Does Paying for It Kill the Wanting
The famous finding that paying people for what they love kills the love is real, and far narrower than the popular version. The two meta-analytic camps that have argued it in print since 1994 agree on praise, on unexpected rewards and on expected tangible ones, and the work evidence runs the other way.
Answer to the question
Is the overjustification effect real at work?
Sometimes, and almost never at work. The overjustification effect holds for expected tangible rewards promised for doing or finishing a task someone already enjoys, measured as time spent on it once the money stops. Praise does the opposite, unexpected rewards do nothing, and the laboratory category built to imitate a salary shows no significant effect at all.
Key takeaways
- An expected tangible reward for doing or finishing an already-interesting task cut free-choice persistence by around a third of a standard deviation.
- Verbal praise moved motivation the other way, and rewards not tied to the task, the paper's own salary analogue, did nothing across seven studies.
- Both meta-analytic camps agree about praise, about unexpected rewards and about expected tangible ones; the dispute is four methodological points.
- Three work-domain syntheses find incentives raising how much gets done and least positive where the measure is quality.
Does paying people for what they love kill the love? That is the question everyone asks about the overjustification effect, and it is the wrong one. The answer is sometimes, and at work it depends on the arrangement. The better question is narrower: what the money was attached to, and who was doing the task?
Start with the sentence everyone quotes
The claim reaches most readers through Chapter 2 of Daniel Pink's Drive, titled "Seven Reasons Carrots and Sticks (Often) Don't Work". Pink reports that the authors of a 1999 meta-analysis reanalysed nearly three decades of studies, and quotes their conclusion straight:
> "careful consideration of reward effects reported in 128 experiments lead to the conclusion that tangible rewards tend to have a substantially negative effect on intrinsic motivation"
"Tend to" survives the trip from journal to paperback. Pink prints Mark Lepper's inconvenient null in the same chapter, gives a whole companion chapter to the circumstances in which rewards do work, and writes that not all rewards at all times are bad.
Chapter 2 presents the 1999 meta-analysis as settled. It is not settled. The paper has a moderator structure that switches the finding on and off depending on what the reward was attached to, and an opposing meta-analytic tradition has been arguing the contrary case in print since 1994. Both are below, with citations a reader can check.
The vault's read of the book was photographic and the edition's notes carry no page for the sentence, so a chapter is as precise a locus as this piece can give. One housekeeping point: the meta-analysis is Edward Deci first, Richard Koestner second, Richard Ryan third, and this piece credits it that way.
Read the moderator table before quoting it
The 1999 paper covers 128 studies and reports study counts but no pooled number of participants, so no total sample exists to quote. Its main outcome is free-choice persistence: how long someone keeps going at the activity later, when nobody is watching and nobody is paying. More than nine in ten studies measured that in seconds. The second outcome is what people say afterwards about the task.
Those two outcomes do not move by the same amount.
| What the reward was | Free-choice persistence | Stated interest |
|---|---|---|
| Verbal praise and positive feedback | Moderate increase, clear for college students and absent for children | Moderate increase |
| Unexpected tangible reward, nine studies | Nothing detectable | Nothing detectable |
| Reward not tied to the task | No significant change | No significant change |
| Expected tangible reward for engaging | Moderate reduction, about four tenths of a standard deviation | Small but significant reduction |
| Expected tangible reward for finishing | Moderate reduction, a bit over a third of a standard deviation | Small but significant reduction, after two outliers were removed |
| All tangible rewards together | Moderate reduction, about a third of a standard deviation | Small reduction, significant |
| Reward tied to level of performance | Moderate reduction on average, a shade under three tenths | Not significant as a block |
| All rewards pooled, praise included | Moderate reduction, about a quarter of a standard deviation | Nothing at all |
One scale runs through this piece: under about two tenths of a standard deviation is small, two to five tenths moderate, more than that large. The behavioural effect is the larger and much the better established: time on task falls once the money stops. The self-report effect is smaller, and it is real. Taking every tangible reward together, stated interest fell by a small amount whose confidence interval excludes zero, and the authors' own summary is that tangible rewards undermined intrinsic motivation on both measures. The abstract says the same of all expected rewards.
So the bottom row is the trap. It reads flat only because it pools verbal praise, which raises what people say, with pay, which lowers it, and the two cancel. Pooling is what made the self-report effect look like nothing, and anyone quoting a single number off this paper is quoting a cancellation. The behavioural and stated-interest measures track different things, a distinction [Wanting Is Not Liking](https://thewhyweproject.com/articles/wanting-is-not-liking) takes apart at length.
