This article uses laboratory data from a series of first-price (FP) and second-price (SP) sealed bid auctions in which the number of bidders is unknown to test for possible deviations of individual behavior from theory and study the source of heterogeneity in bidding. In SP auctions we find a substantial amount of coincidence with theory. We observe systematic deviations from risk neutral bidding in FP auctions and show theoretically that these deviations are consistent with risk averse preferences. We find essentially no heterogeneity in bidding in SP auctions where risk preferences and the number of bidders do not affect the optimal bid, while in the FP auctions heterogeneity in bidding persists with experience. We find that heterogeneity in bidding in FP auctions is consistent with heterogeneity in risk preferences, the attempt to count the number of bidders in the auction, and bidder specific noise. (JEL D44, C91)