Updated on
August 27, 2026
The overjustification effect describes a possible fall in intrinsic interest after an expected external reward is removed. Explore the classic studies, disputed evidence and limits of educational application.
What is the overjustification effect?
The overjustification effect is a possible reduction in intrinsic interest after an expected external reward is attached to an activity that a person already finds interesting and the reward is later removed. It is conditional rather than universal: findings vary with reward type, expectancy, contingency, prior interest and how motivation is measured.
The overjustification effect is a proposed reduction in intrinsic motivation after an expected external reward is introduced for an activity that a person already finds interesting. The clearest prediction concerns what happens after the reward is removed. The person may then spend less free-choice time on the activity than someone who was not offered the reward.
Lepper, Greene and Nisbett (1973), for example, studied 51 preschool children. The study compared an expected award, an unexpected award and no award.
The effect is part of a wider debate about interest, rewards and behaviour. Findings vary with the task, reward, comparison group and measure. It is a conditional research claim, not a universal law or classroom programme.
Intrinsic motivation comes from the task itself. A person may keep going because it is fun or worth doing. A reward adds a reason, such as money, a prize or points. The hypothesis says this outside reason can change how the person explains their choice.
Three conditions matter. The task must already attract free choice. The reward must be clear and usually expected.
Interest must be tested after the reward rule changes or ends. A dull required task is not a clean case. Nor is faster work while points are on offer.
The term is narrower than the split between intrinsic and extrinsic motivation. People can value an outcome and enjoy the task at once. Self-determination theory also sets out several forms of regulation, not a simple inner-or-outer switch.
Researchers use several signs of intrinsic motivation:
These measures are not the same. A learner may do more work for a reward, then choose the task less often. Another may report the same level of fun while doing more work.
Deci's 1971 paper used Soma puzzles with undergraduates, not drawing tasks with children. In the best-known test, people attended three sessions. One group got one dollar for each puzzle shape in session two. A control group did not.
When the researcher left, both groups could use the puzzles or choose other tasks.
Free-choice time with the puzzles was the measure of intrinsic motivation. After pay ended in session three, the paid group used the puzzles less than before, relative to the control pattern. Deci read this as evidence that task-linked money could reduce interest in a task that was interesting at first (Deci, 1971).
The test was short and used one puzzle with undergraduates. It did not test attainment, a school term or the same response in all people.
Lepper, Greene and Nisbett studied 51 preschool children who already liked drawing with felt-tip pens. The children were placed in an expected-award, unexpected-award or no-award group. The first group was promised a certificate for drawing. The second group got the same award with no promise.
One to two weeks later, staff logged free drawing during normal play. Children promised the award drew less than those in the other groups. The authors said the expected award gave a clear outside reason for an act that had seemed self-chosen (Lepper, Greene and Nisbett, 1973).
The study tested an expected award and later free drawing. It did not test all praise, curriculum learning, long-term results or class policy. The preschool drawing result belongs to Lepper and colleagues, not Deci.
The studies show patterns under set conditions. They do not reveal one clear cause. Several accounts seek to explain what a reward means to the person who gets it over time.
The first account draws on self-perception theory. People may infer why they acted by looking at the setting. If a clear reward is present, they may give it more weight and give prior interest less weight. This is one account of a shift in cause.
It is not a direct view of each person's thoughts.
The idea is linked to cognitive dissonance, but it is not the same. Work on insufficient justification asks what people infer when an outside reason seems too small. Overjustification asks what can happen when an outside reason is so clear that it masks an inner one.
Cognitive evaluation theory later became part of self-determination theory. It explains reward effects through felt choice and skill. A reward that feels controlling may make an act feel less self-chosen. Feedback that shows progress without pressure may have a different effect.
Ryan, Mims and Koestner (1983) found that both the reward rule and the social setting mattered (Ryan, Mims and Koestner, 1983).
This account does not make praise safe by default. Praise can help, harm or have no clear effect. Its effect can vary with tone, trust, pressure and social rank. Henderlong and Lepper (2002) set out these limits (Henderlong and Lepper, 2002).
Behaviour analysts have asked whether intrinsic motivation is needed as a cause in this account. A change after reward withdrawal can reflect exposure to the task, contrast between study phases, other choices or the real value of the reward. A preferred item is not a reinforcer unless its use after an act increases that act.
This issue links the debate with operant conditioning, but the ideas stay distinct. Operant work asks how an outcome changes an act. Overjustification work often asks if later free-choice time falls after a reward phase. Roane, Fisher and McDonough (2003) found a result against the hypothesis in one single-case study and tested other accounts (Roane, Fisher and McDonough, 2003).
Reviews do not give one agreed answer. They group rewards and outcomes in different ways. Results also change with the studies included, the test of prior interest and the choice to join or split free-choice and self-report data. The dispute is about both the findings and the methods.
