In the absence of vaccination card data, Expanded Program on Immunization (EPI) managers sometimes ask mothers for their children's vaccination histories. The magnitude of maternal recall error and its potential impact on public health policy has not been investigated. In this study of 1171 Costa Rican mothers, we compare mothers' recall with vaccination card data for their children younger than 3 years. Analyses of vaccination coverage distributions constructed with recall and vaccination-card data show that recall can be used to estimate population coverage. Although the two data sources are correlated (r = .71), the magnitude of their difference can affect the identification of the vaccination status of an individual child. Maternal recall error was greater than two doses 14% of the time. This error is negatively correlated with the number of doses recorded on the vaccination card (r = -.61) and is weakly correlated with the child's age (r = -.35). Mothers tended to remember accurately the vaccination status of children younger than 6 months, but with older children, the larger the number of doses actually received, the more the mother underestimated the number of doses. No other variables explained recall error. Therefore, reliance on maternal recall could lead to revaccinating children who are already protected, leaving a risk those most vulnerable to vaccine-preventable diseases.