How Rewards Can Destroy Motivation

In 1972, a researcher named David Greene stood in a classroom at a Stanford-area nursery school and watched a group of three- and four-year-olds drawing with felt-tip markers. They were absorbed, unhurried, drawing because drawing was fun. Some of these children had been promised a Good Player certificate if they drew; others had been promised nothing. Two weeks later, Greene came back and counted how much each child chose to draw during free time, when no certificates were on offer and nobody was watching for them. The children who had once drawn for the certificate now spent noticeably less time with the markers than the children who had only ever drawn for the joy of it.

Nothing had been taken away from these children. They had been given a reward, a small and pleasant one, and it had quietly drained the activity of the very pleasure that made it worth doing. The markers had become work, and work, the children had learned, is something you do because you get paid.

That uncomfortable finding sits at the center of one of the more counterintuitive results in psychology. We tend to assume that rewarding a behavior makes people want to do it more, and often that is true. But under specific and surprisingly common conditions, paying someone to do what they already love can leave them wanting to do it less. This article is about when that happens, why, and what it means for parents, teachers, and employers who reach instinctively for the carrot.

A Surprising Crack in the Standard Picture of Reinforcement

The intuition that rewards strengthen behavior is not naive folk wisdom but a serious scientific claim with a long pedigree. The classical operant view, associated above all with B. F. Skinner and his experiments in which animals learned to press levers for food, holds that reinforcing a behavior increases its frequency, and that the schedule on which the reinforcement is delivered shapes how persistent that behavior becomes. Reward a rat for pressing a lever, and it presses more. The principle generalizes widely and underwrites much of how we structure schools, workplaces, and the training of pets and children alike.

The overjustification effect is the complication that operant theory did not predict. It applies specifically to behaviors that people already engage in for their own sake, the things they would do unpaid, unprompted, and unsupervised. For these behaviors, introducing an extrinsic reward can paradoxically reduce the behavior once the reward is withdrawn. The lever-pressing rat had no prior love of lever-pressing to lose. A child who already adores drawing does, and that prior love turns out to be fragile in a way the standard reinforcement picture never anticipated. The puzzle is not whether rewards work, because frequently they do, but why they sometimes backfire precisely with the activities we would most like to encourage.

A Cube, a Puzzle, and the First Clean Demonstration

The first rigorous experimental demonstration of the effect came not from a nursery school but from a laboratory. In 1971, Edward Deci ran a study with undergraduate participants and a then-popular interlocking puzzle called the SOMA cube, a set of irregular blocks that can be assembled into various shapes, the kind of task many people find genuinely engaging on its own.

Deci had participants work on the puzzle across several sessions, and in one of them some participants were paid for each configuration they solved while others were not. The crucial measurement came later, during a free-choice period in which participants were left alone in a room with the puzzle and some magazines, with no instruction and no payment on the table. The question was simple: when nobody is paying you and nobody is watching, do you reach for the puzzle? Participants who had earlier been paid spent less of that free time on the puzzle than those who had never been paid. The payment, which during the paid session had presumably motivated them perfectly well, had reduced their later willingness to do the task for nothing, and Deci had shown under controlled conditions that extrinsic payment could erode intrinsic motivation.

The Nursery School Study That Made It Famous

Two years later, in 1973, Mark Lepper, David Greene, and Richard Nisbett published the study that became the canonical demonstration of the effect, the one most people encounter first. Working at the Bing Nursery School, they began by identifying children who already loved drawing with markers, children who would clearly choose the activity for its own sake. Then they randomly assigned these children to three conditions, and the design of those conditions is what makes the study so illuminating.

In the expected-reward condition, children were told beforehand that they would receive a Good Player certificate for drawing, so the reward was promised in advance and explicitly tied to doing the activity. In the unexpected-reward condition, children drew without any promise and then received the very same certificate afterward as a surprise. In the no-reward condition, children simply drew, with no certificate at any point. Two weeks later, the researchers measured how much each child spontaneously chose to draw during free play.

Only the expected-reward children showed reduced spontaneous drawing. The children who had received an identical certificate as an unexpected surprise drew just as much as the children who had received nothing. This is the detail that rules out lazy explanations. It was not the certificate itself, not the paper or the praise or the act of being singled out, that did the damage. It was the prior contract, the sense that drawing was now something one did in order to obtain a reward. The reward had to be expected and contingent to corrode the motivation.

Why a Reward Rewrites the Story We Tell Ourselves

What, then, is actually happening? The mechanism is at least partly cognitive, a matter of how people explain their own behavior to themselves. Human beings are relentless interpreters of their own actions, silently answering the question, why am I doing this? A child who draws for pleasure carries an implicit answer: I do this because I enjoy it. Introduce an expected, contingent reward, and a competing answer becomes available and even more obvious: I do this for the certificate. When the reward then disappears, that second explanation no longer applies, and the first, genuine one has been crowded out and weakened, leaving the activity without a compelling reason to continue.

