How to reward yourself so the habit outlives the reward

Search for how to reward yourself and one of the first things you will find is a number that is not in the paper it cites. A reward system guide I read on 27 August 2026 says an immediate bonus "increased the rate at which people kept up the activity after the reward was removed by 35%," crediting Kaitlin Woolley and Ayelet Fishbach's 2018 study in the Journal of Personality and Social Psychology. I opened the study. There is no 35 percent in it.

What is in it is more useful. In the third experiment, 223 people on Mechanical Turk did a spot-the-difference task for a flat 30 cents. A third were told a 60 cent bonus would land the moment they finished, a third that the same 60 cents would arrive a month later, and a third expected nothing at all. Everyone was then offered the chance to keep working for no extra money. Of those paid immediately, 84.2 percent carried on, against 70.3 percent of the people promised the identical bonus in a month and 52.1 percent of the people paid nothing. Cornell, writing up its own researcher's work, called that an almost 20 percent increase.

The bonus was sixty cents. Same money, same task, and the only difference was whether it landed the second they stopped clicking or four weeks after they had forgotten the study existed.

So the answer this literature gives is duller than a treat menu, and it is worth having up front. Reward yourself inside the hour you did the work, at a size you could repeat next Tuesday without thinking about it, and let the visible record of having shown up count as part of the payment. In the field experiments that deliberately took the reward away afterwards, that is the combination that left people above where they started. None of it needs money.

What happened when the researchers stopped paying

Gary Charness and Uri Gneezy ran the cleanest test of that, published in Econometrica in 2009. The design was blunt. At one university, 120 students were split three ways: one group got nothing, one got $25 to visit the campus gym once, and one got the $25 plus a further $100 for attending eight more times inside a month.

Then the money stopped. The researchers kept reading the turnstile records for another seven weeks after the last dollar was handed over, which is the only way to find out whether a reward leaves anything behind.

The control group drifted down. It went from 0.59 visits a week to 0.56, and the group paid once for a single visit barely moved either, from 0.70 to 0.76. The eight-visit group went from 0.60 visits a week before any money changed hands to 1.24 after all of it had been spent, an increase of 107 percent over their own baseline. A second study at UC San Diego tracked 168 students for 13 weeks after payment ended and found the same shape: 0.52 weekly visits before, 1.46 after.

Two details matter more than that headline. The effect "largely vanishes" for students who were already regular gym-goers, so the money did not buy extra exercise from people who exercised, it recruited the people who weren't going at all. And nothing decayed over the follow-up, with attendance moving by an insignificant 0.004 visits per period in one study and minus 0.010 in the other.

The case against rewards is real, and narrower than it sounds

You have probably met the opposite claim. Edward Deci, Richard Koestner and Richard Ryan pooled 128 experiments for Psychological Bulletin in 1999 and found that expected tangible rewards did real damage: rewards for engaging in a task, for finishing it, or for hitting a standard undermined free-choice persistence at d = -0.40, -0.36 and -0.28 respectively.

Notice the setup: someone is doing a puzzle they already find interesting, they get paid for it, and then they are watched to see whether they still do it in their own free time. That is not the position of a person who has been meaning to start running since March.

Praise ran the other way. Positive feedback raised free-choice behaviour at d = 0.33 and self-reported interest at d = 0.31, which is a decent return on something that costs nothing and can be repeated daily without losing its value.

There is an honest limit here, and the authors are the ones who state it. When they tested whether the undermining was merely temporary, nearly all the delayed measurements came from studies of children, at d = -0.49 within a week and -0.53 after one, and they wrote that there is "no basis for generalizing these results to adults." Cameron and Pierce, whose 1994 meta-analysis reached a friendlier verdict on rewards overall, still reported task-contingent rewards undermining free-time behaviour at d = -0.32 before outliers came out. The argument is about scope.

Timing beats size, and size is where people overspend

Woolley and Fishbach put the two against each other in their fifth experiment. Timing won on its own. Their 2017 paper in Personality and Social Psychology Bulletin then tested the same idea against real goals, and found that immediate rewards predicted whether people were still keeping their New Year's resolutions while delayed rewards predicted nothing, even though the delayed payoff was the reason people gave for having the goal.

