This story of a psychology experiment was told as part of a Commencement Address by author Michael Lewis. It has been quoted in a number of places around the web.

They began by grabbing students, as lab rats. Then they broke the students into teams, segregated by sex. Three men, or three women, per team. Then they put these teams of three into a room, and arbitrarily assigned one of the three to act as leader. Then they gave them some complicated moral problem to solve: say what should be done about academic cheating, or how to regulate drinking on campus.

Exactly 30 minutes into the problem-solving the researchers interrupted each group. They entered the room bearing a plate of cookies. Four cookies. The team consisted of three people, but there were these four cookies. Every team member obviously got one cookie, but that left a fourth cookie, just sitting there. It should have been awkward. But it wasn’t. With incredible consistency the person arbitrarily appointed leader of the group grabbed the fourth cookie, and ate it. Not only ate it, but ate it with gusto: lips smacking, mouth open, drool at the corners of their mouths. In the end all that was left of the extra cookie were crumbs on the leader’s shirt.

Is this account of the experiment accurate?

  • 5
    Lewis attributes the experiment to "a pair of researchers in the Cal [University of California, Berkeley] psychology department", "a few years ago" (in 2012). That would be useful information to include for people trying to track it down. Commented Sep 10, 2021 at 20:51
  • 3
    thenewstatistics.com/itns/2019/05/22/… says the researchers were Dacher Keltner and Dan Ward; the year was sometime prior to 1996; the study was actually done at the University of Wisconsin, Madison; the paper was never published; and the original data were lost. Commented Sep 10, 2021 at 20:56
  • 2
    The best published source seems to be Keltner, D., Gruenfeld, D. H., & Anderson, C. (2003). Power, approach, and inhibition. Psychological Review, 110(2), 265–284. doi.org/10.1037/0033-295X.110.2.265 Commented Sep 10, 2021 at 20:57
  • Sounds like an answer. Commented Sep 10, 2021 at 21:41
  • 4
    @TaW: And this is claimed to have happened with "incredible consistency". I guess that's what the New York Times mean when they say that "no one writes with more narrative panache" than Lewis...
    – Schmuddi
    Commented Sep 11, 2021 at 8:40

1 Answer 1


An experiment like the one described happened. How accurate that depiction is of what actually happened is unknown, as no paper was ever published covering the study and now the details are lost to faulty memory. Without the details, we cannot assess the quality of the experiment either.

The best explanation I found is The Cookie-Monster Study: The highly influential memory of a long-lost study. To summarize:

  • The first source to mention the experiment (a review paper) was "Power, approach, and inhibition" by Keltner et al. (2003). It cited an unpublished manuscript by Ward and Keltner (1998).
  • Ward had all the details of the experiment, but Keltner lost contact with him (apparently in 1996).
  • Keltner has said there were 4 cookies sometimes when asked, and 5 at other points.
  • A graphic in Keltner et al. (2003) on how many cookies low power and high power men and women each ate shows that high power men ate fewer cookies than low power men. However, this contradicts what Keltner says happened and it even contradicts the description of what happened in the 2003 paper (apparently not being noticed in review).
  • The website mentions Legitimacy Crisis? Behavioral Approach and Inhibition When Power Differences are Left Unexplained as another experiment along the same lines as the original experiment, but I couldn't find a way to read it (for free) to verify that. (The set up was not exactly the same.)
  • from the article: "Replicating Ward and Keltner (1998), high-power participants (M = 1.03, SD = 0.61) ate more cookies than did low-power participants (M = 0.68, SD = 0.75), F(1, 65) = 5.00, p = .03." in this experiment, there were 3 cookies for 2 people, but people were free not to eat a cookie if they weren't hungry.
    – Avery
    Commented Sep 11, 2021 at 0:08
  • Also in "Legitimacy Crisis?": "When no explanation for the power differences was provided, high-power participants (M = 1.27, SD = 0.44) ate more cookies than low-power participants (M = 0.45, SD = 0.69), F(1, 21) = 12.19, p = .002. When there was an explanation, high-power (M = 0.90, SD = 0.66) and low-power participants (M = 0.77, SD = 0.77) did not differ, F < 1. [...] High-power participants behaved in a disinhibited fashion when no explanation was provided for the power role assignments, and only in this condition did high-power participants eat more cookies than low-power participants." Commented Sep 11, 2021 at 5:55
  • 2
    My 2¢: You never let multiple participants interact with each other in a study like this. Too many confounding variables arising from unpredictable interpersonal interactions between participants. The proper experimental design is one participant, two confederates, and the "randomness" is probably fake too (so that you always assign the genuine participant to the leader role, or to a non-leader role if that's what you want to study). If they didn't explicitly describe any of that, then that's probably (one reason) why their study didn't get published.
    – Kevin
    Commented Sep 11, 2021 at 8:45
  • @Kevin I think if you are trying to show consistency across different group dynamics you would want different group dynamics. Commented Sep 11, 2021 at 10:01
  • 2
    Here is a google drive link to the paywalled paper. drive.google.com/file/d/18SO4NyvfOerIOAVZoDeY3RLKkWCfgsMJ/… Commented Sep 11, 2021 at 14:28

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .