Quite often in discussions about negative reinforcement, someone brings up a plethora of examples from human life that sound harmless and benign. Here are some of the items that are often mentioned:
- Scratching an itch
- Washing your hands to remove dirt
- Drying your hands on a towel to get the water off
- Trimming your fingernails to reduce their length
- Taking out the trash when the kitchen can gets full
- Turning on the windshield wipers in the car to remove rain from the windshield
- Taking an aspirin if you have a headache
- Covering your ears if there is a loud sound
- Putting on a coat when the temperature drops
- Using an umbrella to stay dry
Reading these, one can come away with the impression that negative reinforcement is just no big deal. What’s the fuss about and why do some people try hard to avoid it in animal training?
All of the above examples have something in common. They are what B. F. Skinner termed “automatic reinforcement.” Here’s a typical definition:
Automatic reinforcement occurs when a person’s behaviour creates a favourable outcome without the involvement of another person (Cooper, Heron, & Heward, 2007).
Automatic reinforcement can be either positive or negative. The above examples are negative since they deal with removing an aversive condition. (For a review of the four processes of operant learning, you can read my post Operant Learning Illustrated By Examples.”)
We learn to do the things in that list, usually as children or teenagers, to make ourselves more comfortable. Some undesired condition develops, we take action to change it, and if successful we personally reap the benefit. Check out this post if you’d like to see a movie with 16 automatically negatively reinforced behaviors (by a human). They can be pretty invisible until you make an effort to notice them.
In applied behavior analysis, one analyzes operant behaviors like this: there is an Antecedent, a Behavior, and a Consequence. The antecedent and the consequence are in or from the environment. The behavior is that of the subject person or animal. Applying these “A B Cs” can be quite helpful in seeing what is going on.
In negative reinforcement, the antecedent is the undesirable condition. So for an example of automatic reinforcement:
- Antecedent: There is dirt on Mike’s hands
- Behavior: Mike washes his hands
- Consequence: Dirt is gone from hands
Socially Mediated Reinforcement
The other type of reinforcement is called “socially mediated” reinforcement.
If another person is involved with the function of the behaviour this would be defined as “social reinforcement” or “socially mediated reinforcement” (Cooper et al., 2007).
In negative reinforcement this means that another person or group removes the aversive stimulus. And most important, they can intervene in the reinforcement process and can determine what behavior is required to get the aversive to stop or reduce in intensity.
In automatic negative reinforcement, the reinforced behaviors are directly related to solving the problem. The actions of opening the aspirin bottle and taking an aspirin are reinforced by the relief the aspirin provides from a painful condition.
But in socially mediated negative reinforcement, perhaps someone else has the keys to the medicine cabinet. That person could require some unrelated behavior from you (say, clapping your hands three times) before you got access to the aspirin. If they were consistent, that behavior could be reinforced by the relief provided the aspirin, and would increase. When you had a headache and needed an aspirin, you would probably clap your hands three times. That’s a big difference from being able to walk in the bathroom and get your own pill.
Connection to Dog Training
Back to automatic negative reinforcement. Here are some examples for dogs:
- shaking off water when wet
- biting their own toenails when they get too long
- scratching an itch
- scooting on their butts when their anal glands bother them
- coming in out of the rain
- getting in the shade when hot
- lying down in the kiddie pool to cool off
Those don’t sound so bad either, do they? Dog gets a little uncomfortable, takes action, gets comfortable again.
- Antecedent: There is water on Fido’s coat and skin
- Behavior: Fido shakes off
- Consequence: There is less water on Fido
Reading these lists, you could come away mystified that so many people disapprove of using negative reinforcement in training. It sounds like no big deal.
The problem is that the application of negative reinforcement in training is quite different. In training we are doing the equivalent of “socially mediated reinforcement.” The animal is no longer autonomously in charge of removing an aversive condition by performing a behavior directly connected to it. There is a human requiring a certain behavior before the aversive can be turned off or escaped.
Negative Reinforcement in Dog Training
In dog training, the human has control of the reinforcers (and the aversives, if used). So in negative reinforcement, rather than a dog performing a natural behavior to remove an aversive condition, the human has influence over the stopping of the aversive. And like the clapping hands for aspirin example above, the human can choose a behavior that is unrelated to the natural way the dog might escape the aversive.
