June 4, 2009

Delaying Gratification Is Easy If You Don't Try


One marshmallow or two?

The New Yorker has an article about the famous Mischel experiments.

A researcher offers a four year old child a marshmallow.  He is told he could eat one now, or, if he is willing to wait until the researcher returns from running an errand, he can have two marshmallows.

Some children eat the marshmallow as soon as the researcher leaves.  Others can delay for varying amounts of time.  About 30% are able to patiently sit there for up to 15 minutes, holding out for two marshmallows.  How?


Mischel doesn't see this as a test of willpower, but as a test of the cognitive ability to determine what works, and what doesn't, to delay gratification.

Breaking it down, there are two actions in play here: the first is the action of waiting for the man to return.  The second action is not eating the marshmallow.  It is this second action that Mischel thinks is important in succeeding: "strategic allocation of attention."  The best way to avoid eating the marshmallow is to not to think about it.

Kids who can delay gratification have an intuitive understanding of this. So they cover their eyes, sing a song, etc.  The kids who can't delay do/did the wrong thing: they stare at the marshmallow, trying to beat it in a battle of wills.

To emphasize: it is not that some kids have more willpower, but that they have a better ability to think about something else.  More precisely: they have a better ability to know that distraction is what will work.


Mischel describes ways to improve delay, e.g. watching videos of other kids successfully waiting.

"This is where your parents are important," Mischel says. "Have they established rituals that force you to delay on a daily basis? Do they encourage you to wait? And do they make waiting worthwhile?" According to Mischel, even the most mundane routines of childhood--such as not snacking before dinner, or saving up your allowance, or holding out until Christmas morning--are really sly exercises in cognitive training: we're teaching ourselves how to think so that we can outsmart our desires


It occurred to me that a more obvious way of delaying gratification would be to show the kid what the reward is.  Staring at one marshmallow but having to imagine the future reward of two marshmallows perhaps is too abstract for a 4 year old.  Better to make the reward real:  show him one marshmallow and a pair of marshmallows and tell him if he waits he gets the pair.  Show him what he has to look forward to.


I dug up Mischel's papers, and it turns out Mischel did show them both choices (e.g. two cookies vs. five pretzels; one marshmallow vs. a pair of marshmallows, etc) so the reality of the reward was not an issue.

However, I was in for a surprise:

The results were the opposite of those [we researchers] predicted: attention to the rewards consistently and substantially decreased delay time instead of increasing it.  Preschool children waited an average of more than 11 minutes when no rewards were exposed, but they waited less than 6 minutes when any of the rewards were exposed.

Showing them "what they could look forward to" consistently sabotaged them.  Take that, "anything but" virgins waiting for marriage. 

I wasn't the only one fooled; mothers of preschool kids also erroneously predicted that seeing the future reward would be helpful. Reality, apparently, doesn't encourage you towards the future; it's a reminder that you are hungry now.

(So if parents are trying to teach their kids how to delay gratification, then they should be doing it with the temptations/rewards in plain sight, so as to make the training more difficult.)


If seeing the reward makes it harder to wait, what does help?

What helps is seeing an abstraction of the reward.

This would mean seeing a picture of the reward; or thinking about (in the case of marshmallows) clouds, or (in the case of pretzels) long sticks.  This is subtle, but important; doing any of these is even better than completely distracting oneself from the reward. It's almost pornographic; you're entertaining yourself with abstractions of the thing, which is sufficiently interesting to you that you're not actually thinking about the real thing.  (Virgins, start your laptops?)

The longest delay time (almost 17 minutes) occurred when suggested thoughts were also about [non-reward] objects but with regard to their arousing qualities (for example, children waiting for marshmallows who had been cued to think about the salty, crunchy taste of pretzels.)
But it all comes down to distraction.  In order to get the better, but delayed reward of two marshmallows, instead of just the immediate one, don't look at any of the marshmallows.


But then I had another thought: why doesn't the kid just eat all three marshmallows?

Think about this.  The game here is to maximize the reward; the delay is specifically for that purpose, it serves no benefit in itself.  A child looking at this scenario should be able to see  that the choice is really between eating all of them vs. participating in some bizarre nonsense contrived by an adult that always results in getting only some of the marshmallows.  Even a puma knows not to play this game.

Any kid who holds out for two isn't choosing two over one, but two later over all three now-- and that part isn't even conscious.

This means that the marshmallows are not the only motivators.  There is a value to obedience, that exists in four year olds but not in pumas.  This value may be less than the marshmallows, but it isn't negligible, it isn't even small.  In fact, to some it is worth two marshmallows. and fifteen minutes of time.

Evolutionary psychology, economics, and behavior studies in general often fail to account for what may be an innate, or strongly socialized, motivating variable.  "Rational people will seek to maximize their gain."  Sure.  Now define gain.

In many discussions about behavior and economics, we do not account for obedience and social pressure.  This is a mistake, as it is evident that it is a highly significant, though invisible, determinant.