It seems that most people are familiar with the Stanford Marshmallow experiment from 1970: “one of the best known studies in the history of psychology” is the line I see used.

Well good for everyone else. Because I only encountered it in a book by Richard Wiseman that I read in my mid-twenties. And by then, it was too late.


Anyway, this post is going to do two things:

  1. Explain how Walter Mischel’s test actually worked; and
  2. Engage in some obnoxious self-congratulation around the outcome of a subsequent revision of the experiment.

Walter Mischel’s Stanford Marshmallow Experiment

Here is the basic concept (glossing over all of the controls and tests to make sure that the child in question could understand english and instructions):

  • Take a sample of 32 nursery school children aged between 3½ and 5½, equally split between boys and girls;
  • Place them, one at a time, in a room that consists mostly of a table, a chair, and a plate with a marshmallow on it (or an Oreo biscuit – allow the child its sugar rush of choice);
  • Offer the child a deal
  • As in: “So here’s the deal. This marshmallow is for you. You can either have this marshmallow now; or if you wait for me to come back, I’ll give you another one and then you can have two. It’s up to you. You either eat one now; or if you wait to eat it, you can eat two when I come back. Okay?”
  • Then leave the room.
  • Keep tabs on the children as they grow up.
  • See how they do in life. And how they do in school. And when the technology becomes available, perform the test again to see if there are differences in brain function.

Here’s a clip of children trying not to eat the marshmallow, with one entertaining young blond who is already well-versed in the art of denial.

The Results

  • A small number of children ate the marshmallow immediately, but most tried to wait.
  • Waiting strategies included: covering eyes, stroking the marshmallow, and attempting to take small bites from the bottom and then hide it*.
    *Which worked well until they applied the same principle to the top. 
  • A third of the children managed to wait the full 15 minutes.
  • Later on in life, the children that managed to exercise self-control scored significantly higher on their SAT scores (on average, 210 points higher – which I think works out to about 13% of the total possible score based on the scoring system at the time), and were described as more competent by their parents.
  • When those children were eventually tested for different brain functioning in 2011 (41 years later, I guess they weren’t children any more), they did indeed have different brain functioning: the prefrontal cortex was more active in those that were able to delay gratification; while the ventral striatum (an area associated with addictions) was much more active in the instant-gratifiers.
  • Which only stands to reason.
  • Interestingly, the instant gratifiers were shown to be just as capable of restraint when they were tested on dry and boring subject matter – it was only when the reward was “emotionally hot” that the distinction in brain function began to emerge.
  • Which also only stands to reason. Why would you get addicted to something that offers you no reward?
  • For more on the subsequent studies of these children through their lives, here’s a TIME magazine article: The Secrets of Self Control: The Marshmallow Test 40 Years Later

So obviously, this experiment began to be held up as the paragon of self-control. All that’s needed is a little self-restraint, and success is yours, baby.

But when I was planning this post, being the argumentative individual that I am, it occurred to me that this test seems a bit, well, naive. Because self-control is a highly irrational choice when there’s only one packet of marshmallows, and blondie from up top is furiously shoving them down while you wait for 15 minutes like a knob.

Wouldn’t you know it – there’s actually a study that’s been done on exactly that by the University of Rochester…

What Happens When You’re Constantly Disappointed

Celeste Kidd (how’s that for an appropriate surname?) was the lead author on a 2012 paper called “Rational Snacking: Young Children’s Decision-Making On The Marshmallow Task Is Moderated By Beliefs About Environmental Reliability“.

The premise, based on her work in homeless shelters, is that children’s waiting times are very much dependent on what environment they come from.

To test this, the marshmallow experiment was repeated on two separate groups of children (14 in each group), under two different conditions.

Group 1: The Reliable Group

Before the marshmallows were introduced, the children in this group were primed to believe that their environment was one where they could rely on the experimenter.

Each child started with an art project, where there was a container of used crayons in the room. The experimenter told the child that she would go and get newer and better stuff to colour with, left the room, and then returned with a rotating tray full of art supplies.

Then the child was given a small sticker to use on the art project. And again, the experimenter told the child that she would go and get better stickers, and returned a short while later with several large die-cut stickers.

Finally, the child was subjected to the marshmallow test.

Group 2: The Unreliable Group

The same set-ups applied, only each time the experimenter left the room, she would return and apologise that there were no more art supplies, or stickers, and would encourage the child to use the used crayons and small sticker that were already in the room.

And then, as with Group 1, the marshmallow task was set…

As you might guess, 9 of the children in the reliable group were able to hold out, while only one was able to from the unreliable group. Also, the average waiting times:

Reliable Group: 12 minutes and 2 seconds

Unreliable Group: 3 minutes and 2 seconds

And in the original experiment, mean waiting times were around the 6 minute mark.

In the world of psychology, I’m given to understand that this was a resounding statistical win. Resounding enough not to require a control group.

So perhaps it is trust in one’s environment that’s the better indicator of success: precisely because it permits self-control to take place when it’s appropriate?

How I Think That Works In An Office Environment

I have worked with a number of businesses that were experiencing financial difficulty. And unsurprisingly, that made their employees particularly nervous.

And what would happen is exactly what you’d expect:

  • people cash in on their leave (“if I don’t take it, I’ll maybe not get it”)
  • there are suddenly more staff loans being applied for (“let me take what I can now”)
  • the standard of work begins to fall (“why should I work for something that’s not working?”)
  • tempers flare (I’m sure it’s stress, but also “why should I be pleasant to my boss if he/she isn’t going to be my boss for much longer?”)
  • theft (also “let me take what I can now”)

All of which make failure more of an inevitability than a possibility.

And while I realise that I’m observing from the fringe, I can’t help but think that the same thing applies even in successful companies. If you get promised promotions, but don’t get them; or get promised bonuses, but they don’t arrive; then how much more likely are you to take a sick day?

As the second experiment demonstrates, it would be better if those promises were never made at all.

It’s all about expectation management.

I’m just saying.