In Bayesian-Nash equilibrium formulations of auctions the bidders are assumed to know all the relevant structural elements of the game they are in. We investigate the empirical implications for bidding behavior of weakening this assumption in the context of sealed bid first price auctions. We study a model in which the bidders are assumed to estimate the unknown distributions that affect their payoff function by processing bids on same or similar items in previous auctions. We present Monte Carlo experimental results on the distribution of bidding strategies when a bidder substitutes the empirical analog of the unknown distribution of bids by the rivals. We also explore the performance of Bayesian- Nash equilibrium based empirical approach to estimate the unknown distribution of bidder valuations when the proposed behavioral model rather than a Bayesian-Nash equilibrium process generates the observed bid data. 20 Pages