This study assessed the accuracy of self-reported Pap smear utilisation over four different time frames, examining the magnitude of errors in self-report and sociodemographic predictors of accuracy. Self-report data on women's cervical screening was collected by interview in a random household survey (Hunter Region, NSW, Australia), with pathology laboratory data collected by a search of records within laboratories. The magnitude of error in self-report was assessed by comparing it against longer intervals in pathology laboratory data. Sociodemographic predictors of accuracy were explored using chi square analyses. Low values for specificity and positive predictive value across all four time frames indicate a considerable degree of inaccuracy in women's reporting of those instances where, in truth, screening has not occurred. Of women reporting a smear within the last three years, only 61.2% were verified within pathology laboratory records. Allowing women some "leeway" in their reporting, comparing self-report to longer intervals of pathology laboratory data, did not greatly improve the accuracy of reporting, confirming that the magnitude of inaccuracy involved is of real clinical significance. Demographic variables were not related to the accuracy of self-report and, while a woman's certainty of her response was predictive, this had little impact on the measures of agreement. Self-report of Pap smear histories consistently results in over-reporting of screening. Other means of assessing the prevalence of screening may be preferable to self-report. Where self-report data is collected, techniques to improve accuracy should be employed, and care should be taken in comparing screening rates obtained by different methods.