This Just In: Reinforcers Reinforce Behavior! (A Summary of Our Research On Rewards, Contracts, and Praise)

This Just In: Reinforcers Reinforce Behavior! (A Summary of Our Research On Rewards, Contracts, and Praise)

We’ve been grappling with the issue of rewards for years. They seem straightforward, helpful, and obvious to us, but we often meet criticism, specifically by readers of Alfie Kohn. His voice is the strongest of the few researchers who assert that rewards, even verbal, should be abandoned; that they damage “intrinsic motivation” by turning the activity into something that is motivated only by an outside reinforcer.

Here we share the most recent meta-analysis on the subject, its conclusion, and our summary (TL:DR). We’ll start with our short conclusion first:

Kohn is wrong. Reinforcers reinforce behavior, rewards make excellent reinforcers, and researchers have found no damaging effects when they are used as prescribed.

Now, a lengthier conclusion, excerpted from the meta-analysis (I’ve added the formatting/bold for emphasis):

Behavior analysts have spent their time exploring the variables that affect the efficacy of reinforcement. They have always admitted that under certain circumstances reinforcement is more effective (e.g., in the presence of an establishing operation; Michael, 1993) or less effective (e.g., when discriminative stimuli are not present to signal the availability of reinforcement). Cognitive researchers have focused on those instances when reinforcement is less effective, and attributed this lack of efficacy to a general problem with programmed reinforcement, claiming that such reinforcement alters an inner propensity called intrinsic motivation. In examining the methodology of these experiments, however, it often becomes clear that the reinforcement programs are simply bad programs — that is, they do not exploit those strategies that we know make for effective reinforcement programs. When tangible rewards are not delivered immediately after behavior, when an individual’s baseline performance is not taken into account in intervention design (as in when students who are already performing a task at a high frequency are put on programmed reinforcement), and when generalization strategies are not used, it is hardly surprising that “intrinsic motivation” is lowered. The logical solution is not to eliminate programmed reinforcement, but to use effective programmed reinforcement strategies.

Bribery is defined in the dictionary as an inducement to engage in illegal or inappropriate behavior (Woolf, 1980). When education personnel, including school psychologists, extol the use of extrinsic reinforcement in the classroom, the motive is clearly not to “bribe” children and youth, but to increase appropriate academic and social behavior. The goal is obviously not to decrease intrinsic motivation, although it is unclear that the construct exists or is useful in the science of psychology. It is apparent through an examination of the data that any decrease occurs under specifically circumscribed conditions, conditions that are easily avoidable. Best practice would suggest that children and youth deserve interventions based on sound, empirical findings. The positive effect of the use of reinforcers in the classroom is one such conclusion.

If teachers are implementing “reinforcement” programs without knowing how to do so, the worries of intrinsic motivation researchers seem reasonable. But it is the practice and not the principle that is suspect and open to misapplication and abuse, and the corresponding prophylactic is more teacher training in behavioral methods, not less. In the meantime, we can only try to correct the misconceptions that have led to unwarranted criticism of programmed reinforcement, and take some solace in the fact that classroom teachers will continue to learn from contingent consequences what works and what does not.

“Extrinsic Reinforcement in the Classroom: Bribery or Best Practice” by
K. Angeleque Akin-Litte, Tanya L. Eckert and Benjamin J. Lovett, and Steven G. Little. 
First published in School Pyschology Review, 2004, Volume 33, No. 3, pp. 344-362.

Finally, our summary of the information found in this meta-analysis and elsewhere, in lay terms:

In the beginning (1950s), there was B.F. Skinner, who demonstrated that behaviors can be “reinforced” or “punished” by the introduction and removal of various stimuli. If you “pay” the pigeon with food when it pecks the light, that behavior will increase.

As more researchers studied behaviorism, they found lots of “quirks” in human behavior. Some of the findings were surprising and counter-intuitive. For example, one study found that if a school fines parents for picking-up their children late, more parents will do so. Another found that paying people to give blood can cause a decrease in that behavior. Researchers started studying the “best” methods to reinforce behaviors in different environments. The public began growing wary of the idea of professional, scientific, behavior “manipulation”.

In the 70’s, Edward Deci found more of these quirks. If you have a group of young people who are enjoying solving puzzles, and then you pay them for it, they will actually spend less time solving puzzles than the people who aren’t paid. He did about 3 studies in this area, (the example above is one), and concluded that rewards decrease behaviors in humans.

The word spread quickly: “Stop the presses! Behaviorism is wrong!” Humans have a more complex mechanism in their minds, Deci claimed, called “intrinsic motivation”, that can only show itself in the absence of outside influence. Children are a bit like the potted cactus you have in the window: You’d better not feed it, water it, or move it! It will motivate itself to grow, and any other interference is damaging to that motivation.

