The oft-replicated risky-shift phenomenon in decision-making groups was examined under conditions where one or all of the group members expected to serve as a delegate representing and defending their group's decisions in meetings with delegates from other groups. Three separate delegate conditions and a control were employed. In two of the conditions a single delegate was selected--designated by the group in one case and chosen at random in the other. In a third condition, all members expected to serve as delegates. The control condition consisted of the standard group discussion-to-consensus without any delegate manipulation. No mean differences in magnitude of risky shift between conditions were obtained. An examination of the two single-delegate conditions revealed other kinds of differences. Whereas the behavior of the delegate selected randomly appeared to have little effect on the size of the risky shift, this was not the case for the group-designated delegate. Correlational analyses indicated that the latter's influence and communication input and output were related to risky-shift magnitude.