Although research on reproductive behavior depends heavily on information from surveys, abortions are characteristically underreported in such data. Estimates of the level of reporting are made for each of the recent major surveys of U.S. women: the 1976, 1982, and 1988 cycles of the National Survey of Family Growth, the 1976 and 1979 National Surveys of Young Women, and the National Longitudinal Surveys of Work Experience of Youth. The estimates are based on comparisons with external counts of abortions taking place. We examine variation by characteristics of women, trends over time, and the possible effects of length of recall and of the way in which questions about abortion are asked. Abortion reporting is found to be highly deficient in all the surveys, although the level varies widely. Whites are more likely to report their abortions than nonwhites. Special, confidential questioning procedures hold promise for improving the results.