Deci wasn’t entirely wrong. His experiments were reproducible, in fact, behaviorists had found similar effects prior to his work. It’s true if you pay someone for something they are already enjoying doing, they will enjoy it less. A lot of us are already aware of this and other Tom Sawyer-esque motivational “hacks”. It’s the reason I had to quit cooking for pay; I enjoy cooking too much. The behaviorists were able to offer many alternative hypotheses to explain Deci’s results. But Deci insisted that this mechanism, “intrinsic motivation”, was outside of the realm of the study of behavior. As an internal “cognitive” state, it can’t really be observed or defined, and is defined only as behavior existing in the absence of reinforcers.

The cracks were beginning to show in Deci’s findings. To this day, there is no consensus on the definition of “intrinsic motivation”. As Stephen Flora points out, psychology, like any science, is the study of observable phenomena. But there is a trend in Western attitudes and beliefs that tells us that our noble behavior should arise without extrinsic motivation, and that the person paying for what he wants is using a form of manipulation and control. This may explain why Deci and Kohn’s conclusions are so tempting for parents and teachers. We want to believe that our heaven-sent children will do right, no matter the consequences. Another hypothesis: parents and teachers are often just too lazy to study and create consistent systems of rewards.

Behaviorists continued to poke holes in Deci’s framework, from the definition, to the methodology, to the conclusion. They continued to show, as did Skinner, that of course reinforcers reinforce behavior, by definition! A reinforcer is a stimulus that increases a behavior. There are certainly ways to get someone to spend more time on a task. Every good manager, teacher, parent, already knows this. But Deci and Kohn repeatedly ignore the difference between the definitions of a reinforcer and a reward. So when a certain reward doesn’t work as a reinforcer, they essentially conclude “rewards don’t work”.

(Click here for our short, scathing review of Kohn’s book: “Beyond Discipline: From Compliance to Community”)

You’ll notice we’re focusing on Deci and Kohn. That’s because their names comprise almost the entirely list of researchers who conclude that rewards don’t work. Contrast that with the hundreds of researchers with consistent, opposing findings. Of the the flaws in the methodology of the cognitive researchers’ work, the greatest may be the failure to study effects within-groups, instead focusing on between-group differences. An example of the difference:

Imagine you’re in a room with 30 kids, and 30 dirty dishes, and I say, “Do something that will motivate them all to wash one dish”. Perhaps you try offering them a video upon completion, and perhaps it doesn’t work very well. If another group had better success without a video, that would be a between-group finding.

Now imagine you have just two kids, and two dirty dishes, 30 days in a row. In the beginning you struggle to find the motivator, but with the right negotiating and incentives, you find a solution that works for them, a way to reinforce the dish-washing behavior. Congratulations, you have solved the problem within the group you were focusing on. This is the reality of parenting and education, and the methodology behaviorists use to solve problems of motivation.

Somehow, and this has also been studied by researchers, Kohn’s evidence-thin, emotionally-laden accusations became mainstream. In 1991, the National Education Association told teachers to never use rewards, as they destroy intrinsic motivation. And this has been a common idea among teachers and parents ever since, not just in our experience, but among the researchers published in this meta-analysis.

Deci made a comeback in 1999 with his own meta-analysis. He found that even verbal rewards can have a negative effect on intrinsic motivation. According to the writers of the meta-analysis being summarized here, “the assertion… is in direct opposition to the available data”. Also “none of the studies that utilized single-case [within-group] design found any detrimental effects with the use of reinforcement contingencies”. Keep in mind, we’re talking about reviews of over 100 studies. This horse died a long time ago, or as the authors of this meta-analysis more tactfully put it: “Perhaps the time has come to accept these findings and cease attempts to damage these data by offering alternative, invalidated conclusions”.

Skinner wasn’t wrong. Reinforcers reinforce. They don’t damage your children or their love of learning. There are no long-term negative effects. And as for the most important metric for parents who follow the NAP: paying someone for their work is not immoral. For most children, there is no “intrinsic motivation” for learning algebra or cleaning the toilet. So we can choose to motivate them to succeed and develop healthy habits, or we can let them navigate the world by their own whims. But kids can’t raise themselves, so failing to prepare and plan appropriate environments that encourage them to succeed, is nothing more than planning for them to fail.

We have work to do, shaping the world that our children will live in. That means finding what works for them, what motivates them, and using that to inspire them to greatness, knowledge, health and wealth.

Links to two of the meta-anlysese cited here


Leave a Reply

Your email address will not be published. Required fields are marked *