Y11W20VC What the marshmallow test really predicted

In the late 1960s, Walter Mischel left four-year-olds alone with a marshmallow. They could eat it now, or wait and get two later. Decades later, researchers followed up and found that the children who had waited had better life outcomes on almost every measure. The study became famous. The conclusion was: willpower matters. Then, in 2018, a replication changed the story. This week's article examines what the research actually predicted.

Core Vocabulary

delay

/dɪˈleɪ/|de·lay

noun

The act of putting something off to a later time; a period by which something is postponed or held back.

Word family: delay (v.), delayed (adj.), delaying (v. progressive)

Synonyms: postponement, deferral, wait

Collocations: delay of gratification, without delay, brief delay

Example: The marshmallow test was designed to measure a child's capacity for delay — the ability to wait for a better outcome.

In the articleThe ability to delay gratification at four, the story says, predicts most of what matters in adult life.

gratification

/ˌɡrætɪfɪˈkeɪʃən/|grat·i·fi·ca·tion

noun

The pleasure or satisfaction obtained from getting what one wants or needs; often used specifically of immediate pleasure.

Word Breakdown: -ation (noun-forming suffix); from Latin *gratificari* "to show favour"

Word family: gratify (v.), gratifying (adj.), gratified (adj.)

Synonyms: satisfaction, pleasure, reward

Collocations: delay of gratification, immediate gratification, gratification of desire

Example: Young children typically seek immediate gratification and find it difficult to resist the marshmallow placed in front of them.

In the articleThe ability to delay gratification at four, the story says, predicts most of what matters in adult life.

correlate

/ˈkɒrəleɪt/|cor·re·late

verb

To have a mutual relationship or statistical connection with something else; to be associated without necessarily causing one another.

Word Breakdown: co- (prefix "together") + relate (to connect)

Word family: correlation (n.), correlated (adj.), correlating (v.)

Synonyms: relate, associate, connect

Collocations: correlate with, strongly correlate, correlate positively

Example: Children who waited longer in the marshmallow test were found to correlate with better academic outcomes — but correlation is not causation.

confound

/kənˈfaʊnd/|con·found

verb

To complicate or distort the results of a study by introducing factors that muddle the interpretation; to interfere with a clear conclusion.

Word Breakdown: con- (prefix "together") + from Latin *fundere* "to pour"

Word family: confounding (adj./n.), confounded (adj.), confounder (n.)

Synonyms: distort, muddle, complicate

Collocations: confound the results, confound interpretation, confounding variable

Example: Socioeconomic status was shown to confound the original marshmallow results — children from wealthier homes could wait longer for reasons unrelated to self-control.

disadvantage

/ˌdɪsədˈvɑːntɪdʒ/|dis·ad·van·tage

noun

An unfavourable circumstance or condition that reduces one's chances of success or makes a situation harder.

Word Breakdown: dis- (prefix "not") + advantage (a favourable condition)

Word family: disadvantaged (adj.), disadvantageous (adj.)

Synonyms: drawback, handicap, setback

Collocations: socioeconomic disadvantage, at a disadvantage, overcome disadvantage

Example: Children from backgrounds of socioeconomic disadvantage were more likely to eat the marshmallow immediately — not from weak self-control, but because their environment made caution rational.

robust

/rəʊˈbʌst/|ro·bust

adjective

Well-supported, reliable, and holding up consistently across different studies, conditions, or populations.

Word family: robustly (adv.), robustness (n.)

Synonyms: strong, reliable, well-established

Collocations: robust finding, robust evidence, robust methodology

Example: The original marshmallow test findings appeared robust for decades — until a large replication study called them into serious question.

predictive

/prɪˈdɪktɪv/|pre·dic·tive

adjective

Having the ability to forecast or anticipate future outcomes; providing evidence that something can predict what will come.

Word Breakdown: -ive (adjective-forming suffix)

Word family: predict (v.), prediction (n.), predictability (n.)

