I frequently read variations of the following online:
People shouldn’t object to the use of negative reinforcement! It’s just stuff like washing my hands when they are dirty or drying them when they are wet. What’s wrong with that?
This is a fairly common defense of using negative reinforcement (R-) in training. The defender points out that R- is common in life and trots out a benign-sounding example or two.
Here’s a quick review of the definition of negative reinforcement:
In negative reinforcement, a behavior is strengthened by the removal, or a decrease in the intensity of, a stimulus.—Chance, 2013, p. 134.
In lay terms: something is bugging, annoying, scaring, or hurting me. I behave in such a way that I escape from or stop that aversive stimulus. My behavior is reinforced by the cessation of the aversive. I am likely to perform it again in similar situations.
Negative reinforcement happens under the radar all the time. Think of all the times we scratch an itch, shift in our seats, take off or put on clothing to be more comfortable, and perform other small movements to relieve discomfort. Not to mention the more intense instances when we escape or avoid things that are threatening or hurting us.
(Throughout this post I am using the convention of describing certain scenarios involving aversives as negative reinforcement. But keep in mind that we never know whether any reinforcement process has occurred until we see a behavior increase or maintain.)
Hand washing is a good example of the everyday kind of negative reinforcement. The analysis looks like this.
- Antecedent: There is dirt on my hands
- Behavior: I wash my hands
- Consequence: Clean hands
Problem solved. Negative reinforcement doesn’t sound so bad then, right? Why should I and others argue against using it in training?
Automatic vs. Socially Mediated Negative Reinforcement
One way to classify negative reinforcement scenarios is based on how the aversive is removed.
Instances where we take action with a behavior that directly removes the aversive are called automatic or natural negative reinforcement.
[In] automatic reinforcement, the consequence is produced directly by a response independent of the actions of another.—Cooper, Heron, and Heward, 2014, p. 315.
The consequence in the hand washing case is having clean hands. The person with the dirty hands washes them. This behavior directly removes the aversive stimulus of the hands being dirty.
But when a trainer uses an aversive in training to reinforce specific behaviors, it is no longer automatic negative reinforcement. She has inserted herself into the process. This version is called socially mediated or contrived negative reinforcement.
[In] socially mediated reinforcement, the consequence results from the action of another person.—Cooper, Heron, and Heward, 2014, p. 315.
No longer can the human or animal respond with a behavior that directly relieves the discomfort. The trainer uses the aversive by putting a contingency on escaping it. She decides what behavior will stop the aversive stimulus. It may be something completely unrelated to what the natural escape response would be. In fact, that natural escape response is usually prevented.
Cooper, Heron, and Heward write that automatic vs. socially mediated negative reinforcement is a classification by source. In other words, in socially mediated negative reinforcement, the third party (trainer, in our case) is the source of the negative reinforcement. Many people who use negative reinforcement try to dodge this truth. They argue that since they use aversives that are already in the environment, that exempts them from the ethical problems with using negative reinforcement scenarios. Yes, the aversive may already be there. But if they are removing the aversive contingent on a certain behavior, they are the source of the negative reinforcement.
One day I watched Zani hop onto the chaise lounge in the backyard, take a couple of steps around on it, and hop off again. It was 100º Fahrenheit out and the vinyl was hot to the touch. Then Clara approached it and I pulled out my camera, expecting her to jump off as well. I thought I was going to capture an example of automatic negative reinforcement. Instead, she settled down and stayed there for six minutes. She then got up, apparently not out of discomfort, but because Summer barked at something. I realized I wouldn’t have known that Zani would be more sensitive to the hot plastic than Clara. (See #4 below.)
Equating socially mediated, training-centered negative reinforcement with automatic negative reinforcement is inaccurate and disingenuous.
In the movie linked here, I demonstrate five differences between socially mediated negative reinforcement and automatic negative reinforcement. In socially mediated negative reinforcement:
- A third party controls access to the reinforcer and sets a contingency on escaping or stopping the aversive stimulus.
- The animal doesn’t generally escape the aversive one time and get to move on and do something else. The trainer usually reapplies the aversive, exposing the animal to it multiple times.
- The trainer forces the animal to stay in the area. She will generally prevent the animal from performing the natural escape response. For instance, gun dog trainers who teach a force fetch with an ear or toe pinch often have the dog tethered on a bench. People who use negative reinforcement in exposure to fear-inducing triggers have their dogs on leash or in an enclosed space.
- The trainer can’t know exactly how much discomfort she is causing the animal. She has interrupted the natural sequence for the animal of “feel discomfort—do something about it.” She may cause the animal to endure a much larger magnitude of the aversive than it would have in automatic negative reinforcement. As I mentioned above—I never would have predicted that Zani was uncomfortable with the briefest touch of hot vinyl if I had first seen Clara be fine with it.
- The behavior required to escape the aversive can be anything at all. The animal often has to figure it out on the fly, while in the presence of the aversive.
So Is Socially Mediated Reinforcement Bad in General?
The automatic and socially mediated classifications apply to both negative and positive reinforcement. I’ve pointed out some dishonesty in the ways socially mediated negative reinforcement is presented by some who use it. Does that mean I disapprove of socially mediated positive reinforcement?
Of course not. All our training is socially mediated. We are seeking to change an animal’s behavior. The choice between positive and negative reinforcement is generally a choice between something that is fun for the learner and something that is, at best, a drag. Intervening in a dog’s spontaneous behavior carries ethical imperatives. We need to make it fun and pleasant for the dog. We also should be honest about the times we might use an aversive stimulus instead of telling fantasy stories about it
Well Then, Is Automatic Negative Reinforcement Fine?
I’m painting a pretty hard picture of socially mediated negative reinforcement. Does that mean that automatic negative reinforcement is fine?
That’s not what I’m saying. Automatic negative reinforcement scenarios can run the gamut from barely noticeable, such as when we shift in our chairs, to life-threatening, when we run for our lives. Automatic negative reinforcement is a way all animals learn new behaviors, and quite often not a fun one. It’s just that it is irrelevant to training. As soon as we train, we are in the world of socially mediated negative reinforcement.
It’s simply disingenuous to equate low-level automatic negative reinforcers with training with negative reinforcement.
In the movie, I show an example of an automatic negative reinforcement scenario with a very low-level aversive stimulus. Something you wouldn’t think twice about if it happened to you. Then I show what happens when that same minor aversive is applied in a socially mediated negative reinforcement scenario.
I recently updated this article to use the current ABA terminology. I’m not remaking the movie, though. So you will hear me saying natural (same as automatic) and contrived (same as socially mediated) negative reinforcement.
I haven’t discussed the fallout from the use of aversives in this post. I do in several other posts and pages. (Yeah, I know, I usually won’t shut up about it.) But do take a look at the movie and consider how you would feel about the person who had the remote control in her hand and had tethered you to a chair.
- Chance. (2013). Learning and Behavior. 7th edition. Cengage Learning.
- Cooper, Heron, and Heward. (2014). Applied Behavior Analysis. 2nd edition. Pearson.
- Fallout from the Use of Aversives
- Discussing Negative Reinforcement Responsibly
- How Did the Aversive Get There? A Call for Honesty
- Let’s Talk about Aversives in Training. I’ll Go First!
- Leaving the Scene: Clarifying the Science of Negative Reinforcement
- What Does Shower Mold Have To Do With Dog Training?
© Eileen Anderson 2015, 2017