Some examples of human controlled negative reinforcement in dog training:
- A dog moves forward from sit stay. The human walks or leans forward into their space until they sit back down.
- A dog is wearing a shock collar. The shock is turned on and stays on until the dog performs a specified behavior.
- A dog is nervous about the teeter in agility. You lead the dog close to the teeter and mark, then turn around and take her away while giving her a treat. (The treat is potentially positive reinforcement, getting farther away from the scary teeter is potentially negative reinforcement.)
- A dog is being trained to hold a retrieve dummy using a force fetch technique. The trainer pinches the dog’s ear continuously until the dog opens its mouth and accepts the dummy.
- Your dog is reluctant to take the steps down into the back yard. You walk into the dog’s space until they go down.
- You teach your dog to yield to pressure on the leash when you pull gently on it (often combined with a treat, which is generally positive reinforcement).
- You want to teach your dog the trick of covering their face with their paws. As a first step to get your dog to touch their face, you put a piece of tape or post it note on their face. The dog paws at her face to get the tape off. (People usually use this one as well in tandem with positive reinforcement of giving the dog a treat after she paws her face.) The tape comes off either through the dog’s efforts or because you help.
- A herding dog has its first exposure to sheep. The handler uses a staff to create pressure to keep the dog from rushing in too close to the sheep.
- Your dog has issues with other dogs. You take her close enough to see another dog and start evidencing slight concern. When she performs a certain behavior, for instance a head or body turn away, or a ground sniff, you lead her farther away from the dog. (e.g. see at 0:44 – 0:51 and 1:17 – 1:27)
- You do agility with your dog. Sometimes you push into her space to influence her path. (Clearest example in the video is around 1:24 – 1:29, where the handler is using her whole body to push. She also often “pushes” with an arm held high.)
- Your dog doesn’t respond on the first cue to sit, so you repeat it again and again, keeping up the verbal nagging, perhaps even raising your voice and fussing at her, until she sits. There’s an example of this one in my movie that has examples of the four procedures of operant learning at 3:11 – 3:37.
- You want to teach your dog to get out of your way or back up, so you continue walking toward her until she moves.
Each one of these begins with particular situation or condition and requires the dog to perform a specific behavior to get it to stop or decrease.
So here’s what a sample ABC looks like now:
- Antecedent: Human is pinching dog’s ear
- Behavior: Dog opens mouth and accepts retrieve item
- Consequence: Human stops pinching
Perceive the difference? The handler either creates the aversive condition, or utilizes one occurring in the environment. The handler cues or waits for a specific behavior. When that behavior is performed, the handler stops the aversive condition or moves the dog away from it.
The handler gets to create a contingency.
People often think I am some sort of purist since I write critically about negative reinforcement. But it’s not the R- itself I’m a purist about. I’m a purist about being honest about it. We live in the real world, with our dogs, and it’s very hard to go through life without that quadrant sneaking in now and then. But we make choices all the time. Using an aversive to train our dogs, or to get through a tough situation, is a choice.
In the spirit of honesty, here are some things I have done.
- I do agility, and pushing into the dog’s path with your body on occasion is pretty hard to avoid. But I’m learning to be a better handler, and there’s always a different way to handle just about any sequence.
- I have taught response to leash pressure (put a tiny pressure on the leash, and when the dog yields to the pressure, treat). It is a combination of negative and positive reinforcement. See below for my comments about it.
- I have used body pressure in the past to get my dogs to move or return to stay position or to walk down the stairs. I have a movie showing their sad reactions to that. I have a third, less sensitive dog, on whom I occasionally perform a body block, for instance, if she is liable to bash into my other dogs if I don’t take action. However, that may not constitute negative reinforcement, since there doesn’t seem to be a behavioral change on her part and she is not getting more sensitized to the pressure. So it’s just plain application of pressure on my part, in a pinch.
- I have also used negative reinforcement a few times when the other alternative would have been unintentionally positively reinforcing a behavior that I really don’t like. For instance, in the past when I was holding Zani or Summer and they started to struggle, a couple of times I hung on until they became still, and only then released them. I didn’t want to reinforce the struggling by putting them down right then. If I found myself doing something like that repeatedly I would take action about the problem. And actually, I did in that case. We do practice handling and associating being held with great things, and I think it’s been years since I had to apply that kind of restraint.
