This article explores why estimates of arts participation in America diverge dramatically. It focuses on two similar surveys - the General Social Survey (GSS) and the Survey of Public Participation in the Arts (SPPA) - that produced very different estimates of attendance at museums, classical music concerts and dance performances. Comparing the design of each survey, this paper examines several possible explanations to account for the divergent estimates, including sample composition and selection bias, question wording and context, and survey design.