Walter Mischel, E. B. Ebbesen and colleagues · 1970–1972

The marshmallow test

What helps a young child wait for a bigger treat, and does the length of that wait say anything about the person they will become?

  • Bing Nursery School, Stanford University, California
  • Replicated, with caveats

The question

In the late 1960s, Walter Mischel and his colleagues at Stanford were studying how children manage to wait for something they want. The setup was simple: a small reward now, or a bigger or better one later. Most young children want the bigger reward. The real question was what goes on in the child’s head during the wait. Does looking at the reward strengthen resolve, or wear it down? And can a child be helped to hold out?

Only later did a second question take over: does a four-year-old’s waiting time say anything about how that child will do as a teenager?

How it was done

The studies took place at the Bing Nursery School on the Stanford campus, a preschool attended mainly by children of Stanford faculty and students. Between 1968 and 1974, 653 children took part in one version or another.

In the first published study (Mischel and Ebbesen, 1970), 32 children aged between three and a half and almost six were taken, one at a time, to a room with a table. Each child chose which of two treats they preferred: two animal cookies or five small pretzels. The experimenter then explained the deal. He had to leave. If the child waited until he came back on his own, they would get the treat they preferred. If they wanted to stop waiting, they could signal for him to come back at any time (by eating a small piece of pretzel; a later follow-up used a bell), but then they would get the other treat. The experimenter returned after 15 minutes at most.

The researchers changed only one thing: what the child could see during the wait. Both treats were on the table, only one of them was, or neither was.

The 1972 study (Mischel, Ebbesen and Zeiss) ran three further experiments with 92 children aged three to five. This time the treats were a marshmallow and a stick pretzel, and the child rang a desk bell to call the experimenter back. Some children got a toy (a Slinky spring) to play with. Others were told to think about something fun, something sad, or the treats themselves while they waited.

About the name. The “marshmallow test” label came later. The treat varied from study to study: cookies and pretzels in 1970, a marshmallow and a pretzel in 1972, and sometimes even plastic poker chips. Later versions let children pick their own snack, such as marshmallows, M&M’s or animal crackers.

What came out

Seeing the reward made waiting harder. In 1970, children who had both treats in front of them waited on average just over 1 minute. When neither treat was in sight, the average rose to about 11 minutes, and six of the eight children in that group held out for the full 15. Some found their own ways to cope, such as covering their eyes or resting their heads on their arms.

What children thought about mattered more than what they saw. In 1972, children left with no distraction gave up in under a minute on average. With a toy, they waited about 8½ minutes. Told to think of something fun, they waited about 12 minutes. Thinking sad thoughts helped no more than thinking about the food. And when the treats were hidden, children who were told to think about them lasted under a minute, while those told to think of fun things waited almost the whole 15 minutes.

Mischel concluded that at this age self-control is largely a matter of attention: children wait longer when they stop thinking about what they want.

The follow-ups. Around ten years later, the team sent questionnaires to the parents of former participants. In the 1988 report, based on 95 children, those who had waited longer in preschool were described by their parents as more academically and socially competent, more attentive and better at coping with frustration and stress.

The 1990 report added SAT scores, as reported by the parents, for 94 children. The strongest links appeared only in one version of the task: treats in plain view, with no hints on how to wait. In that group, waiting time correlated with SAT maths at r = .57 and with SAT verbal at r = .42. Those figures are based on 35 children. The authors themselves warned that the sample was small and middle-class, and that shared family environments might explain the link. A footnote also showed that the verbal correlation came almost entirely from the girls.

In 2011 a team led by B. J. Casey retested 59 former participants in their forties. Those who had been consistently low in self-control found it harder to hold back responses to happy faces.

What really happened (replications and critiques)

A larger, more varied sample. In 2018 Tyler Watts, Greg Duncan and Haonan Quan tested the same idea with data from a large US study of child development. About 900 children had done a similar task at age four and a half, and their school results were measured at 15. The researchers focused on 552 children whose mothers had not finished college. The task was shorter (7 minutes at most) and children picked their own treat, such as marshmallows, M&M’s or animal crackers.

What they found:

The authors argued that waiting time may reflect broader abilities and circumstances more than self-control on its own, so teaching children only to wait would probably do little.

The response. Other researchers pushed back. Doebel, Michaelson and Munakata wrote a commentary arguing that the ability to wait probably still matters for school achievement. Armin Falk, Fabian Kosse and Pia Pinger reanalysed the comparison and argued that Watts and colleagues had controlled for so many variables that some of them likely absorbed part of the effect itself, a problem they call “bad control”. Keeping only the controls they saw as clearly appropriate, they reported a link that was weaker than Mischel’s but still substantial, and they estimated that the shorter 7-minute limit alone explained about a third of the gap between the two studies. Watts and Duncan replied, defending their choice of controls. The disagreement is mostly about which variables to control for and how to read the results.

Not waiting can be a rational choice. In 2013 Kidd, Palmeri and Aslin at the University of Rochester first gave 28 children an art activity. For half of them, the adult promised better art supplies, then better stickers, and broke both promises. For the other half, the adult kept both promises. In the marshmallow task that followed, children in the unreliable condition waited about 3 minutes on average, and only 1 of 14 lasted the full 15 minutes. With the reliable adult, they waited about 12 minutes, and 9 of 14 lasted the full time. For a child who has learned that promises are often broken, eating the treat now is a reasonable bet.

Children have not got worse at waiting. In 2018, Carlson and colleagues, including Mischel himself, compared 840 children tested on comparable versions of the task in the 1960s, the 1980s and the 2000s. Children in the 2000s waited about 2 minutes longer than those in the 1960s, contrary to what most adults in their survey expected.

