The Reward That Stops the Drawing

In 1973 Lepper, Greene and Nisbett watched nursery children during free play and identified the ones who spontaneously chose to draw. Those children clearly liked drawing; nobody was making them.

They were then split into three groups. One was told in advance they would get a "Good Player" certificate for drawing. One got the certificate unexpectedly afterwards. One got nothing.

Two weeks later, during free play, the children who had been promised the reward drew about half as much as the other two groups.

This is the overjustification effect, and the mechanism is self-perception. A child who draws for a prize, and observes themselves doing so, infers that drawing is the kind of thing one does for prizes. Remove the prize and the reason is gone.

Note the conditions, because they are what makes this a real finding rather than a slogan. The effect appears for expected, task-contingent rewards given for activities the person already found intrinsically motivating. Unexpected rewards after the fact did not produce it. Rewards for things nobody wanted to do anyway do not produce it, because there is no intrinsic motivation to undermine.

"Rewards destroy motivation" is a bad summary that has caused as much trouble as the finding has prevented.

The distinction that makes the literature coherent is whether a reward is experienced as informational or as controlling. The same money can be either. A prize awarded for demonstrated skill tells you something about your competence. The identical sum delivered as a condition of compliance tells you who is directing whom. Deci and Ryan's meta-analyses find the undermining effect concentrated in the controlling case, which is why the literature looks contradictory until you sort the studies by framing.

Their broader framework, self-determination theory, proposes three needs whose satisfaction supports durable motivation: autonomy (experiencing your actions as chosen; competence) a sense of growing capability; and relatedness (connection to others).

Its practical value is diagnostic. When motivation is failing, the useful question is which of the three is thwarted, because the fixes are different. A person with no discretion over how they work has an autonomy problem, and a bonus does not touch it. A person who cannot tell whether they are improving has a competence problem, and what they need is feedback. A person working alone on something nobody sees has a relatedness problem.

Financial incentives also fail in a more specific way than most people expect. Dan Ariely and colleagues ran experiments in India where bonuses could reach several months' spending, and found that very large incentives degraded performance on tasks requiring creativity or problem-solving while improving it on purely mechanical ones.

The pattern is consistent with arousal effects: high stakes narrow attention, which helps when the bottleneck is effort and hurts when the bottleneck is insight. So incentives work well for stuffing envelopes and badly for finding a bug, and the common practice of applying them uniformly across a workforce doing both is a mistake with evidence behind it.

Goals have a similar shape: powerful, and distorting in predictable ways.

Locke and Latham's goal-setting research is among the most replicated findings in organisational psychology: specific, difficult goals reliably outperform "do your best". A separate and equally robust literature documents what else they do: attention narrows onto the measured dimension, unmeasured dimensions are neglected, risk-taking increases near a threshold, and falsification appears when the target is both important and hard.

Both bodies of evidence are correct. A goal is a powerful instrument with known side effects, which is an argument for choosing what you measure very carefully rather than for not having goals.

The most striking demonstration that money can make things worse involves no goals at all.

Gneezy and Rustichini studied Israeli high school students collecting donations door to door. One group was paid nothing and told the cause mattered. One received 1% of what they collected; one received 10%.

The unpaid group collected the most. The 1% group collected substantially less than the unpaid group. The 10% group did better than the 1% group and still worse than the unpaid.

Once money enters, the question the person is answering changes. It stops being "am I the kind of person who helps with this" and becomes "is this worth my time at this rate" And 1% answers the second question badly while having destroyed the first. Titmuss found something similar for blood donation decades earlier.

Pay properly or do not pay at all. Paying a little is the one option that is reliably worse than both, and it is the option organisations reach for most often, because it looks like a cheap way to buy some extra motivation.