Synonyms: forecasting, indicative, prognostic

Collocations: predictive of success, predictive value, predictive tool

Example: Mischel claimed that marshmallow test performance was predictive of academic success, SAT scores, and even body-mass index decades later.

reinterpret

/ˌriːɪnˈtɜːprɪt/|re·in·ter·pret

verb

To interpret something again, reaching a different or revised understanding from the original interpretation.

Word Breakdown: re- (prefix "again") + interpret (to explain the meaning of)

Word family: reinterpretation (n.), reinterpreted (adj.)

Synonyms: reassess, revise, reconsider

Collocations: reinterpret the data, reinterpret findings, reinterpret a result

Example: Watts and colleagues reinterpreted the marshmallow data, showing that socioeconomic background — not just willpower — explained much of the original finding.

Technical Terms

delay of gratification

/dɪˈleɪ əv ˌɡrætɪfɪˈkeɪʃ(ə)n/|de·lay of grat·i·fi·ca·tion

noun phrase

the capacity to postpone a smaller immediate reward for a larger later one

Synonyms: impulse control, deferred reward, self-control over immediate desire

Collocations: ability to delay gratification, delay of gratification predicts, failure to delay gratification

Example: The original marshmallow test was designed to measure delay of gratification, but subsequent research revealed that a child's willingness to wait reflected the reliability of their environment as much as their self-control.

In the articleThe ability to delay gratification at four, the story says, predicts most of what matters in adult life.

executive function

/ɪɡˈzɛkjʊtɪv ˈfʌŋkʃ(ə)n/|ex·ec·u·tive func·tion

noun phrase

the set of cognitive processes including planning, inhibition, and working memory

Synonyms: cognitive control, higher-order regulation, self-regulatory capacity

Collocations: executive function develops, poor executive function, executive function and self-control

Example: Executive function — the set of cognitive processes governing planning, flexibility, and impulse inhibition — predicts academic and life outcomes more consistently than IQ across a range of longitudinal studies.

confounding variable

/kənˈfaʊndɪŋ ˈveəriəb(ə)l/|con·found·ing var·i·a·ble

noun phrase

a hidden factor that influences both measured variables, producing misleading correlations

Synonyms: confound, lurking variable, third variable

Collocations: control for a confounding variable, confounding variable distorts, identify the confounding variable

Example: The apparent link between shoe size and reading ability in children is entirely explained by a confounding variable — age — which drives both, rendering the original correlation meaningless.

replication

/ˌrɛplɪˈkeɪʃ(ə)n/|rep·li·ca·tion

noun

the reproduction of a research finding in a new study

Synonyms: independent verification, reproduction of results, scientific confirmation

Collocations: replication attempt, replication fails, replicate the finding

Example: The inability of multiple independent labs to replicate the original marshmallow results using larger and more diverse samples raised serious questions about how broadly the initial findings could be applied.

In the articleThe replication that changed the picture In 2018, a team led by Tyler Watts at New York University, with Greg Duncan and Haonan Quan, did something the original research hadn't: they ran a much larger version, with a more diverse group of more than 900 children, and they controlled for family background.

socioeconomic status

/ˌsəʊsɪəʊˌiːkəˈnɒmɪk ˈsteɪtəs/|so·cio·e·co·nom·ic sta·tus

noun phrase

a composite measure of income, education, and occupation

Synonyms: SES, socioeconomic position, social class standing

Collocations: control for socioeconomic status, socioeconomic status predicts, low socioeconomic status

Example: When researchers controlled for socioeconomic status, the predictive power of the marshmallow test dropped substantially — suggesting the test was measuring the stability of a child's environment rather than an innate trait.

In the articleThe simple correlation between waiting time and later outcomes was still there — but almost all of it disappeared once they accounted for the child's socioeconomic background.

