The Marshmallow Effect | 1

Although the concepts of immediate and delayed gratification have been widely used in both psychology and sociology for over 50+ years, perhaps their most well-known application has been through Mischel’s “Marshmallow Test”. This experiment has seen its influence spread from the world of academic psychology to the wider shores of both the popular imagination and, more-interestingly perhaps, into many schools and colleges.

To understand why this has been the case – and to suggest why this should be a cause for concern – we need to look briefly at the test itself and a range of contemporary psychological and sociological critiques.

What Is It?

Although the ideas underpinning the Marshmallow Test have, in one shape or another, been around since at least the late 1950’s – Mischel’s (borderline racist – maybe to be charitable they were different times…) 1958 study of children in Trinidad arguably being one of the first – the test itself involves a young (4 years old in Mischel’s 1990 study) child being presented with a desired object (such as a marshmallow) and told the examiner will leave the room for 15 minutes. If the child can delay eating the marshmallow until the examiner returns they will be given a second marshmallow to eat.

As Mischel pput it, the main objective of these studies was “to create a conflict for young children between the temptation to stop the delay and the desire to persist for the preferred outcome when the latter required delay”.

Although the 1958 study looked specifically at cultural influences in the behaviour of two distinct ethnic groups (the “presence or absence of the father within the home, age, and socioeconomic status”), later studies focused more on individual gender differences based around measured differences in “willpower” or self-control. The 1990 study, for example, tested “male and female preschoolers in the Bing School of Stanford University, a preschool for mostly middle-class children of faculty and students from the Stanford University community”.

Why Is It Important?

As a stand-alone piece of research the Marshmallow Test is interesting, but what catapulted it into the wider public consciousness was the follow-up studies Mishcel et al (1990) carried-out when their original respondents were 16 and 18. More-specifically they found their respondents SAT scores reflected their ability to delay gratification: in a nutshell, the 4 year old students who were able to delay their gratification had higher levels of academic achievement at age 16 than those who didn’t.

To cut a long story short, the relationship between self-control and academic achievement not only became an accepted part of the discourse – if you wanted to be successful, both academically and in later life, you needed to develop and exercise self-control from a young age – it also spawned a small sub-genre of self-help manuals designed to teach “willpower” and, more-worryingly perhaps, found its way into the curriculum of a significant number of American schools.

Toasted?

So, is the hype justified? Let’s approach this from a few different directions, starting with some general observations about the Tests.

The first thing to note is the relatively small sample size. The original tests took place over a six-year period (1968 -1974) and involved a total of 650 boys and girls who took at least one test.  By the time of the follow-up comparison of these subjects’ SAT scores the sample had degraded by just over 50% (leaving around 190 respondents). So, not only was the original sample small, sample erosion leaves a big question mark hanging over the relationship between the original Test and the SAT comparison. This is because we know nothing about the SAT scores of over 50% of the original sample and this makes drawing conclusions about the relationship between “self-control” and academic achievement tentative at best.

A further dimension here is the sample was drawn from a very narrow pool: “a preschool for mostly middle-class children of faculty and students from the Stanford University community”. This means it’s not possible to generalise the results from one very select, ethnically-homogeneous, affluent community to the wider population – although this hasn’t stopped an awful lot of people from trying…

One Test – or Many Tests?

One of the things about the Marshmallow test is that although it’s often described (and shown) as a child sitting at a table with a single marshmallow in front of them, this was only one form of the test. In all Mischel et al (1990) used four test conditions, where “ideation” refers to the experimenter providing the subjects with ideas about how to stop themselves thinking about eating the marshmallow:

Condition 1. Marshmallow exposed / no ideation

Condition 2. Marshmallow exposed / ideation

Condition 3. Marshmallow obscured / no ideation

Condition 4. Marshmallow obscured / ideation

The important point here is that only the first condition (the child has the marshmallow in view and is given no instructions about how to stop themselves thinking about eating it) resulted in a statistically significant correlation with subsequent SAT scores. Mischel et al also warn that even this set of conditions may “lose its diagnostic potential early in the course of development”.

In other words, the Test may only be an effective diagnostic tool – predicting future academic achievements – when used with very young children. This follows, they argue, because “As children become older, their delay behavior (sic) rapidly becomes less responsive to this manipulation”. To put this another way, as children develop they “become aware of basic rules to facilitate delay, for example, by distracting their attention from the rewards…and can purposefully influence their own ideation to overcome the impact of exposed rewards”.

The key takeaway here is that if the correlation only appears under very specific experimental conditions any attempt to translate the Test results into real world behaviour – such as teaching children “self-control” in schools – is unlikely to work if the aim is to somehow “improve academic achievement”: by the time children reach school age they will have already begun to learn the “techniques of self-control” required to effectively bias the test.

While this is an important qualification, a much more-significant question relates to the relationship between the ability to delay gratification in pre-school children and their subsequent academic development. While the original studies found a correlation, this is a long way from saying the former causes the latter – even if, for the sake of argument, we assume the fact that 50% of the original test takers who were not tracked into adulthood did not materially affect the relationship (and that’s a Really Big If…).

Stay tuned for Part 2…