On a smaller screen, swipe across to read all columns.
| Source | Evidence considered | Main finding | Boundary for interpretation |
|---|---|---|---|
| Wiersma (1992) | Review of work and job studies | Results differed by measure. Free-time task use fell in some tests, while work scores showed added effects. | A change in free choice is not the same as a change in work quality. |
| Cameron and Pierce (1994) | 96 studies and four measures | No overall fall was found. Praise raised scores. Expected tangible rewards for doing a task had a small negative free-choice effect. | The result depends on reward type and how outcomes are coded. |
| Tang and Hall (1995) | 50 studies and 256 effect sizes | The effect appeared in several tests where theory forecast it, but not in all tests where no gap was forecast. | The effect was conditional, not a general result of reward. |
| Deci, Koestner and Ryan (1999) | 128 studies | Expected tangible rewards reduced free-choice time under several reward rules. Effects on reported interest were smaller. Positive feedback raised both measures. | Free-choice behaviour and reported interest must stay separate. |
| Cameron, Banko and Pierce (2001) | New review of more than 100 tests | Negative effects were not broad. They centred on expected tangible rewards for high-interest tasks under set reward rules. | Study choice and grouping help explain the dispute with Deci and colleagues. |
| Peters et al. (2022) | Repeat of Deci's puzzle design with 24 undergraduates | Both groups used the puzzles less over time, and each person's pattern varied. The group result did not support the hypothesis. | One small repeat cannot settle the field, but the first pattern is not automatic. |
| Cerasoli, Nicklin and Ford (2014) | 183 samples and 212,468 people across school, work and sport | Intrinsic motivation had a stronger link with quality. Rewards had a stronger link with the amount of work. The two were not always at odds. | This review covered work scores, not just free choice after a reward. |
Overjustification Effect Study Notes
This one-page note shows the steps in the claim and the factors linked with a later fall in interest.
Direct answer: The effect is a possible fall in later interest. It can occur after a person gets an expected reward for a task they already like, then loses that reward. It is conditional, not universal.
Basic sequence: the task is already interesting; an expected reward starts; the reward ends; later free choice may fall.
More likely conditions: an expected tangible reward, a reward for taking part or finishing, a task with prior interest, and a test after the reward ends.
Less likely conditions: an unexpected reward, useful feedback, a task with low prior interest, or a test of current output or quality.
Evidence limit: early findings and later reviews do not agree. Reward type, advance notice, reward rules and the measure all matter. The note draws on Deci (1971), Lepper et al. (1973), Deci et al. (1999), Cameron et al. (2001) and Peters et al. (2022).
Free-choice engagement asks if a person goes back to a task when it is not required and no reward is on offer. It is a measure of free choice over time. The early studies used it, but boredom and repeated use can affect it. So can the other tasks on offer and the sense of being watched.
In school, a learner who stops drawing during a break after an art prize ends would resemble this measure. That choice would not show that the learner had lost art knowledge, made worse work or would avoid art in another setting. A fair reading needs a useful baseline and comparison, not one event.
Reported interest asks people if they find a task fun or worth doing. It can show an experience that task counts miss. Yet wording, memory and a wish to please can shape the answer. Reviews often find smaller and less stable effects here than in free-choice behaviour.
A learner may say that reading is still fun while choosing a different free-time task. These signs do not clash. One is a judgement. The other is a choice among options at one point in time.
Rewards can raise task completion now even if later free choice falls. These are two separate claims. The wider behaviourism in learning literature and Skinner's theories deal more directly with how a result changes an act.
Performance also has more than one form. Cerasoli and colleagues found a stronger link between intrinsic motivation and quality. Rewards had a stronger link with the amount of work. A class may complete more questions for tokens, but this does not show a change in interest, sound thought or later free choice.
Initial interest: the claim fits best when people already choose the task. A reward can raise work on a dull task without displacing much prior interest. Tests should therefore check prior free choice rather than assume it.
Expected or unexpected reward: a promised reward can be a reason to act before the task starts. An award given later does not set the same rule in advance. This split was central to the Lepper study, but it does not prove that surprise rewards are always safe.
Tangible reward or verbal response: money, prizes, awards and points are not the same as spoken feedback. Praise also varies. Feedback can show progress, while praise used to control or rank can add pressure. Henderlong and Lepper (2002) also stress trust, fair goals and the meaning given to the praise.
Reward rule: a reward for starting, finishing, meeting a set goal or beating others creates a different deal in each case. Reviews disagree about some rewards tied to work quality. The broad label reward-based learning is too vague for a sound claim.
Context and duration: short lab sessions do not copy a school term. Lessons involve staff ties, test rules, peer rank, prior skill and tasks that learners must do. The evidence gives schools a question to test, not a way around those facts.
The evidence base joins small lab studies, field tests and later reviews that use different rules. Peters and colleagues' 2022 repeat of Deci's puzzle design did not find the forecast fall for the reward group. This is key evidence against a fixed effect. Yet one small repeat cannot settle each reward type, group or outcome.
Claims about wider use also need care. A short puzzle test with undergraduates and a preschool drawing study do not show effects on grades, SEND support, behaviour plans, token schemes, long-term results or subject choice. Each claim needs direct evidence from that setting.
Schools can infer that the meaning and form of a reward matter. They can split short-term compliance from later free choice. More work done is not proof of more interest. This fits careful use of formative and summative assessment, since feedback on learning is not the same as a prize for taking part.