A richer account comes from self-determination theory, a framework developed in large part by Deci and his collaborator Richard Ryan. The theory proposes three basic psychological needs that sustain healthy motivation: autonomy, competence, and relatedness. The overjustification effect, on this view, works by undermining the first two. An expected, controlling reward shifts what is called the perceived locus of control from inside the person to outside, so the behavior comes to feel externally driven rather than self-chosen, which undermines autonomy. The reward can also carry an implicit message that the person's own competence requires external validation, that they need a certificate to confirm they are doing something worthwhile, which undermines competence. Strip those supports away and the intrinsic motivation that depended on them weakens too.

The Conditions Under Which Rewards Stay Safe

It would be a serious misreading to conclude that rewards are simply poison for motivation. They are not, and the overjustification effect is sharply bounded, so understanding its limits matters more than memorizing the headline.

Several conditions reliably keep extrinsic rewards from undermining intrinsic motivation. Rewards do not undermine when the activity was uninteresting to begin with, because there is no prior intrinsic motivation to lose. They tend not to undermine when they are informational rather than controlling, conveying useful feedback about how well someone did rather than functioning as a lever to direct behavior. They are far less corrosive when unexpected rather than promised in advance, exactly as the surprise certificates in the nursery school study demonstrated. And they tend to support rather than sabotage motivation when they are tied to the quality of performance rather than to mere participation, since a reward earned for doing something genuinely well can reinforce a sense of competence instead of replacing it. The damage comes most reliably from a specific recipe: a reward that is expected, contingent on the activity itself, controlling in tone, and unconnected to how well the work is done.

What Happens When the Theory Meets the Real World

Laboratory effects do not always survive contact with classrooms and street corners, so it is worth looking at where the framework has been tested at scale. Two bodies of work stand out, and together they show both the reach and the messiness of the idea.

Between 2007 and 2010, the economist Roland Fryer ran large randomized controlled trials in New York City, Chicago, Dallas, Houston, and Washington, paying students for academic outcomes. The design distinguished between paying for outputs, such as test scores, and paying for inputs, such as reading books, attendance, or good behavior. Paying for outputs generally did not work, which fits the framework, since test scores are a distant and hard-to-control target and the reward easily becomes a controlling pressure detached from the actual learning. Paying for inputs produced modest positive effects, plausibly because the rewarded actions were concrete steps within a student's control and functioned more like scaffolding than bribery. The results are consistent with the theory while showing that the empirical picture is more nuanced than any universal slogan allows.

The sharper real-world demonstration comes from a 2000 study by Uri Gneezy and Aldo Rustichini involving Israeli high-school students who went door to door collecting charitable donations. Some were paid nothing, doing it for the cause and the sense of contribution; others were paid a small fraction of what they collected. The students paid a small commission collected less money than those paid nothing at all. A modest payment had displaced the intrinsic and social motivation to do good, replacing it with a transaction that, at that price, simply was not worth the effort. Paying volunteers, the study showed, can push their effort below the unpaid baseline, the canonical demonstration of overjustification outside the classroom.

Designing Rewards Without Killing the Thing You Want

The practical reach of this framework is broad. It touches parents who pay children for chores, teachers who pay students for grades, employers who attach bonuses to work their staff already find meaningful, and organizations that try to incentivize volunteers. In each of these settings, the instinct to reach for an extrinsic reward can quietly undercut the very motivation it was meant to strengthen.

But the lesson is not the simplistic one that rewards are always bad and should never be used, a conclusion that ignores everything the boundary conditions tell us. The honest takeaway is that extrinsic rewards are a powerful tool with a specific and predictable failure mode. When you attach an expected, controlling reward to something a person already loves to do, you risk converting play into work and watching the love drain away once the reward stops. When the activity was tedious to begin with, when the reward carries genuine information about a job well done, or when it arrives as an unanticipated thank-you rather than a pre-arranged price, the danger largely disappears. The skill, for any parent or teacher or manager, lies in knowing which situation you are in and designing accordingly, rather than assuming that more reward always means more of the behavior you want.

Key Takeaways

The overjustification effect describes how extrinsic rewards, when attached to behaviors people already pursue for their own sake, can paradoxically reduce those behaviors once the rewards are withdrawn, a complication the classical Skinnerian view of reinforcement did not predict. Edward Deci's 1971 SOMA-cube study established the effect first, showing that paid participants later chose to do the puzzle less in free time, and the 1973 Lepper, Greene, and Nisbett nursery-school study made it famous by demonstrating that only an expected, contingent certificate, not an identical surprise one, reduced children's spontaneous drawing. The mechanism is partly cognitive, as a person's self-explanation shifts from "I do this because I enjoy it" to "I do this for the reward," and partly motivational, as self-determination theory describes the reward undermining the basic needs for autonomy and competence. Crucially, the effect is bounded: rewards do not corrode motivation when the activity was uninteresting to start, when they are informational rather than controlling, when they are unexpected, or when they are tied to quality of performance. Real-world tests, from Roland Fryer's pay-for-performance school trials to the Gneezy and Rustichini finding that lightly paid charity collectors raised less than unpaid ones, confirm both the effect and its nuance, leaving us not with the rule "never reward" but with the harder task of understanding when rewards support intrinsic motivation and when they quietly destroy it.