Until recently, most surveys of insurance coverage have classified people as uninsured if they have not been assigned some coverage in response to one of a series of questions about specific types of insurance. This "residual" approach to measuring uninsurance rates has not required respondents to either verify their insurance status or confirm that they are uninsured. Using the 1997 National Survey of America's Families, this paper examines the impact of a question confirming whether individuals for whom no insurance coverage is reported are, in fact, uninsured. The results of our analysis suggest that a confirmation question as part of a telephone-based survey works to lower estimates of the uninsured.