Ethics

The children were not deceived. The rules were explained, each child practised calling the experimenter back before the wait began, and every child received a treat at the end, either the preferred one or the other. The early papers do not describe parental consent procedures or ethics review.

The discomfort was mild: up to 15 minutes alone with a temptation, with a way out always within reach. The 2013 Rochester variant went further, since an adult deliberately broke promises to the children, a small deception that would need clear justification and a kind explanation afterwards.

What it means today

The most solid finding is also the most practical one: young children wait better when they are helped to stop thinking about what they want, by hiding it, having something else to do or thinking about something fun.

The long-term prediction is far less certain. A four-year-old’s waiting time is linked to later school results, but weakly, and it is tangled up with family circumstances, early abilities and whether the child has reason to trust adults’ promises. Treating one short wait as a forecast of a child’s future goes beyond what the evidence supports.

A few ideas for everyday life, which go beyond what these studies tested directly:

Would you have passed?

Coming soon. A treat sits on the screen with a timer and a bell. Choose: ring now, or wait for two. Along the way you can try the tricks the Stanford children used (cover the treat, think about something fun) and see how your wait compares with the 1970 and 1972 averages. One honest caveat: an adult at a screen is not a four-year-old in a quiet room, and a picture of a marshmallow is far less tempting than a real one.

Questions people ask

Was the treat in the marshmallow test always a marshmallow?

No. In the 1970 study children chose between animal cookies and pretzels. The 1972 study used a marshmallow and a stick pretzel. Other studies in the series used pretzels or even plastic poker chips, and later versions let children pick snacks such as M&M's or animal crackers.

Does the marshmallow test predict success in life?

Only weakly, as far as current evidence goes. A 1990 follow-up found strong correlations with SAT scores, but in just 35 children. A 2018 study of about 900 children found a smaller link with school results at age 15; among the 552 children whose mothers had not finished college, it shrank by about two-thirds once family background was taken into account. No link was found with behaviour.

What helped children wait longer?

Not looking at or thinking about the treat. Children waited far longer when the treats were out of sight, when they had a toy, or when they were told to think about something fun. Thinking about the treats themselves made the wait much shorter.

Is a child who eats the treat right away lacking self-control?

Not necessarily. In a 2013 study, children who had just seen an adult break two small promises waited about 3 minutes on average, against about 12 minutes for children whose adult had kept them. Not waiting can be a sensible choice when the promised reward seems unlikely to arrive.

Sources

  1. Mischel, W., & Ebbesen, E. B. (1970). Attention in delay of gratification. Journal of Personality and Social Psychology, 16(2), 329–337. doi:10.1037/h0029815
  2. Mischel, W., Ebbesen, E. B., & Zeiss, A. R. (1972). Cognitive and attentional mechanisms in delay of gratification. Journal of Personality and Social Psychology, 21(2), 204–218. doi:10.1037/h0032198
  3. Mischel, W., Shoda, Y., & Peake, P. K. (1988). The nature of adolescent competencies predicted by preschool delay of gratification. Journal of Personality and Social Psychology, 54(4), 687–696. doi:10.1037/0022-3514.54.4.687
  4. Shoda, Y., Mischel, W., & Peake, P. K. (1990). Predicting adolescent cognitive and self-regulatory competencies from preschool delay of gratification: Identifying diagnostic conditions. Developmental Psychology, 26(6), 978–986. doi:10.1037/0012-1649.26.6.978
  5. Casey, B. J., Somerville, L. H., Gotlib, I. H., Ayduk, O., Franklin, N. T., Askren, M. K., Jonides, J., Berman, M. G., Wilson, N. L., Teslovich, T., Glover, G., Zayas, V., Mischel, W., & Shoda, Y. (2011). Behavioral and neural correlates of delay of gratification 40 years later. Proceedings of the National Academy of Sciences, 108(36), 14998–15003. doi:10.1073/pnas.1108561108
  6. Kidd, C., Palmeri, H., & Aslin, R. N. (2013). Rational snacking: Young children's decision-making on the marshmallow task is moderated by beliefs about environmental reliability. Cognition, 126(1), 109–114. doi:10.1016/j.cognition.2012.08.004
  7. Watts, T. W., Duncan, G. J., & Quan, H. (2018). Revisiting the marshmallow test: A conceptual replication investigating links between early delay of gratification and later outcomes. Psychological Science, 29(7), 1159–1177. doi:10.1177/0956797618761661
  8. Carlson, S. M., Shoda, Y., Ayduk, O., Aber, L., Schaefer, C., Sethi, A., Wilson, N., Peake, P. K., & Mischel, W. (2018). Cohort effects in children's delay of gratification. Developmental Psychology, 54(8), 1395–1407. doi:10.1037/dev0000533
  9. Doebel, S., Michaelson, L. E., & Munakata, Y. (2020). Good things come to those who wait: Delaying gratification likely does matter for later achievement (A commentary on Watts, Duncan, & Quan, 2018). Psychological Science, 31(1), 97–99. doi:10.1177/0956797619839045
  10. Falk, A., Kosse, F., & Pinger, P. (2020). Re-revisiting the marshmallow test: A direct comparison of studies by Shoda, Mischel, and Peake (1990) and Watts, Duncan, and Quan (2018). Psychological Science, 31(1), 100–105. doi:10.1177/0956797619861720
  11. Watts, T. W., & Duncan, G. J. Controlling, confounding, and construct clarity: Responding to criticisms of 'Revisiting the marshmallow test'. Psychological Science.