For a long time the famous "marshmallow studies" by Walter Mischel were regarded as prototypical experiments to measure self-control. However, in some studies only low correlations to other measures of self-control were observed. Thus, in recent years alternative hypotheses were tested. For example: If the delay of gratification task is regarded as problem solving situation, longer waiting times could be explained by problem solving skills (intelligence). Other scientists hypothesize, that the decision to wait might be linked to the amount of trust in the experimenter (Kidd et al. 2013). But are these new research approaches adequate to explain the amazing predictive power of the marshmallow experiment?
Keywords: Marshmallow Test - Marshmallow Experiment - Delay of Gratification - self-control - intelligence - social skills - grade point average - frustration tolerance - predictive validity - Duckworth - Kirby - replication - social trust - social psychology - cognitive psychology - educational psychology - mind - brain - developmental psychology - delay discounting - Walter Mischel
Hello and welcome back to the world of psychology.
Is it really self-control?
The predictive power of the delay of gratification task.
In the last episode we heard about the amazing predictive power of the marshmallow experiments.
Using a quite simple experimental design, in essence telling a four year old kid,
well, you can either have one marshmallow right now or you wait until I come back and then you can have two marshmallows.
It was possible to predict the children's academic competence, social competence and coping competence ten years later.
And even more, children who were able to wait longer in the delay of gratification task on average scored higher in the SAT test.
Even though we heard that these correlations were based on really small sample sizes
and probably overestimated the real correlations between delay time and SAT performance,
I also reported a very new study published in 2013 by Duckworth, Tsukayama and Kirby
in which a correlation of 0.21 between grade point average and a correlation of 0.27
between standardized achievement test scores and delay times were found.
So the correlations were not that high, but they were strongly significant.
So this new study is a really important replication of the original results by Walter Mischel and colleagues.
But Duckworth, Tsukayama and Kirby were not only interested in replicating the original results,
they were examining another really important research question, which was, and this is the title of their study,
and I have to confess I stole the title of this episode from them,
the title of their study was, is it really self-control?
Examining the predictive power of the delay of gratification task.
But how on earth could it be something else?
I mean, if you look at the videos of the kids who are fighting with themselves,
trying to distract their attention, closing their eyes, but then again smelling at the marshmallow and
no, I can't eat it, I have to wait, I have to fight, I want two marshmallows.
And then again they turn their back to the marshmallow and they are really having a hard time
showing all their self-control in this experiment.
So how on earth could it be something else?
Well, think about it. In a way, the delay of gratification task is a problem.
You want to have the second marshmallow and you have to find a way to get it.
So you have to develop strategies.
So maybe the child, which is able to come up with a good strategy,
like closing his or her eyes or like daydreaming about other distracting activities,
the kid that comes up with these clever strategies might be more intelligent.
And as we all know, intelligence is a really good predictor for academic success,
but also many other factors in life.
So maybe the predictive validity, the predictive power of the delay of gratification task
stems from intelligence and not from self-control.
To find out what is really going on here,
Duckworth, Tsukayama and Kirby report data of 966 4-year-old participants
who not only took part in a delay of gratification task,
but were also assessed in respect to self-control.
So on the one hand, the children were asked themselves how impulsive they are.
And on the other hand, they also had teacher ratings.
The teachers had to answer items like, and now I quote from the study,
this student's mind wandered when he or she should have been listening.
But not only self-control was assessed,
they also conducted an intelligence test with the 4-year-old kids.
And all they had to do now was to correlate the delay times of the kids
with on the one hand self-control and on the other hand intelligence.
And surprisingly, intelligence in fact correlated higher with delay times than self-control.
So children being more intelligent are in fact able to wait longer,
probably because they find better strategies.
But is intelligence also responsible for the strong predictive power of the delay of gratification task?
Yes, but only for predicting standardized achievement test scores.
Grade point average, risky behavior and body mass index
10 years later were much better predicted by self-control.
So intelligence is important to find suitable strategies in the delay of gratification task,
but it doesn't explain the predictive power of the marshmallow experiments.
All alone, in fact, self-control seems to be,
especially in respect to the grade point average of the children,
self-control seems to be much more important.
Another recently published and very interesting study
that questions whether it's really about self-control in the delay of gratification task
is the study by Maguire and Cable published 2013 in Psychological Review.
Maguire and Cable hint at an aspect of the marshmallow experiments that is often overlooked.
They say, isn't it strange that children who waited 4 or 5 minutes
suddenly decide, no, I'm not going to wait any longer, I want the marshmallow right now.
Because they already waited 5 minutes,
so they could expect that they don't have to wait very much longer.
It seems irrational to suddenly give up.
But Maguire and Cable argue that this behavior was only irrational
if the children had known how long they have to wait.
In the original studies by Walter Mischel and colleagues,
children were not told how long they have to wait.
And maybe, after waiting 5 minutes, they think,
well, probably this will take very long,
I'll better ring the bell and eat the one marshmallow now.
I don't want to wait 2 hours or something.
So the question is, how long do people estimate they have to wait
if they don't know when a positive result is going to occur.
To examine this question, Maguire and Cable conducted several experiments
in which all in all 646 participants took part
and they asked the participants to imagine different scenarios.
For example, imagine you are on a diet
and after 5 days of diet, nothing happened.
You didn't shed one single pound yet.
What do you think, how many days you will have to wait until something happens?
Then, on average, the participants said, well, probably 9 or 10 more days.
And when the participants were asked,
okay, now imagine you are on a diet for 15 days now and nothing happened.
What do you think, how much time has to pass until the diet shows first results?
Then the participants said, 15 days.
And after 30 days without any results, they said,
well, I probably have to wait 30 more days.
