Correlations between statistically measured wage discrimination and confidential survey reports of discrimination are used to estimate the relative importance of statistical evidence in determining women's self-assessed discrimination levels. Statistical evidence is found to be less important than other, "nonstatistical" evidence. The estimates indicate that the percentage of women who report discrimination would fall from 15.4 to 10.2 percent in Canada, and from 9.9 to 6.3 percent in the United States if all conventionally measured pay discrimination were absent.