Most self-reward plans fail here. A weekend away in three months belongs to a version of you who has already made it, and the you sitting here at 9 p.m. with the thing finally finished gets nothing at all. Good coffee at the end of the run is doing more work than it looks.

Put the reward inside the task rather than after it

The strongest form of immediacy is a reward you can only collect while doing the thing. Katherine Milkman, Julia Minson and Kevin Volpp handed 226 students and staff at a university gym iPods loaded with page-turner audio novels, in a study published in Management Science in 2014. One group could only listen at the gym, because that is where the iPods stayed. One group got the books on their own devices with a suggestion to save them for workouts. One group got neither.

Visits in the locked-iPod group ran 51 percent above control at first, and 29 percent above in the encouraged group. That advantage then eroded by about 0.07 visits a week across the nine weeks, and it collapsed over Thanksgiving when the gym closed and the routine broke. Something else survived, though. When the study ended, 61 percent of participants chose to pay for continued gym-only access to those iPods, which says they wanted the restriction kept rather than merely tolerated.

Bundling works, and it wears off. Use it to buy yourself the first six weeks, and expect to want something new by the time the novelty of the audiobook has gone.

Reward turning up, not turning up on schedule

There is a tempting refinement here: pay yourself only when you do the thing properly, at the hour you planned. It quietly backfires. John Beshears and colleagues paid workplace gym users either for exercising on any weekday or for exercising inside a two-hour window they had chosen themselves, and the rigid version produced fewer visits both during the intervention and after the payments stopped. Four weeks after the money ended, the flexible arms were still making 0.25 more visits a week than the control group, and were 12 percentage points more likely to get there at least once in a given week.

Rigid rewards break easily. A reward pinned to a specific hour turns one late train into a failure, which is why a morning routine needs a two-minute version of every line instead of a timetable, and why a streak is worth designing around its restart.

Do not spend the whole thing in week one

Mariana Carrera, Heather Royer, Mark Stehr and Justin Sydnor ran an experiment on employees of a Fortune 500 company that answers a question nobody asks themselves: how should the payments be spread out? Everyone earned $10 a visit. The cap was two visits a week, and the arms differed only in the shape of the schedule: one paid steadily across eight weeks, one front-loaded the money on the theory that a big early reward breaks through procrastination, and one scattered the same eight paid weeks across sixteen.

Front-loading did not increase participation, and sustained less exercise over time. Scattering did slightly better. The authors also report no sign that the payments crowded out intrinsic motivation once they stopped, including among the employees who were already exercising heavily before any of it began.

For one person with no budget, the translation is short. Unpredictable is fine. A brilliant fortnight of treats followed by nothing at all is the design that fails, and it is the one most people build the evening they decide to change something.

Recognition is the reward you can afford every day

Deci's group found praise enhancing where goods were corrosive, but that was lab work with children and undergraduates. The largest recent test of the same idea was the STEP UP trial, published in JAMA Internal Medicine in 2019, in which 602 adults with a body mass index of 25 or higher set themselves a step goal and were randomly assigned to a control group or to one of three games made of points and levels: one with a supportive partner, one with a collaborative team, one a competition against other participants.

All three raised daily steps across 24 weeks, and competition did most, at 920 steps a day above control. Then the games stopped. Twelve weeks later competition was the only arm still significantly ahead of control, at 569 steps a day, so the version with other people watching was also the version that left something behind.

A secondary analysis in PLOS ONE found one group for whom all three games held up through the follow-up, between 941 and 1,061 steps a day. It was the participants who started out less active and less social. If you have already decided that points and leaderboards are not for people like you, that is probably the finding to sit with.

Points are not the mechanism. Somebody seeing them is, which is the same lever that makes reporting to a named person beat tracking privately, and it is the cheapest thing on this list to hand yourself again tomorrow.

How to reward yourself, in four rules the evidence supports

None of this makes the hard thing easy. It changes what happens in the ten seconds after you finish, which is the only part of the whole business you can reliably control.

• Pay yourself inside the hour rather than at the weekend, because 60 cents now beat 60 cents in a month.
• Pick something you can deliver on an ordinary Tuesday, since a reward has to outlast the novelty of having one.
• Reward the fact that you turned up, not that you turned up at 7 a.m.
• Count being noticed as payment: a logged day, a streak, a friend who saw it.