It has been observed that enumerators sometimes fail, for one reason or another, to record a zero answer properly on the schedule, thus leading to the eventual tabulation of the answer as “not given” and introducing error in the data. This article indicates the existence of this error in the data on children ever born which have been collected in many censuses. A study of the covariation of the proportions tabulated as childless and as “not given” in the various age groups has shown the existence of a linear or nearly-linear relationship between the two proportions in many instances. This linearity can be utilized, as explained here, to detect the existence of the error and to adjust the data, under certain conditions.