A central prediction of information economics is that market forces can lead businesses to voluntarily provide information about the quality of their products, yet little voluntary disclosure is observed in the field. In this paper, we demonstrate that the inconsistency between theory and reality is driven by a fundamental failure in consumer inferences when sellers withhold information. Using a series of laboratory experiments, we implement a simple disclosure game in which senders can verifiably report quality to receivers. We find that senders disclose less often than equilibrium would predict. Receivers are not sufficiently skeptical about undisclosed information – they underestimate the extent to which no news is bad news. Senders generally take advantage of receiver mistakes. We find that providing disclosure rates by quality score helps to improve receiver inferences.