Delayed gratification

principle · origin: study · evidence: contested

In short

Delayed gratification means choosing a larger but later reward over a smaller but immediate one, and sticking with that choice while you wait. The classic experiments with preschoolers, known as the “marshmallow test”, showed that waiting is easier when attention is directed somewhere other than the reward. The link these studies promised, namely that a child who waits longer at age 4 will do better as a teenager, turned out to be much weaker in a replication with a larger and more diverse sample. And how long a child waits also depends on how much they trust that the promised reward will actually appear.

What it says

How the idea arose. Mischel, Ebbesen and Zeiss (1972) described three experiments with preschool-age children. Each child could get a less preferred reward right away, or wait, without knowing for how long, for a more preferred one. If they gave up waiting, they got only the smaller reward.

The results, as the authors summarise them:

• Children waited much longer when they were distracted from the rewards than when they looked directly at them.
• Not every thought worked as a distraction. Thinking about “fun things” helped. With sad thoughts, children waited only a short time, just as when they thought about the rewards.
• Even when the rewards were not physically in front of them, thinking about them shortened the wait rather than lengthening it.

The authors’ conclusion: whatever makes the reward more “vivid” in the mind shortens the wait, while distraction, external or mental, makes it easier. The emphasis was therefore on attention mechanisms, not on a fixed trait of the “you either have willpower or you don’t” kind.

The long-term follow-up. Shoda, Mischel and Peake (1990) contacted the parents of the children tested in those experiments, more than ten years later. They obtained data on 185 adolescents. They found significant correlations between how long the children had waited in preschool and their cognitive and academic competence and their ability to cope with frustration and stress, as rated by their parents. The study was often read as “whoever waits at 4 does better in life”.

The detail that usually gets lost: the links appeared mainly in a single version of the test, the one in which the rewards were in sight and the children were not given any strategy. The authors called it the “diagnostic” condition. In this condition, waiting time correlated with SAT scores (the American college admission test) at r = 0.42 for the verbal part and r = 0.57 for the quantitative part, but in only 35 children (Shoda et al., 1990). The correlation coefficient r shows how closely two measures go together: 0 means no link, and 1 a perfect link. The authors themselves pointed out that, given the small sample, the correlations may overstate the real association and that replications in other populations are needed.

Example

A four-year-old sits alone in a room, with a small treat on the table. They are told that if they wait until the adult comes back, they will get two. They can call the adult back at any time by ringing a bell, but then they get only one.

Based on the results above, the child will probably wait longer if they keep their attention away from the treat, and less if they look at it or think about it.

The same situation for an adult (an illustrative example, not from a study): you want to save money, but you have a shopping app on your phone, with products picked “for you”. The immediate reward is, in effect, on the table, in plain sight.

How to apply it

The suggestions below are an extrapolation from the experiments with preschoolers to adult life. The studies cited did not test these applications.

• Take the immediate reward out of sight. Attention to the reward shortens the wait (Mischel et al., 1972). Don’t keep sweets on your desk, hide the app, unsubscribe from discount emails.
• Prepare a concrete distraction, not just “I’ll resist”. Decide in advance what you will do at the moment of temptation: a short walk, a phone call, a small task. The implementation intentions method offers a format for this.
• Don’t dwell on the reward, neither the small one nor the large one.
• Check whether it’s worth waiting. The study by Kidd et al. (2013), described below, suggests that not waiting can be a reasonable choice when the promise is not reliable. Before you judge yourself for a “lack of willpower”, ask yourself whether the delayed reward is real and certain.

Limits and nuances

The 2018 replication: the association is much smaller. Watts, Duncan and Quan (2018) revisited the question of the 1990 study using a large US sample (the NICHD Study of Early Child Care and Youth Development). The children were tested at age 4 and a half (54 months), with a version of the task in which the maximum wait was 7 minutes, and outcomes were measured at age 15. The analysed sample had 918 children, and the authors focused on the 552 whose mothers had not completed college. The authors give two reasons: they wanted to see whether Mischel and Shoda’s results also apply to the populations of greater interest to researchers and policymakers concerned with interventions, and in the group of children whose mothers had a college degree the waiting measure was too truncated (the measure was capped at 7 minutes, so differences between children who waited a long time no longer showed up) for the link to be estimated reliably. According to Watts et al. (2018), the sample of 552 children is 10 times larger than the one in the study by Shoda et al. (1990).

What they found, according to the article’s abstract:

• Each additional minute of waiting at 4 and a half was associated with better school achievement at 15, but only slightly: about a tenth of a standard deviation (the standard deviation shows how spread out the results are around the average; a tenth of it is a small difference).
• The raw link, before any adjustment, was about half the size of the one reported in the original studies.
• It shrank by two thirds when the family’s social background, early cognitive ability and the quality of the home environment were statistically taken into account.
• Most of the difference in school achievement came from whether a child could wait at least 20 seconds.
• The links with behaviour at 15 were much smaller and rarely statistically significant.

Our interpretation, in plain terms: a good part of the link attributed to “self-control” disappears when you take into account things that go along with it, such as the family environment and early cognitive ability. The replication does not show that delayed gratification doesn’t matter. It shows that the marshmallow test, on its own, says much less about a child’s future than was believed.

Trust in the environment matters. Kidd, Palmeri and Aslin (2013) tested 28 children, with an average age of 4 years and 6 months. Before the marshmallow test, the children worked on an art project, and the experimenter twice promised them something better: first drawing supplies, then stickers. In the “unreliable” condition, the experimenter came back each time and said they had made a mistake and had nothing to give. In the “reliable” condition, they came back with what they had promised.

Then came the marshmallow test. Children in the reliable condition waited about 12 minutes on average, those in the unreliable condition about 3 minutes. In the reliable condition, 9 of 14 children waited until the end (15 minutes), compared with only 1 of 14 in the unreliable condition. The authors’ interpretation: waiting time also reflects a rational decision, based on how certain it seems that waiting will pay off, not only self-control. They also say that their data do not show that self-control is irrelevant, only that it is premature to attribute most of the differences between children to self-control.

Other limits to keep in mind:

• Shoda et al. (1990) worked with children from a university community, whom they describe as middle-class, who had not been selected for self-regulation difficulties, and within a narrow age range. The authors themselves warn against generalising.
• All these studies are on children. The transfer to adults in the “How to apply it” section is an interpretation, not a finding.
• The study by Kidd et al. (2013) has a small sample (28 children). It convincingly supports an important nuance, but it does not by itself establish how much of the variation between children is due to trust and how much to self-control.

Sources

• Walter Mischel, Ebbe B. Ebbesen, Antonette Raskoff Zeiss (1972). Cognitive and attentional mechanisms in delay of gratification
• Yuichi Shoda, Walter Mischel, Philip K. Peake (1990). Predicting adolescent cognitive and self-regulatory competencies from preschool delay of gratification: Identifying diagnostic conditions
• Tyler W. Watts, Greg J. Duncan, Haonan Quan (2018). Revisiting the Marshmallow Test: A Conceptual Replication Investigating Links Between Early Delay of Gratification and Later Outcomes
• Celeste Kidd, Holly Palmeri, Richard N. Aslin (2013). Rational snacking: Young children's decision-making on the marshmallow task is moderated by beliefs about environmental reliability

See also: Compounding, Ego depletion, First-order negative, second-order positive, Implementation intentions