The finishing row needs its working shown. Across the fifteen studies with a self-report measure the raw figure sat near zero and the studies disagreed sharply; the authors removed two outliers to bring the set into line, and the remaining thirteen gave a small reduction that was significant. They conclude that rewards for finishing undermine intrinsic motivation on both measures. The figure depends on that removal, and a reader is entitled to know it does.
Then the row that decides an article about pay. Rewards not tied to task behaviour are the closest laboratory analogue of a wage, and the paper says as much, describing a study in which participants were paid two dollars simply for turning up:
> "The idea was to simulate the typical work condition in which people are paid a salary without having their pay tied in any direct way to the specific tasks they perform."
Seven studies, no significant effect. That sentence sits inside the paper the popular account is built on.
Where each camp is strongest
Deci's camp makes its strongest case on performance-linked pay. Because those results disagreed with one another, the analysis split them four ways by kind of control group and kind of feedback. In the slice where not everyone received the maximum reward, the undermining was large, close to nine tenths of a standard deviation. The authors call that group the one that "represents the type of performance-contingent rewards that one would typically find in the real world", and the "largest undermining effect of any category used in the entire meta-analysis".
Six studies.
That breakdown, and the next finding, come from the authors' own 2001 restatement in Review of Educational Research, because our read of the 1999 paper stopped partway through its tables. The restatement reports that the reduction did not fade: measured immediately, days later or a week later, it held, and was if anything largest a week out.
Cameron's camp makes its strongest case on scope. Judy Cameron and W. David Pierce's 1994 meta-analysis covered 96 experiments. Their own later summary: "rewards can be used effectively to enhance or maintain an individual's intrinsic interest in activities", and "the only negative effect of reward occurs under a highly specific set of conditions, circumstances that are easily avoided".
Then they did something new. In a footnote to their 1996 reply they converted their own effect size into minutes, using the spread of scores from an earlier eight-minute free-choice study, and calculated that the average person promised a tangible reward not tied to performance will spend about 41 seconds less time on the task once the reward stops.
Forty-one seconds is the overjustification effect in units a reader can hold.
Cameron and Pierce converted their own effect size into minutes. About 41 seconds less, in an eight-minute free period.
Their 2001 meta-analysis with Katherine Banko states the position at full strength: "Overall, the pattern of results indicates that reward contingencies do not have pervasive negative effects on intrinsic motivation."
Count where the disagreement actually sits
Begin with what both camps find. Verbal praise and positive feedback enhance intrinsic motivation, and the 1999 paper says so of its rivals' work explicitly. Unexpected tangible rewards do not undermine it. Expected tangible rewards for merely engaging with an interesting task reduce free-choice persistence, with Cameron and Pierce's figure the smaller. Both traditions found all three.
That is most of the overjustification effect, agreed across the aisle.
Praise enhances, unexpected rewards do nothing, expected tangible rewards reduce persistence. That is most of the overjustification effect, agreed across the aisle.
The genuine disagreements are methodological, and there are four of them.
- Performance-linked rewards. Deci's team splits the category by control group and finds large undermining in one slice. Cameron's team finds motivation holds up when rewards track performance, and argues that the slicing produces the effect.
- Whether dull tasks belong in the analysis. One camp excludes them by design; the other includes them and reports enhancement.
- Whether self-report counts. Deci's team finds a small reduction in stated interest for tangible rewards; Cameron's team finds none.
- Whether several effect sizes from one experiment may enter a meta-analysis separately. Cameron and Pierce argue the errors are correlated and that separate entry inflates the result; Deci's team argues that an aggregate obscures the moderators.
Two printed exchanges, in 1996 and again in 2001, worked the question over in the ordinary way of science.
The agreement is the finding. The popular question has no answer. The precise one does, and it is in the table above.
Check what the experiments actually paid for
The 1999 analysis excluded uninteresting tasks from its primary comparisons, and said why: the phenomenon "has always been specified as applying only to interesting tasks insofar as with boring tasks there is little or no intrinsic motivation to undermine". Within the thirteen studies that manipulated interest directly, the reduction on the interesting task was large and the dull task showed nothing. Cameron's camp, working from a different study set, finds that rewards on low-interest tasks raise free-choice motivation instead. [Self-Determination Theory at the Till](https://thewhyweproject.com/articles/self-determination-theory-at-the-till) sorts the same split for shoppers.
So the overjustification effect is defined over things people already enjoy, and measured as whether they go on doing them for free. Most of a job is not that.
The size of the original manipulations matters too. The first published experiment paid participants one dollar for each three-dimensional puzzle completed in the time allowed, a description taken from the 1999 meta-analysis because the 1971 paper's only open copy is an unreadable image scan.
A dollar a puzzle is a long way from a salary.