- When I go to bed at night and Clara is in my place, I tell her to move and I move into her space. I have a rule about not using treats on the bed so haven’t been able to think of a different way to get her to move.
- And of course through daily life with multiple dogs I probably unconsciously use body pressure more often than I know, although I make an effort to pay attention. And again, the dog I am most likely to use it on is not showing any particular behavior changes that I can see, so there may not be negative reinforcement going on.
It’s important to me to be honest about it. But I also want to make it clear: I strive not to use these things. Please don’t interpret the above items as condoning negative reinforcement. I’m always looking for better ways. I hope you are too.
Even some of the more benign-sounding techniques in the list of links above I have seen to be quite unpleasant to dogs. I used to do leash pressure exercises with my dogs as part of leash training, and when I look at video footage of those training sessions, I can tell they didn’t enjoy the training as much as most of the other things we do, especially at first, even though they got a treat every time they yielded to the (tiny!) pressure. Perhaps more skill on my part would have helped, but nothing would remove the fact that I was putting physical pressure on their necks.
Defenses of Using Negative Reinforcement in Training
Some people say, “I’d rather take something away that my dog doesn’t like (negative reinforcement) than take away something that he does like (negative punishment).”
This comment erases that important distinction: the contingency from the handler. “Taking something away that the dog doesn’t like” sounds so benign. But they are omitting that they are putting a contingency on that removal. They are not taking it away until the dog does something they want; they are using the unliked thing as a threat. I take away things my dogs don’t like frequently., but I don’t make them do something first! I don’t want to come to view an untoward environmental condition as an opportunity to get behavior out of my dog. My dogs rely on me for their safety and happiness. If something unpleasant happens, I do my best to get them out of the situation.
Others stipulate that they do not cause or create aversives, but only use them when they occur naturally in the environment. This is a red herring. I think a more important point is that they are using the aversive to get behavior. Wherever the aversive is coming from, they are choosing to use it. And usually there are other choices.
I wrote this piece because I think equating the type of negative reinforcement used in most training with scratching an itch or washing one’s hands is seriously confusing and misleading. I hope it is helpful.
Related Posts
Copyright 2013 Eileen Anderson
Interesting observations as usual. I think the important part of training that I want to incorporate is not to commit to a quadrant but rather to commit to the relationship I have with my dog. Information is known to be one of the most powerful reinforcers available to us and when using R- gives the dog information that is useful to him, and he is not distressed, then that will be the tool that I use. You seem to be equivocating here; the message I am getting from this blog is “I don’t like using it, but sometimes I do and sometimes the dogs don’t mind”. If the dog’s relationship with you is improved, if he gains skills that will keep him safe, then why worry about what quadrant you are in? This seems counter intuitive to me.
I’m interested in hearing your response on this as well. For instance, the video with the dog on the teeter… I don’t see anything wrong with that. You said “lead the dog to the teeter” but the dog was in no way forced onto the teeter, he was completely off leash and choosing to go on his own. It seemed an effective way to get him past his fear, honestly. You also mention BAT – do you not think BAT is a humane way to train?
Hi Gabriellevannini–one of the reason I posted all the links was so that people could both the differences and commonalities between lots of applications of negative reinforcement. I actually chose to put the teeter one first because I think it is comparatively benign. I like that the trainer knows and states that she is using R-. Also, she is skilled. My intent is certainly not to lump a bunch of criticism on her.
You are right that I shouldn’t have said “lead” the dog to the teeter. She encouraged the dog to the teeter. However, I don’t agree that he was choosing to go on his own. If the owner had been sitting in a chair in the corner of the room otherwise occupied and the dog was released and climbing on and exploring the teeter of his own accord, then I would say he was choosing it. As it is, she is strongly influencing his choices by 1) her body position; 2) calling him; 3) clicking and treating when he walks past a certain point on the board; 4) at least once using her body movement as a lure. In the grand scheme of things, this is mild pressure, but in my observation it is pressure nonetheless. Watching the whole video, one can see exactly what the dog is saying: he doesn’t want to get close to that pivot point. He pauses, he jumps off, he does what appears to be stress sniffing. The trainer’s criterion is very clear: take one step out of the contact zone higher onto the teeter and you’ll get to leave and get a treat. His response is often, “no thanks.” I do agree that such training can be effective. So can lots of things that might not be my or your first choice.
