Concern exists that hypothetical willingness to pay questions overestimate real willingness to pay. In a field experiment, we compare two methods of removing hypothetical bias, a cheap talk approach and a certainty approach, with real purchases. We find evidence of hypothetical bias for unadulterated contingent valuation. Contingent valuation with certainty statements removes the hypothetical bias, but the cheap talk approach has no significant impact. Our findings suggest that willingness to pay can be accurately estimated by adding a simple follow-up question about the certainty of responses and that cheap talk is not a generally effective approach.