The field study everyone cites is Mark Lepper's 1973 work with children, setting an expected award against an unexpected award and against none. The expected-award children showed less subsequent interest. The unexpected-award children did not differ from controls, a null the meta-analytic record corroborated a quarter of a century later across nine studies. Lepper is the first author.
The better citation for a working reader is the same group's 1976 token economy, led by David Greene with Betty Sternberg. Two classrooms, 44 children, real tokens exchanged for prizes at assemblies, three reinforced groups against one control. During treatment all three spent more time on their target activities than controls. After the reinforcement stopped, two of the three spent significantly less.
Follow the money into actual workplaces
Here the overjustification effect stops behaving as the popular account implies.
Two syntheses of incentives and performance at work converge on one shape. A 1998 synthesis of 39 studies, led by Douglas Jenkins, found financial incentives moderately related to how much work got done and unrelated to how well. A 2014 synthesis led by Christopher Cerasoli, covering more than 212,000 people, found the same division of labour: intrinsic motivation predicts quality, incentives predict quantity, and the two are "not necessarily antagonistic".
Then the paper that adjudicates this article's question. Ji Hyun Kim, Barry Gerhart and Meiyu Fang took the primary studies from two earlier syntheses that disagreed by almost a full standard deviation, Jenkins's among them and the other a 2010 public-sector review reaching us at second hand, and re-analysed them together. So this is not a third dataset. Incentives improved performance in interesting tasks by a large margin, and in uninteresting tasks by not much less. The exception counts: the relationship is least positive where performance is measured as quality, most so in interesting tasks.
Two datasets and a re-analysis of one of them, pointing the same way. Money buys more of the thing. What it does not reliably buy is the thing done well.
The counterweight, cautions attached. Safelite Glass moved from hourly pay to piece rates in the mid-1990s, and output per worker rose by 44 per cent in Edward Lazear's published figure. His earlier working paper puts it at 20 to 36 per cent depending on what is held constant, and his own theory predicts the firm also attracted more able workers, so selection sits inside the headline number. Nobody claims installing windscreens is fascinating work.
Weigh the null against the bias correction
The best-designed field test of the overjustification effect is a null, and it came with a diagnosis of why this literature disagrees with itself. Constanca Esteves-Sorenson and Robert Broce reviewed more than 100 prior tests and reported that the typical one ran on about fifteen participants, and that not a single one jointly reported time on task, output and productivity, so the outcome chosen could decide what a study found. [What Forty-One Participants Buys](https://thewhyweproject.com/articles/what-forty-one-participants-buys) makes the same point about small samples.
Then they ran the canonical two-session design properly. The detail that follows comes from a 2020 interview with Esteves-Sorenson, not from the paper, which we read at abstract level only. The task was tasting and rating cookies; volunteers had turned up for it unpaid; the paid group was promised seventy-five cents a cookie, then told at the start of the second session that the money had stopped. In session one they rated 62 per cent more cookies, 29 per cent faster. In session two, with pay withdrawn, they produced no less than the unpaid controls and were no less productive.
What suffered was the quality of their written evaluations, which the authors read as a response to unmet pay expectations. Her own summary is the sentence to carry: the harmful effects of pay on interesting activities "are not as easily produced as suggested by prior research".
One task, cookies, in one field setting. A null, and not a refutation.
Then a correction that disciplines everybody. Petr Cala and colleagues, in a CEPR discussion paper that has not been peer reviewed and reaches us through the authors' own summary column, corrected 44 economics experiments for publication bias and concluded that "after correction for the bias, the mean experimental result suggests a negligible effect of financial incentives on performance". That cuts at economists sure money raises output and at anyone sure it lowers it.
The hole in the middle of all this is a field experiment that randomises real wages for work people call interesting and then measures whether they keep going for free. We searched for one and found none, which is the condition this literature shares with the neighbours [The Psychology You Learned Didn't Replicate](https://thewhyweproject.com/articles/the-psychology-you-learned-didnt-replicate) covers.
None of the famous field results concerns wages. A fine of ten shekels for late collection, introduced at six of ten Haifa day-care centres, roughly doubled lateness, and removing it did not bring the rate back down. A fine works the opposite way round from a reward, the behaviour is a chore nobody finds interesting, and the authors decline to choose between two explanations, saying a large enough fee would have worked.
In Gothenburg, 262 undergraduate volunteers were offered the health examination that qualifies a blood donor, with no payment, with about seven dollars, or with the option of passing it to a children's cancer charity. Overall, nothing significant. Among men, nothing. Among women the rate fell from about 52 per cent to 30 per cent when money entered, and recovered when the money could be given away. Those figures come from the working-paper version, whose authors give three reasons their students may represent nobody else.
Duty, reputation and gifts, then. Wages are a different question.
