We study the extent to which communication can serve as a collusion device in one-shot first- and second-price sealed-bid auctions. Theoretically, communication between bidders does not change the set of equilibrium outcomes in first-price auctions, but substantially expands the set of equilibrium outcomes in second-price auctions. In an array of laboratory experiments we vary the amount of interactions (communication and/or transfers without commitment) available to bidders. We find that the auctioneer’s revenues decrease significantly when bidders can communicate. When, in addition, bidders can make transfer promises, revenues decline substantially, with 70% of our experimental auctions culminating in the object being sold for approximately the minimal price. Furthermore, the effects of communication and transfers are similar across auction formats.