As for BAT: I’m a humane hierarchy trainer. I always try to use absolutely the most humane method for my dog, judged by information about learning theory and observation of my dog. I use desensitization and counterconditioning with my feral dog and seeing her develop absolute joy for the things she used to be afraid of is incredibly rewarding. So while I wouldn’t go so far as to classify the particular BAT video I linked to as “inhumane” training, I find DS/CC to be more humane.
Hi Sue, glad you found it interesting. Equivocation? Maybe. I thought it important to be frank about my own practices but I consider that more of a progress report than a platform statement. If I wanted to model any behavior for readers it would be willingness to learn and improve and reassess. I would much rather people have that as the takeaway message than “avoid this quadrant” or “watching your dog’s body language is crucial” although those are indeed my values. If my skill level were better I could perhaps avoid R- entirely. Or if my observation skills were better I might always know exactly what was the most comfortable method for my dog, of whatever quadrant.
Improving as a trainer means improving my understanding of learning, improving my mechanical skills, and becoming a better observer in real time. The last one is definitely not the easiest, and I don’t think I’m alone in that. I have had some unpleasant surprises when watching videos of stuff I thought my dogs were fine with. Hence it is very important to me to analyze and heed the particular learning process (quadrant) that is happening.
Very thought provoking post! Thank you! And I love your transparency about how you do occasionally use -R. I don’t think I’ve ever commented before, but I really enjoy reading your blog.
As an aside, we have had a dog force fetch trained by a trainer considered “gentle” in the upland community, and I’ll never subject any of our dogs to that again. When he returned from training, I compared his “hold” with another of ours around the same age who I’d taught a hold with shaping/clicker training. The difference? The FF trained dog’s eyes would glaze over and he’d become obviously fearful–he performed the behavior like a robot. My clicker trained dog, on the other hand, was eager and happy to hold–for him, it was just a great game. This just confirmed my beliefs even more that I made the right choice “crossing over” to kinder world of training….
Hi Champ’s Mom! Glad you enjoy the blog.
That’s a sad story about the force fetch training. I still have a post in the works about “shut down dogs.” Dogs who are able to perform behaviors but for whom every movement is minimal and careful, generally because of aversive methods. I know what you are describing.
Glad you crossed over, too!
Oh, I’m you are going to be discussing this in Part 2! I used to consider him shut down in FF, too, but had been reconsidering, especially after reading your Part 1, since he was still performing.
ha ha. you’ve been reading too many threads on facebook 😉
well done and great examples.
I think we all must be reading the same threads…
Who, me? Casts eyes skyward and looks innocent. One of these days I’m going to write a post on Negative Reinforcement in Blogging, or “Someone is Wrong on the Internet!” But truly, I’m glad about this topic. I finally figured out a distinction I’ve been struggling to make.
Glad you liked it, Diana. Thanks for saying so.
This is one of the best posts on negative reinforcement that I’ve seen in a long time. Very refreshing.
Thank you thank you Anne!
Oh dear here we go once again. Except in BAT type work, and as you can see in the chosen video, the human is not applying the antecedent, is she? The dog is walking of her own free will towards the dog (note the dog chooses to walk ahead and alone) and then is mildly interested in the other dog, not really even showing slight concern, and then chooses when to disengage and then chooses when to leave. The set up could just as easily be done off leash, when the dog understands he is in control, indeed many set ups are. The long leash only performs a way to guard against magnetization in the early stages. And all this begs the question, if any supposed R was working, why is the dog choosing to walk back towards the dog, closer and closer and not giving ‘a required behavior ‘ faster to avoid the ‘applied unpleasantness’? Simple. It’s not -R. Or it’s ‘automatic negative reinforcement’ but even then, there is no problem to solve.
Hi Helen,
As I mentioned in the post, where the aversive comes from is immaterial (but I do imagine that in this case the decoy dog was there by arrangement of the trainers). What the dog does between iterations is also immaterial to the behavior analysis of the specific sequence. There are many reasons why a dog might approach something they don’t feel good about, and the notes in the video make it clear that the subject dog, Lexi, doesn’t feel good about other dogs in certain situations.
