Skip to content

← Psychology

Self-control and habits

Chapter 15 · Feeling and wanting

In this chapter 5 sections
  1. Waiting for the second treat
  2. The test, told again
  3. Is willpower a tank that runs dry?
  4. How a habit forms
  5. What a habit doesn’t care about

In the late 1960s, at the nursery school on Stanford University’s campus in California, a small child would be offered a choice between two treats, one liked better than the other. The lesser one could be had at any time. The better one meant waiting until the researcher, who was leaving the room, came back. Then the researcher left, and the child was alone with the choice.

It sounds like a party game. Known ever since as the marshmallow test, it became one of the best-known experiments in psychology, and its story, told and retold, was that the children who could wait grew up to do better at almost everything. That story turned out to be too simple, and the way it fell apart says a good deal about how self-control really works.

Here is the short version. Self-control is choosing what you want more in the long run over what you want right now. There are two ways to manage it. You can fight the urge when it comes, by effort. Or you can arrange things so the urge is weak, or never comes at all. The second way, done often enough in the same place at the same time, turns into a habit: an action that the place and the moment set off by themselves, with no decision needed. The people who seem to have the most willpower turn out, in several studies, to be the ones who need it least.

Waiting for the second treat

The psychologist Walter Mischel and his team ran these tests on hundreds of children, and much of their early work asked what made waiting possible. They changed the room, or what the children were told, and watched what happened to the wait.

The first surprise was that looking at the prize made waiting harder. When the treats were out of sight, children waited much longer than when they sat facing them. Distraction helped too, and some distractions worked better than others. Children told to think about fun things waited a long time. Children told to think sad thoughts gave up quickly, and so did children told to think about the treats themselves. Thinking about the very thing you were waiting for, the thing that was supposed to keep you going, made you give in sooner.

Mischel and a colleague later described this as two systems pulling against each other. One is hot: quick, emotional, set off by whatever is in front of you, saying go. The other is cool: slow, thoughtful, able to plan, saying wait. The trick of waiting, on this view, is to keep the hot system from being lit, by not looking at the sweet, by thinking about something else.

Then came the part that made the study famous. Years later, Mischel’s team got back in touch with some of the children, now teenagers. In 1988 they reported that parents of 95 of them, asked about their children a decade after the test, rated the ones who had waited longer at four or five as more competent at school and with friends, more able to plan, and better at dealing with frustration and stress. In 1990 they reported on the American college entrance exam. Among the 35 children who had waited with the treat in plain view and no hints on how to pass the time, the link between seconds waited and exam scores was strong, a correlation of about .4 to .6. A few minutes alone with a sweet seemed to look a long way into a child’s future.

This is the version that went into textbooks and talks. It is worth noticing what it rested on. The children all came from one nursery school, in a university community. And the strongest result came from 35 of them.

The test, told again

In 2018 three researchers tried the idea on a much larger and more varied group: 918 American children from ten places across the country, followed from birth as part of a national study of child care. At four and a half, each did a version of the marshmallow test, with a seven-minute limit. At fifteen, each took achievement tests. The researchers paid most attention to the 552 children whose mothers hadn’t finished college, a group quite unlike the families of a university nursery.

The link was there, but it was about half the size of the one in the 1990 study. Then the researchers took into account the child’s family background, their early ability at thinking and language, and what their home was like. Two thirds of what was left of the link disappeared. Most of the difference between children came from whether they could wait at least 20 seconds, not from the long waits. And the waiting was barely linked at all to how the teenagers behaved.

This doesn’t mean poorer children have less self-control, or that the test measured nothing. It means that how long a four-year-old waits is tied up with everything else about that child’s life so far, and most of what it predicted later was those other things. The test was picking up a whole life, not a fixed gift.

A follow-up of the original Stanford children added to the doubt. In their late forties, 113 of them answered questions about their money, savings, debts and schooling. How long they had waited at four predicted none of it: on average, the correlation was 0.02, about zero. A combined measure of their self-control, built from questionnaires at 17, 27 and 37 together with the preschool test, did predict these things, modestly.

Meanwhile other researchers asked a different question: what is a child deciding, while waiting? In 2013 Celeste Kidd and her colleagues ran a small experiment with 28 children aged three to five. Before the marshmallow, each child did an art project with the same researcher who would run the test. For half of them the researcher promised better crayons and then some stickers, and came back with them. For the other half the researcher promised the same and, both times, came back without them.

The adult had kept the promises9 of 14 waited all 15 minutesaverage wait: about 12 minutesThe adult had broken the promises1 of 14 waited all 15 minutesaverage wait: about 3 minutes
Figure 15.1 Each circle is one child. Filled circles waited the full 15 minutes for the second marshmallow. Kidd, Palmeri and Aslin, 2013.

