This study was conducted to examine whether the amount of unshared informat
ion (i.e., information that only one group member or another possesses prio
r to discussion) exchanged within groups is related to group-judgment accur
acy when the correct response is not apparent to the members prior to discu
ssion. Thirty-nine 3-person groups were asked to make a series of 36 judgme
nts regarding the probability that hypothetical high school dropouts would
return to school. These judgments were based on a set of information, part
of which was given to all group members prior to discussion (shared informa
tion) and part of which was divided among them (unshared information). More
over, this information was distributed to the members in such a way that th
eir individual prediscussion preferences would tend to be either inaccurate
(hidden profiles) or accurate (manifest profiles), relative to the optimal
group judgment based on all of the information that was given to the group
as a whole (i.e., both shared and unshared information). Results indicated
that there was no relation between the amount of unshared information disc
ussed and group accuracy on hidden-profile cases. Instead, the results sugg
ested that group accuracy was determined by how accurate members were prior
to discussion and that the vital role of group discussion was not to excha
nge information but to aggregate member judgments into a consensual group j
udgment.