Jack Treynor's jelly-bean experiment

Jack Treynor's jelly-bean experiment is a compact demonstration of a central paradox of collective judgment: a crowd can estimate an unseen quantity with striking accuracy, then become less accurate when given additional information that steers its members toward the same interpretation. The added facts were true; what damaged the estimate was their shared influence on the students’ perspectives.

The First Estimate

Treynor, a finance professor, first asked students in his class to estimate how many jelly beans were in a jar. The jar held 850 beans, and the students’ average guess was 871—within three percent of the actual number. Only one student estimated more accurately than the group as a whole. The result echoed Francis Galton’s earlier cattle-weight experiment: averaging independent estimates can yield a surprisingly close approximation, even when most individual guesses miss the mark.

At this stage, the students had no special direction imposed on their thinking. Their estimates differed, and those differences mattered: the group’s average drew on varied individual judgments rather than repeating a single common interpretation. The experiment established a baseline against which Treynor could test what happened when the students were given a new frame for considering the jar.

A True Clue, a Worse Result

Treynor then asked the students to estimate the number of jelly beans again, but this time urged them to consider two details: there was air space at the top of the jar, and the jar was plastic rather than glass. Both observations were true, and they suggested that the jar might hold more beans than students had initially assumed.

The average estimate shifted from 871 to a figure 15 percent away from the actual count. It was considerably less accurate than a number of the individual guesses. The extra information did not simply add useful knowledge to the first round. Because it pointed in a common direction, it altered many students’ judgments in a similar way, so their errors no longer canceled out as effectively.

How Information Can Reduce Collective Knowledge

The contrast between the two rounds turns on more than whether information is true. Treynor’s account emphasizes how it is presented: instead of merely identifying a plastic jar containing jelly beans, the explanation stressed why the plastic mattered. That emphasis gave the detail force as an interpretive cue. Students were not just told more; they were guided toward a shared conclusion about what the new fact implied.

For a crowd’s average to be useful, its judgments must combine information without allowing one common influence to make everyone err in the same direction. The students’ first estimates included individual differences that could offset one another. In the second round, the shared clue narrowed those differences and weakened that corrective effect. In this sense, more information produced less collective knowledge: it supplied a true fact while diminishing the independence that had made the first average accurate.

The Experiment’s Broader Implication

Treynor’s comparison warns against treating every public disclosure as an uncomplicated improvement in group judgment. Societies and economies depend on sharing information, but commentary that overemphasizes the significance of one detail can make separate observers respond alike. The issue is not that the information is false; it is that a common framing can convert a diverse set of private judgments into a coordinated mistake.

The experiment therefore gives concrete form to a broader account of collective intelligence. A group needs information, but it also needs members whose judgments are not all being shaped in the same way. The first estimate shows how independent errors can offset one another; the second shows how a seemingly helpful explanation can disturb that balance. For Treynor, the jar’s air space and plastic sides were not merely details about a container: once emphasized together, they changed how the crowd saw the problem—and made its answer worse.