The difference was large. After the kept promises, 9 of the 14 children waited the whole 15 minutes, against 1 of 14 after the broken ones. A child who has learned that adults don’t keep their word has a perfectly sensible reason to eat the marshmallow now. It was a small study, but a later one found the same thing from a different angle: children waited less for a reward promised by an adult they had just seen treat someone else badly. Children also waited longer when told that their own group waited, and another team found that Japanese children waited longer for food than for a present, while American children did the opposite. The researchers’ explanation is that waiting to eat is stressed more in Japan, and waiting to open presents more in the United States. So what looked like one fixed quality in a child turns out to depend on trust, on what other people do, and on habits learned at home. Even the times are changing: American children tested in the 2000s waited about two minutes longer, on average, than children in the 1960s.

None of this means self-control doesn’t matter. A study in New Zealand followed about 1,000 children in the city of Dunedin from birth to the age of 32. Their self-control was rated again and again through childhood, by parents, teachers, observers and the children themselves. At 32, those who had had less self-control as children were more likely to have health problems, depend on drugs or alcohol, struggle with money and have a criminal record. This held after their intelligence and social class were taken into account, and in a second study in Britain the twin with less self-control at five was doing worse than the other twin by twelve. The children whose self-control grew between childhood and adulthood did better at 32 than their childhood scores predicted. These are links, not proof that self-control was the cause, and the measure was years of ratings, not a few minutes with a sweet. What fell apart was the idea that one short test could read a child’s future. The idea that self-control matters held up.

Is willpower a tank that runs dry?

In 1998 the psychologist Roy Baumeister and his colleagues published an experiment that seemed to show what self-control is made of. Some people had to make themselves eat radishes instead of the tempting chocolates in front of them. Others didn’t have to hold back. Then everyone was given puzzles that couldn’t be solved. The radish-eaters gave up sooner.

The explanation was that self-control draws on a limited store, like fuel in a tank. Resisting the chocolate used some up, so there was less left for the puzzles. Baumeister called the state that follows ego depletion. The idea was easy to believe, because it felt true: everyone knows the evening when, after a long day of being good, they give in to everything. By 2010 a review of 83 studies found a clear effect.

But later reviews questioned how big the effect was, and whether it existed, and found signs of possible bias in what had been published. A field that publishes the experiments that work and quietly drops the ones that don’t can build a solid-looking result out of very little. So in 2016, 23 laboratories ran the same carefully planned experiment, with 2,141 people between them. The effect came out close to zero. In 2021 a second project, 36 laboratories and 3,531 people, tried again with a new design, each laboratory using one of two procedures, everything planned in advance. The planned analysis found no clear effect. When every person was counted, including those the plan had said to leave out, there was a tiny one, and a further look hinted that it was a little larger among people who said they felt more tired.

So being worn out may still make it harder to keep going at something, though that is only a hint. What hasn’t held up is the tank: the claim that each act of self-control drains a single store, so that eating a radish leaves you weaker at puzzles.

If willpower isn’t a fuel, what are people with good self-control doing? A growing body of evidence says they don’t spend more of their day fighting temptation. If anything they do less of it. In six studies with 2,274 people in all, the psychologists Brian Galla and Angela Duckworth found that people with better self-control tended to have more good habits, for healthy snacks, exercise, regular sleep, study and homework. And it was the habits that went with how well they did. They didn’t seem to be winning more battles. They seemed to be having fewer of them.

Duckworth and her colleagues gave the other half of the answer. You can deal with a temptation at any point as it grows, and the earlier the better, because an urge gets stronger the longer it is in front of you. The steps are much like those for handling a strong emotion, and one of the authors, James Gross, drew up both: choose the situation, change it, move your attention, change how you see it, and only last hold back the action. The partygoer who sits far from where the drinks are being poured, the student who goes to the library and leaves the phone at home, never have to resist anything in the moment. The children who waited longer with the treats out of sight, or thinking about something fun, were helped in the same way. It is so effective that people tend not to notice it as self-control at all, because it doesn’t feel like a struggle.

So “try harder” reaches for the last tool, not the best one. Effort in the moment is real, but stepping in earlier may work better.

How a habit forms

The American psychologist William James, writing in 1890, opened his chapter on habit by saying that living creatures, seen from outside, are “bundles of habits”. He called habit “the enormous fly-wheel of society”, and his examples show what he meant: it “keeps the fisherman and the deck-hand at sea through the winter”, holding people to the work they grew up in.

A habit, as psychologists now describe it, forms when you repeat the same action in the same setting while going after some goal. It is learning by consequences, done over and over in one place. Gradually the setting itself starts the action, before any thought. The goal that started it all can drop away, and the habit runs on.

How much of life runs this way? When students in Texas wrote down, every hour of the day, what they were doing and thinking, between a third and a half of what they did were things done almost daily in the same place. During those actions they tended to be thinking about something else, and they felt less stressed than during the rest. It was one group of American students over a day or two, but it gives a sense of how much runs by itself.

The best-known study of how long it takes was done at University College London. Phillippa Lally and her colleagues asked 96 volunteers to pick something to eat, drink or do every day at a set moment, such as a piece of fruit with lunch or a run before dinner, and keep it up for twelve weeks. Every day each person said how automatic the action had become. For most people it grew quickly at first and then leveled off, like a curve rising toward a ceiling. The time to get nearly to that ceiling varied enormously, from 18 days for one person to about 254 for another, the longer times worked out from the curve since the study stopped at 84 days. The middle value, the median, was about 66 days, and exercise seemed to take longer than eating or drinking, though in so few people that may have been chance. That figure comes from a fairly small group: of the 96, the curve fitted the data well for only 39.

So the claim you may have heard, that a habit takes 21 days, has no study behind it. It seems to come from a 1960 book by Maxwell Maltz, a plastic surgeon, who wrote that it usually took a minimum of about 21 days to change a mental image, as when his patients got used to a new face. Getting used to something is not the same as forming a habit.

Lally’s study also bears on one of James’s own rules. James told his readers: “Never suffer an exception to occur till the new habit is securely rooted in your life.” In plainer words, never allow yourself a single day off. A single slip, he wrote, was like dropping a ball of string you were winding, which undoes more than many turns can wind again. In Lally’s study, missing a single day made no real difference. Doing it more regularly went with a clearer pattern of growing habit. On this evidence, one slip is not the disaster James feared.

What a habit doesn’t care about

Once a habit runs on its own, it can keep going after the reason for it has gone. In one experiment, people at a cinema were handed popcorn, some of it fresh and some of it stale. People who seldom ate popcorn at the cinema ate by how much they liked it. People who always ate popcorn at the cinema ate as much stale popcorn as fresh. Their hunger and how much they liked it hardly mattered. But the habit had limits. Away from the cinema, the regulars went back to eating by taste. And when they were asked to eat with the hand they didn’t normally use, the habit lost its hold too. It was one pair of experiments, but they show how a habit is tied to the setting that cues it, and even to the exact movements.

That suggests a way to break one. Change the setting. A study of 115 students moving to a new university in Texas looked at their habits of exercise, reading the newspaper and watching television a month before and a month after the move. The habits survived only where something of the old setting stayed the same, such as reading the paper with other people. Where the setting changed, the habits broke, and in some cases people then did what they actually intended, not what they had always done. That suggests a move, a new school year or a new room may be a good moment to start a new habit, while the old ones are loose.

In Kavre, in Nepal’s middle hills, researchers tried changing settings on purpose. In 2013, in eight wards of two neighboring village areas, they set out to change five habits of food hygiene among mothers of small children, such as washing hands with soap before feeding a child and heating leftover food thoroughly. In four of the wards, over three months, trained local volunteers ran rallies, games, stories, plays and competitions that played on mothers’ care for their children, disgust at germs and pride. They also gave kitchens a makeover, with paint, bright bunting and reminders hung at eye level, to make the kitchen a new setting. Six weeks after it ended, observers saw all five behaviors done together by 43% of mothers in those wards, against 1% before and 2% in the other wards. Washing hands with soap before feeding went from 5% to 67%. The authors say plainly what the study can’t tell: so much was done at once that nobody knows which part worked, mothers who knew they were being watched may have been on their best behavior, there were only eight wards, and whether the new habits lasted nobody yet knows.

A smaller tool works inside a single head. If you tie a plan to a moment, “when I get home from school, I’ll do ten math questions before tea”, the moment can start the action much as it starts a habit. Psychologists call these if–then plans, and a review of 642 tests found that they helped people do what they meant to, more so when the plan really had an if and a then, when people cared about the goal, and when they had rehearsed it. There is a tip on this site built on it, along with others on changing what sets a habit off and on how long a new one takes.

Habits that involve drugs, alcohol or gambling are a different matter. They are not something to handle alone with a plan, and the first step is to talk to someone you trust or to a health worker.

So the old picture, of a person straining against temptation with a store of willpower that slowly runs down, explains less than it seemed to. Self-control matters, and it can grow. But much of it is done before the moment of temptation ever comes: by choosing where you are and what is in front of you, by trusting that waiting will pay, and by repeating the good thing in the same place until it does itself. In Mischel’s room, waiting came easiest not to the children who stared at the treat and held on, but to those who couldn’t see it, or were thinking of something else.

Found a mistake? Please tell us.