I was going nuts tonight trying to figure out what I did with two pieces of rye bread I had put on the kitchen counter. I didn't have to call in CSI after seeing the shredded paper towel on the floor.

Maybe I should try the "Jalapeno Training Method". :D
DISCLAIMER:

My posts are my own opinions unless otherwise stated. They are not necessarily correct for all dogs or all owners.
a-bull
I live here
 
Posts: 2926

Next

Return to Training & Behavior

Who is online

Users browsing this forum: No registered users