Figurative Phrases

the test of time

long

Etymology/Type: term validation — idiom; not a literal test

Synonyms: durability under scrutiny, continued validity after sustained examination, survival of ongoing review

Example: The marshmallow test was assumed to have stood the test of time, but the replication study using a far larger and more diverse sample suggested the original finding was more contingent on circumstance than previously believed.

delay gratification

postpone pleasure

Etymology/Type: non-literal; 'gratification' replaces 'pleasure' with specific resonance

Synonyms: wait for a larger reward instead of taking an immediate smaller one, exercise patience over impulse, defer satisfaction

Example: Children who could delay gratification in the original study were tracked over decades, producing findings that appeared to predict everything from academic performance to health — until researchers began asking what else had been predicted simultaneously.

In the articleThe ability to delay gratification at four, the story says, predicts most of what matters in adult life.

hold out

resist temptation

Etymology/Type: idiom; not physical holding

Synonyms: wait, resist the temptation to take the immediate option, persist without giving in

Example: The ability to hold out for the second marshmallow turned out to depend heavily on how reliable the experimenter had been in a prior interaction — trust in the environment predicting patience as much as internal self-control.

stand the test

survive examination

Etymology/Type: idiom from quality-testing

Synonyms: endure scrutiny, remain valid under examination, survive critical assessment

Example: Few psychological findings have been as celebrated as the marshmallow study, and few have failed so publicly to stand the test of large-scale independent replication.

go down in history

become permanently famous

Etymology/Type: idiomatic; no literal going

Synonyms: be remembered, be recorded as significant, be cited as an important example

Example: The marshmallow test went down in history as one of psychology's most compelling demonstrations — which made the replication failure all the more important to understand and communicate.

fall apart

collapse

Etymology/Type: idiom; not literal disintegration

Synonyms: collapse, fail under scrutiny, disintegrate when examined

Example: The original interpretation began to fall apart when researchers found that controlling for socioeconomic background removed most of the predictive power that had made the study famous.

Confusing Words

correlate vs cause

This is one of the most important distinctions in scientific reasoning — two things being related is not the same as one producing the other.

  • correlate — to be statistically associated; to vary together in a consistent pattern. Two variables correlate when changes in one tend to accompany changes in the other, without any claim about which produces which, or whether a third factor produces both. Correlation is symmetric and observational.
  • cause — to produce an effect; to be the mechanism by which one event brings about another. Causation is directional and requires more than statistical association — it requires that the relationship holds under experimental manipulation, with confounds controlled. Causation makes a stronger and harder-to-establish claim.

If describing a statistical relationship between two variables, use correlate. If claiming that one variable produces or brings about another, use cause.

confound vs confuse

In everyday language these words overlap, but in scientific contexts confound carries a precise technical meaning that is quite different from simple confusion.

  • confound — in research, to introduce a variable that is correlated with both the independent and dependent variables, making it impossible to determine what is actually causing the observed effect. A confound does not merely make things unclear; it systematically misleads by providing an alternative explanation for results.
  • confuse — to cause a lack of clarity or understanding; to mix things up in a way that makes comprehension difficult. Confusion is general — it can apply to a person's state of mind, to a tangled argument, or to an ambiguous signal. It does not carry the specific methodological meaning of confound.

In research contexts, use confound for a variable that provides an alternative causal explanation. In general contexts, use confuse for anything that creates a lack of clarity.

reinterpret vs misinterpret

Both words describe arriving at a different meaning than the original or expected one, but they differ in whether that alternative reading is legitimate or erroneous.

  • reinterpret — to interpret again, arriving at a new or different meaning that is legitimate and considered. To reinterpret data is to examine it with a fresh analytical lens, producing a different but defensible reading. Reinterpretation is a scholarly and creative act.
  • misinterpret — to interpret incorrectly; to arrive at a meaning that is wrong. Misinterpretation is an error — the understanding produced is false, not simply different. Where reinterpretation opens new valid readings, misinterpretation closes off the correct one.

If describing a legitimate alternative reading that is defensible on its own terms, use reinterpret. If describing an incorrect understanding that fails to capture the true meaning, use misinterpret.