Two worlds sit behind the one point the camps have not settled. In the first, tying pay to how well the work is done slowly costs an employer the part of the worker that would have done it anyway. In the second, the money buys quantity while the wanting carries quality. Those are different products.
Nobody is pricing them differently.
Common questions
Is the overjustification effect real at work?
Sometimes, and it turns on what the money is attached to. The laboratory category built to imitate a salary showed no significant effect across seven studies. But the slice Deci's team calls the kind of performance-contingent reward one would typically find in the real world undermined by close to nine tenths of a standard deviation, the largest effect anywhere in that meta-analysis, on six studies. Both rows describe the overjustification effect at work.
Does praise undermine motivation the way money does?
No, and both camps agree here. Verbal praise and positive feedback produced a moderate increase in free-choice persistence and in stated interest in the 1999 meta-analysis, and the opposing tradition reports enhancement as well. One qualification is worth keeping: the increase was clear for college students and not significant for children. Pooling praise with pay is what hides the self-report effect.
Should managers stop paying for work people enjoy?
The one arrangement with a large undermining effect behind it is pay that tracks how well the work is done, and it rests on six laboratory studies, which is where the worry belongs. A wage not tied to the task is the case with no undermining behind it, and no field experiment randomising real wages for interesting work was located on our search. The work evidence shows incentives raising quantity and being least positive for quality in interesting tasks.
Sources
Deci, E. L., Koestner, R., & 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.
Deci, E. L., Koestner, R., & Ryan, R. M. (2001). 'Extrinsic Rewards and Intrinsic Motivation in Education: Reconsidered Once Again.' Review of Educational Research, 71(1), 1-27.
Cameron, J., & Pierce, W. D. (1994). 'Reinforcement, Reward, and Intrinsic Motivation: A Meta-Analysis.' Review of Educational Research, 64(3), 363-423.
Cameron, J., & Pierce, W. D. (1996). 'The Debate About Rewards and Intrinsic Motivation: Protests and Accusations Do Not Alter the Results.' Review of Educational Research, 66(1), 39-51.
Cameron, J., Banko, K. M., & Pierce, W. D. (2001). 'Pervasive negative effects of rewards on intrinsic motivation: The myth continues.' The Behavior Analyst, 24(1), 1-44.
Eisenberger, R., & Cameron, J. (1996). 'Detrimental effects of reward: Reality or myth?' American Psychologist, 51(11), 1153-1166.
Lepper, M. R., Greene, D., & 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.
Deci, E. L. (1971). 'Effects of externally mediated rewards on intrinsic motivation.' Journal of Personality and Social Psychology, 18(1), 105-115.
Greene, D., Sternberg, B., & Lepper, M. R. (1976). 'Overjustification in a Token Economy.' Journal of Personality and Social Psychology, 34(6), 1219-1234.
Jenkins, G. D., Mitra, A., Gupta, N., & Shaw, J. D. (1998). 'Are financial incentives related to performance? A meta-analytic review of empirical research.' Journal of Applied Psychology, 83(5), 777-787.
Cerasoli, C. P., Nicklin, J. M., & Ford, M. T. (2014). 'Intrinsic motivation and extrinsic incentives jointly predict performance: A 40-year meta-analysis.' Psychological Bulletin, 140(4), 980-1008.
Kim, J. H., Gerhart, B., & Fang, M. (2022). 'Do financial incentives help or harm performance in interesting tasks?' Journal of Applied Psychology, 107(1), 153-167.
Weibel, A., Rost, K., & Osterloh, M. (2010). 'Pay for Performance in the Public Sector: Benefits and (Hidden) Costs.' Journal of Public Administration Research and Theory, 20(2), 387-412.
Esteves-Sorenson, C., & Broce, R. (2022). 'Do Monetary Incentives Undermine Performance on Intrinsically Enjoyable Tasks? A Field Test.' The Review of Economics and Statistics, 104(1), 67-84.
Cala, P., Havranek, T., Irsova, Z., Matousek, J., & Novak, J. (2022). 'Financial Incentives and Performance: A Meta-Analysis of Economics Evidence.' CEPR Press Discussion Paper No.
Gneezy, U., & Rustichini, A. (2000). 'A Fine is a Price.' The Journal of Legal Studies, 29(1), 1-17.
Mellstrom, C., & Johannesson, M. (2008). 'Crowding Out in Blood Donation: Was Titmuss Right?' Journal of the European Economic Association, 6(4), 845-863.
Lazear, E. P. (2000). 'Performance Pay and Productivity.' American Economic Review, 90(5), 1346-1361.
Pink, D. H. (2009/2018). Drive: The Surprising Truth About What Motivates Us. Canongate paperback 2018, ISBN 978-1-78689-170-9.