The point is this. Starting at the moment that the subject dog is gazing at the decoy dog:
Antecedent: A dog is in Lexi’s area of perception
Behavior: Lexi turns her head away and down
Consequence: Handler says “Yeah!” and runs with Lexi away from the other dog
There may be other things going on in the classical and operant realms within the session. But that sequence is an example of R-. The notes in the video say that Lexi “wants the dog to go away” and that she is “rewarded by moving away from the trigger.”
That is not automatic negative reinforcement. If it were automatic negative reinforcement, we wouldn’t be waiting for the head turn, the ground sniff, etc. And the handler wouldn’t be marking it.
As Grisha Stewart describes it on the Ahimsa Dog Blog: “For most dogs, especially fearful dogs, I prefer for them to retreat from the Scary Monster as their reward
when they’ve done an acceptable alternative behavior
, versus only having the monster leave.” That is a contingency.
R- is not the end of the world in and of itself. There are plenty of trainers who use it openly, fairly gently, and with skill. I’ve listed ways in which I have used it. I think it is good to be clear about when and why we are using it.
I get what you’re saying, that R- happens on an automatic level and people use that to justify socially mediated R-. But I also don’t see one as relevant to the other. If so, just based on the same principal, I could get away with quite a lot of abuse and neglect just because some things happen naturally. Where is the line drawn at how far you want to take it on this alone? It’s a poor way to consider what you choose to expose your dogs to.
But I can come up with some examples of socially mediated R- in training that I really don’t have a problem with, that fit into my strict criteria for not causing undue stress, or intimidation, fear or pain. For instance, a front attach harness turning the dog off center until they release tension on the lead. Some may argue that this is management but I argue that it still counts as training if the dog actively learns from the stimulus and the behavior is decreased or increased due to it. To break it down into a really obvious level (not to insult anyone’s intelligence), Negative (being turned off center) reinforcement (keeping the leash tension slack).
Here’s one for positive punishment. Spraying bitter apple on chair legs. Dog goes to chew on the chair leg and is punished via the unpleasant taste. If the dog forms a negative enough association to the activity of chewing on the chair legs, the behavior will become extinct. Sure, you can take other proactive steps than to apply the bitter apple to the chair legs (I might use this as well as other measures, particularly if it’s a particularly stubborn case), but it’s just an example of something that technically fits into a quadrant but isn’t as stressful or aversive to the pupil as some would argue that all socially mediated R- and P+ have to be, due to the sole reason of being in those quadrants.
Any who, I’m pretty sure those count as socially mediated since the human sets it up?
I prefer to give each exercise it’s due in regards to careful consideration and critical thinking rather than throw the baby out with the bathwater.
Another great blog.
Lady Chauncey Barkington those are great examples and I love your approach to them.
You got it right about my point in the post: that people use the automatic negative reinforcement to justify socially mediated. And I absolutely agree with your point that just because something is natural (automatic) doesn’t make it good. If my dog has her foot caught in the fence, she might get it out by pulling (R-), but I would be there helping her before that happened, as you would too. (Hopefully we wouldn’t have a situation where that could happen in the first place.)
BTW I’ve used bitter apple as part of training one dog what is OK or not OK to chew. I didn’t use it as the only thing (and I bet you didn’t either). By the time I did apply it, my dog had quite a bit of experience with impulse control, Zen, and being given her own stuff to chew. And she wasn’t in a chew deprived environment.
That’s a really good question about whether that one is socially mediated. If so, then what about all the other things in the environment that are there because of me that might scare my dog accidentally? Something that I put there falling off a cabinet because she jiggles it, for example. Behaviorists don’t usually get inside the black box of “intent” so I think there would technically be no difference. Maybe a behavior analysis guru can weigh in. In the meantime I’ll look it up.
Thanks for great comments!
Thought provoking blog. What i’ve taken away from this is that you try to avoid using methods which could be stressful, no matter how slight, if you can find a less stressful alternative. R+ always seems to be a better choice although often requires more skill and forethought. Being human, punitive methods seem to flow naturally and need no encouragement. Arguing against your philosophy, no matter how pedantic one may percieve you to be, is condoning those who lack the skill, the forethought or the desire to aquire either. Like you, I am no saint, but for the welfare of the dogs in my care, I am at least trying to move that direction. If, as in the words of William G. Mcadoo “it is impossible to defeat an ignorant man in argument”, I guess the answer lies in education. Keep up the good work.