So the more time elapsed, the more they said,
the more time has to pass until anything happens.
And similar results were found for running, for studying or practicing piano
in hopes of playing a piece without mistakes.
In comparison, when participants were asked,
okay, imagine you are watching a movie and 40 minutes have elapsed up to now.
What do you think, how long will it take till the movie ends?
Then participants say, okay, 60 minutes.
And after 100 minutes of watching movie, participants say,
well, probably only 20 more minutes.
So if you know how long something takes,
people know the regular duration of a movie.
If you know how long something takes,
your predictions, how long it's gonna take,
your time predictions decline.
This is common sense.
But in the delay of gratification tasks,
the children don't know how long it's going to take.
And maybe there are inter-individual differences
between the children in predicting how long it will take.
So maybe some children, after two minutes,
begin to think, well, this will take two more hours or something.
Well, let's better stop it here.
Whereas other children, after two minutes,
have another prediction and they think,
well, maybe five more minutes or ten minutes.
I can wait so long.
As plausible all that sounds,
it's not clear how this can predict better academic performance
or better social competence.
So maybe it's similar to what we heard about intelligence before.
Some variance in the delay of gratification task
can be explained by different time predictions.
But it's questionable whether this variance is responsible
for the predictive validity,
the predictive power of the delay of gratification task.
Future studies will have to cover this question.
In the experiments by Maguire and Cable,
participants obviously thought,
well, it probably will take very long until anything happens.
But what happens if participants think,
maybe there won't happen anything at all.
Maybe I won't get my gratification.
Can I trust the experimenter?
Can I trust the guy on the other side
to really give me the gratification I deserve?
So the line of research we will get to know now
is all about social trust.
And there are two studies published in 2013
I want to present to you.
In the first study, two experiments were conducted
with adult participants
who were given three different descriptions
of either a trustworthy, neutral or not trustworthy person.
And then they were asked,
well, imagine person A, the trustworthy one,
offered you $40 right now or $65 in 70 days.
What would you choose?
Would you go for the instant gratification
or would you go for the delayed one,
which is obviously clearly higher.
And the results were not really surprising.
The more trustworthy the person was characterized,
the participants were more willing to delay gratification.
Probably because they were convinced,
well, on this guy I can rely, he will give me the money.
However, one weakness of this study was
that all participants were presented
with the three different vignettes.
Maybe somehow they might have figured out
what the hypothesis of the experimenters was
and maybe they behaved accordingly to this hypothesis.
In fact, this is something that occurs quite frequently
in psychological research.
So Michaelsen and colleagues conducted a second experiment
in which they had three groups of participants
which were presented only one description.
And additional to the descriptions,
now there were also photos, portraits of the other person.
And these portraits were selected based on previous research
according to social trust and facial expressions.
So to the trustworthy description,
a picture of a really trustworthy looking guy was attached.
These weren't real persons, but they were computer generated.
And if you create bigger eyes,
you get a baby face expression
which is much more trustworthy
than smaller eyes that are closer to the eyebrows.
So if you want, take a look in the mirror
and if you have really huge eyes, baby face-like eyes,
you are probably regarded as very trustworthy and nice.
Again, participants were asked,
well, would you prefer, for example, just $40 now
or $65 in 70 days?
If this is the guy you will get the money from.
Replicating their first experiment,
they found if the partner was untrustworthy,
participants were not willing to delay gratification that often.
Surprisingly, in this second experiment,
there was no significant difference
between the neutral condition and the trustworthy condition.
So the effect of the manipulation was only visible
when comparing the untrustworthy condition with the other two conditions.
Delaying gratification obviously does not occur in a social vacuum.
It really depends on social trust.
And this is also confirmed by the last study
conducted by Kitt and Palmieri,
published in the journal Cognition.
In this study, 28 children aged about 4 to 6 years
were randomly assigned to two conditions.
In the one condition, the experimenter was reliable.
In the other one, he wasn't.
Before they conducted the original marshmallow experiment,
they had a, let's say, probe trial.
They showed to children, for example, some crayons
that were in really bad condition.
And they said, well, if you wait two and a half minutes until I come back,
I'll bring you really great crayons.
Many more crayons and brand new crayons.
In both conditions, all children decided to wait.
When the experimenter came back in the trustworthy condition,
the children got the brand new crayons.
In the untrustworthy condition, the experimenter came back and said,
and now I quote from the study,
I'm sorry, but I made a mistake.
We don't have the crayons.
But you can use these instead.
And then the children had to play with the rubbish crayons.
After the children had learned,
well, my experimenter is not that trustworthy.
Or in the other condition, well, I can rely on the experimenter.
The original marshmallow experiment was conducted with them.
And children in the trustworthy condition
waited on average 12 minutes and 2 seconds.
Whereas in the untrustworthy condition,
children waited just 3 minutes and 2 seconds.
So what these experiments tell us
is that the time children wait in the delay of gratification task
depends on how much they trust the experimenter.
And for sure, there are also inter-individual differences
between children in regard to social trust.
So maybe at least some of the predictive validity
of the delay of gratification task depends on social trust.
And I think it really makes sense
that children who are able to show more trust in their environment
are also doing better at school
or show better social competence 10 years later.
On the other hand, these experiments show us something else.
And I think that's important to mention
because some parents might be willing
to conduct the marshmallow experiment with their children.
The performance of the kid in the marshmallow experiment
can strongly depend on who is conducting the experiment
and what happened before the experiment.
So if you want to conduct the experiment with your child,
I mean, I think it can be great fun,
please don't interpret the results too seriously.
The original experiments were conducted
under standardized, controlled conditions.
And if you do it at home, it probably has no predictive validity.
But nevertheless, it's probably much fun.