Intermittent Rewards and Motivation: What the Dopamine Research Supports

Unpredictable rewards grab attention because dopamine neurons respond to surprise and uncertainty, and behavior that is rewarded only some of the time is slow to fade. Rewards you always expect, on the other hand, can reduce your interest in a task. That is the solid science. Advice to randomly withhold rewards from yourself is a reasonable idea but has not been tested as a productivity method. This page is not medical advice.

Not medical advice. This page is about everyday motivation, not treatment for any condition.

What intermittent reinforcement means

Intermittent (or partial) reinforcement means a behavior is rewarded only some of the time. Learning researchers describe four basic schedules:

The best-known effect is the partial reinforcement extinction effect: behavior that was rewarded only some of the time keeps going longer after the rewards stop. In a 2019 series of four rat experiments, rats rewarded on 33% of trials needed approximately 3 times as many trials before they began to stop responding as rats rewarded every time, and rats rewarded on 20% of trials needed 5 times as many. The authors’ explanation is that animals learn roughly how many tries usually pass per reward, so a run of misses does not tell them the rewards have ended. That is useful for building habits and a problem when the habit is one you want to break.

Surprise and uncertainty drive dopamine

Wolfram Schultz’s recordings in monkeys showed dopamine neurons fire for better-than-expected rewards, stay flat for fully expected ones, and dip when an expected reward does not arrive. So a reward that always comes produces little dopamine response once it is predicted.

A 2003 study found a second signal: activity that gradually rose before a possible reward and was greatest when the odds were 50/50. The authors suggested this uncertainty signal may matter for attention-based learning and risk-taking.

Do uncertain rewards motivate people?

The monkey recordings do not show what people will do, but behavioral experiments point the same way, with a catch. In four experiments with real rewards, Shen, Fishbach and Hsee found people invested more effort, time and money to qualify for an uncertain reward (even odds of getting $2 or $1) than for a certain $2, even though the certain reward was worth more. The effect appeared only when people were focused on the process of earning the reward, not on the reward itself. In that case, the authors wrote, “uncertainty generates positive experience such as excitement and hence increases motivation.”

So uncertainty seems to help most while you are already working toward something. It is less clear that it helps you decide to start.

When rewards backfire

This is the “undermining effect.” Murayama and colleagues note that it challenges theories that assume money always increases motivation. Rewards are not bad, but tying an expected prize to every task can shift your focus from the task to the prize.

Immediate rewards keep people going

The rewards that matter most for sticking with a goal may be the ones you feel while doing it. Across five studies, Woolley and Fishbach found that immediate rewards such as enjoyment predicted persistence at New Year’s resolutions, and at single sessions of studying and exercising, while delayed rewards such as better health did not. This held over both 1-week and 3-month time frames. Their pooled analysis found “immediate rewards are more strongly associated with actual persistence in a long-term goal.” The studies measured how well each kind of reward predicted persistence, so they show a link, not proof that adding enjoyment causes it.

This fits the undermining research: enjoyment that comes from the activity itself is a different thing from a prize handed out for finishing it.

The intermittent-reward advice

On Huberman Lab Episode 39, Andrew Huberman suggests not rewarding yourself every time you finish something, and sometimes deciding at random whether to celebrate. He also suggests learning to value effort itself. These ideas follow from the uncertainty research, but they have not been tested in trials of work or study habits.

The downside: variable rewards in apps and gambling

The same principles work on you in apps. An analysis of over one million posts from over 4,000 people found posting patterns matched reward-learning models, with likes acting as rewards. See social media and dopamine.

A 2023 review by Clark and Zack argues that the variable, unpredictable payouts in gambling may drive learning in a way that resembles drug cues, and that gaming, shopping and social media use similar variability. It points to infinite scrolls and personalized recommendations as newer sources of unpredictable reward, and to how often rewards arrive as a second factor. This is a theoretical review, not a trial, but it explains why the same tactic that can make a hard task more engaging can also make a feed hard to put down.

How to apply it

Based on the reward research and the uncertainty experiments:

• Don’t bribe yourself for everything: Expected rewards for every task can reduce interest.
• Use feedback: Noting progress is a form of positive feedback, which increased interest in the meta-analysis.
• Make small rewards occasional: If you use treats, try not attaching them to every finished task.
• Make the work itself more pleasant: Good music, a better setting or a training partner are immediate rewards, which predicted persistence better than distant goals.
• Focus on the process: Notice what you learned, not just the outcome. Uncertainty motivated people only when they were focused on the process.
• Keep variable rewards out of work time: Notifications and feeds are built on unpredictable rewards; silence them during focus blocks. The workspace guide has more.

For background, see the dopamine and motivation guide and what a dopamine reset can and can’t do. The focus and cognition guide has more.

Limits of the evidence

The dopamine studies are mainly in monkeys, the extinction experiments are in rats, and the human motivation studies are short lab tasks or self-reports with small sums of money. Turning them into habits for real work involves extrapolation. No trial has tested random self-rewards as a way to work or study better.