Schools cannot infer that all rewards are harmful, that praise is always safe, or that removing each reward will create intrinsic motivation. They cannot infer one learner's motives from one change in behaviour. They also cannot use a short lab result to predict attainment, wellbeing or long-term subject choice without more evidence.
Consider a reading scheme that awards points for completed books. If learners read more while points are available, the scheme has changed current behaviour. To test overjustification, the school would need proof of prior interest.
It would also need a fair comparison and a later measure of free reading after the reward changed. Even then, the result would apply to that context, not all reading lessons.
The effect is most useful as a warning not to merge all aims into one reward plan. A teacher may use feedback to show progress, a result to support a routine, a grade to report attainment and a prize to invite people to take part. Those aims are not the same.
In art, a prize for any drawing may send a different message from clear feedback on colour mixing after learners have worked. In maths, points for each completed question may raise the amount of work while saying little about the quality of each plan. In reading, public rank may change peer status as well as motivation. These are claims to test, not rules about how each learner will respond.
School leaders should keep this debate apart from wider claims about rewards and behaviour policy. A result that raises a target act meets the behavioural test for reinforcement. That does not settle its effect on interest, trust or transfer. A reward that does not raise the act is not a reinforcer just because an adult meant it to be one.
The same care applies to claims about independence. Helping learners notice their own progress may aid metacognition, but self-assessment is not a proven cure for overjustification. Evidence about growth mindset interventions also cannot prove a claim about rewards. Each framework has its own evidence base and role within the wider learning theories literature.
No. The best answer is conditional. Negative effects are most common when an expected tangible reward is tied to a task that people already enjoy. Tests of later free choice tend to show the clearest effect. Other settings show smaller, no or positive effects.
Rewards can also raise current work or output.
Praise is not one uniform act. Reviews find that it can help, harm or leave motivation unchanged. Tone, trust, control, social rank and the meaning given by the person all matter. Clear feedback about work is not the same as broad approval used as a prize.
No single study can rule out an effect in all settings. Peters and colleagues repeated Deci's puzzle task with 24 undergraduates. Both groups used the puzzles less, and each person's pattern varied. The group result did not support the claim.
This key test shows why the first pattern is not a fixed rule.
It is a useful sign when people have real choices. It is not a full measure of motivation. Repeat use, the appeal of other tasks and the study rules can affect it. Reports of interest, work quality and later acts give other forms of evidence.
No. Grades can report results, select people, set rank and act as rewards at the same time. This research does not split all those roles in real school systems. A claim about grade policy needs direct work on grades, learning, fairness and use. Puzzles and drawing awards are not enough.
Cameron, J., Banko, K. M. and Pierce, W. D. (2001). Pervasive negative effects of rewards on intrinsic motivation: The myth continues. The Behavior Analyst, 24, 1-44.
Cameron, J. and Pierce, W. D. (1994). Reinforcement, reward, and intrinsic motivation: A meta-analysis. Review of Educational Research, 64(3), 363-423.
Cerasoli, C. P., Nicklin, J. M. and Ford, M. T. (2014). Intrinsic motivation and extrinsic incentives jointly predict performance: A 40-year meta-analysis. Psychological Bulletin, 140(4), 980-1008.
Deci, E. L. (1971). Effects of externally mediated rewards on intrinsic motivation. Journal of Personality and Social Psychology, 18(1), 105-115.
Deci, E. L., Koestner, R. and Ryan, R. M. (1999). A meta-analytic review of experiments examining the effects of extrinsic rewards on intrinsic motivation. Psychological Bulletin, 125(6), 627-668.
Henderlong, J. and Lepper, M. R. (2002). The effects of praise on children's intrinsic motivation: A review and synthesis. Psychological Bulletin, 128(5), 774-795.
Lepper, M. R., Greene, D. and Nisbett, R. E. (1973). Undermining children's intrinsic interest with extrinsic reward: A test of the overjustification hypothesis. Journal of Personality and Social Psychology, 28(1), 129-137.
Peters, K. P., Grauerholz-Fisher, E., Vollmer, T. R. and Van Arsdale, A. (2022). An evaluation of the overjustification hypothesis: A replication of Deci (1971). Behavior Analysis: Research and Practice, 22(3), 258-264.
Roane, H. S., Fisher, W. W. and McDonough, E. M. (2003). Progressing from programmatic to discovery research: A case example with the overjustification effect. Journal of Applied Behavior Analysis, 36(1), 35-46.
Ryan, R. M., Mims, V. and Koestner, R. (1983). Relation of reward contingency and interpersonal context to intrinsic motivation: A review and test using cognitive evaluation theory. Journal of Personality and Social Psychology, 45(4), 736-750.
Tang, S. H. and Hall, V. C. (1995). The overjustification effect: A meta-analysis. Applied Cognitive Psychology, 9(5), 365-404.
Wiersma, U. J. (1992). The effects of extrinsic rewards in intrinsic motivation: A meta-analysis. Journal of Occupational and Organizational Psychology, 65(2), 101-114.
