the Bavelas experiment

"During the 1950's and 60's, a series of intriguing experiments were conducted on the nature and effect of feedback on human activity. In one representative study, professor Alex Bavelas simultaneously exposed two participants to a series of pictures of either healthy or sick cells (Watzlawick, 1976). Neither person in the study could observe the other while the experiment was under way and each was given the assignment to learn to distinguish between the two types of cells through trial and error. Small lights marked "right" and "wrong" provided feedback to the participants about their respective choices.
There was just one "wrinkle" in the experiment of which both participants were unaware. Only one of them received accurate feedback about their guesses. When the light in this person’s cubicle indicated they had made the "right" choice, they had indeed guessed correctly. On the other hand, feedback for the second participant was not based on their own, but on the guesses made by the first participant! No matter their choices, this person was told they were "right" if the other person had guessed correctly and "wrong" if the other had been incorrect. Data collected without their knowledge showed, at the conclusion of the experiment, that the first participant had learned to distinguish healthy from sick cells with an 80% rate of accuracy. The second continued to guess at no better than a chance rate.
These were not the only results. The two types of feedback also had a distinct and interesting impact on the theories each participant developed during the study for differentiating between " healthy" and "sick" cells. The participant who received accurate, reliable feedback ended the experiment with a very simple, concrete, and parsimonious explanation. In contrast, the second participant, developed a complicated, subtle, and elaborate theory. This person, it is recalled, had no way of knowing the feedback they received was not contingent on their own responses. Sometimes, as luck would have it, their responses happened to coincide with the correct answer. However, given the inconsistent and unreliable feedback, this participant was prevented from learning anything about their own actions and choices.
Even these results are not all that surprising. Something more troubling occurred when the two participants shared their respective theories with each other. Contrary to what one might hope and expect, the first participant was impressed with the complicated, mysterious, and unreliable theoretical formulations of their co-participant. The second, on the other hand, dismissed the statistically accurate theory of the first as "naive and simplistic." In later retests during which both participants received accurate feedback about their own guesses, the second continued to guess at little better than a chance rate. The performance of the first, however, who was now attempting to put some of the "brilliant" insights of their co-participant into practice, significantly worsened."