In a company takeover experiment, groups placed better bids than individuals and substantially reduced the winner’s curse. This improvement was mostly due to peer pressure over the minority opinion and to group learning. Learning took place from interacting and negotiating consensus with others, not simply from observing their bids. When there was disagreement within a group,nwhat prevailed was not the best proposal but the one of the majority. Groups underperformednwith respect to a “truth wins” benchmark although they outperformed individuals deciding in isolation. We draw general lessons about when to employ groups instead of individuals in intellectual tasks.