Thank you so very much for getting what I’m trying to say and commenting. You said it so well. Made my day!
Enormously helpful post. Everyone seems to understand, or think they understand, positive reinforcement, but negative reinforcement and negative punishment get confusing. Thanks for the clarity!
Hi Eileen – love your blog! But I must say I get very lost when the topic gets “technical” … I really do not “get” R+ and R- ….. but I do understand examples! And some of those that you describe I would not have thought of as negative – for example walking into their space.
I volunteer with an organization and we use positive reinforcement training. In the past when introducing dogs to things like hair dryers or vacuums etc … we would take them into a room off lead and actively entice the dogs towards the object – treating as they got close.
Well that approach has been changed and when rolling out the new way they wanted us to introduce things – they used the following example on us- and boy did it make sense! They used an example of bringing an aquarium into the room filled with something you didn’t like – say snakes – and enticing you to put your hand in the aquarium – if they upped the ante at some point you would be willing to put your hand in to get the reward – but you weren’t ever going to like doing it. Hence our dogs might want to come get the treat but it certainly didn’t mean they liked getting near the hair dryer!
Now what we do is bring the dog into the room – allow them to check out the room – then turn the object on and we just sit with it not saying anything to the dog – allowing the dog to approach when and if it wants to – invariably at some point the dog comes over and checks out what we’ve got – some will come check then move away and then come back around – etc …. but typically they all come over. (we’re not training dogs with any “issues” – we’re just introducing new things to fairly well adjusted puppies)
So I’m assuming the way we used to do it would have been some type of negative reinforcement based? Is that true?
Hi Gayle! Thanks for some great questions!
First let me say I think the change your organization has made is great. And the way they explained it was right on. Dogs (and people!) can get “in over their heads” if they are enticed towards something that scares them by something they really want.
Regarding your last question about whether the former approach employed negative reinforcement? The answer is that in the luring the dog up to the scary thing, probably not. Both negative and positive reinforcement have as part of their definition that the behavior has to increase. If the reason the dog is approaching the scary thing is because of food enticement, not only does that not change their feelings about the scary thing, but it doesn’t make the likelihood of their approaching it on their own increase. Usually. But whenever you have something around that the dog doesn’t like, there is the potential for positive punishment or negative reinforcement. It certainly could be happening. For example, negative reinforcement would be in play if someone was holding the dog close to the scary thing, then rewarded it for some behavior by letting it run away. The dog would likely learn to do what let them escape and that behavior would increase in that situation.
Trying to use examples since you said those were helpful.
About the walking into the dog’s space: It does vary with the dog. Think of when someone who has a different sense of personal boundaries stands too close to you and is talking right in your face and you don’t know them well. Everybody has a different boundary sense, and it’s culturally based too. But it can be a bad feeling when someone is too close, and it makes you want to back up. Dogs are all different, too. One of my dogs likes me to come into her space in just about any situation, but the other two don’t. If you watch dogs, you’ll learn that a lot of them, especially the little guys, really don’t like being loomed over and reached toward.
Back to the food and scary things. If you had a non-scary thing, say, a mat, and treated the dog every time they stepped on it, they’d end up going there a lot. That’s R+. But something aversive or scary has the potential to punish behavior or negatively reinforce it. So that’s one reason that the old method doesn’t work consistently with scary monsters. Another reason is that when dogs are scared they are not in the right frame of mind to calmly learn new behaviors.
The method your organization is using right now sounds like it employs desensitization, habituation, and some natural negative reinforcement (as the pups may approach and retreat at their own pace).
For truly scared dogs, the tried and true techniques are called desensitization and counterconditioning. You might want to look at the website http://fearfuldogs.com . Debbie Jacobs explains all this stuff better than I do! Especially look at the blog, her book, and the force free training articles.
Thanks for commenting! Hope I answered your questions.
Hi Eileen – thanks so much for taking the time to write back in so much detail!
As always very much appreciated!!
Yes – examples are best with me!
You are welcome!
hello Eileen – I wrote to you recently about starting a dog training club in NYC. This post is fascinating. Just wanted to say thanks again. Love your blog. And I am, indeed a member of the failed-bell-ringer club, which is how i got here (followed a link from the bell-ringing post). thanks again.
Hi again! Thanks so much. And good luck with the bell ringing, if you choose to give it another try.