The predictive power of marshmallows

I have gone through the back catalogue of podcasts for WNYC’s Radiolab for a couple of months now. I got into it after ABC radio substituted it for the Science Show for two weeks in late June. It is sensational – great content and entertaining.

I just listened to the Radiolab podcast on Walter Mischel’s marshmallow experiment . The basic idea of the experiment concerned testing the ability of four-year olds to delay gratification. The experimenters left the children in a featureless room with a marshmallow (or Oreo cookie in later experiments) with a promise of more if they waited. The interesting outcome from the experiment has been how the ability to delay is a strong indicator of future success. For example, the average difference between those who waited 10 seconds and those who waited 15 minutes was around 210 points on the SAT. 210 points is roughly the difference between the 60th and 85th percentile.

I was familiar with this experiment from earlier readings, but was not aware on how this had become such a strong indicator of future success. Mischel and his colleagues are still following a number of the subjects, so there should be more to come.

So what we can take from this experiment? One (optimistic) possibility is that by training children to delay gratification (those who delayed used a variety of tricks) there may be lifelong benefits. Another possibility, and the one to which I lean, is that the experiment is an indicator of a broader personality trait, and that teaching a technique to delay gratification in such a case will not have much long-term impact. I consider that it is more like testing children for IQ – you can do something but it will take some effort.

An interesting question is whether patience is a cause of the delayer’s future success or a symptom of the underlying cause. For example, is this test simply a proxy for intelligence, with more intelligent children able to foresee the consequences of their actions and devise methods to help them delay?