OBJECTIVE: To compare estimates based on vaccination cards, parental recall, and medical records of the percentages of children up-to-date on vaccinations for diphtheria, tetanus, and pertussis; polio; and measles, mumps, and rubella. METHOD: The authors analyzed parent interview and medical records data from the Baltimore Immunization Study for 525 2-year-olds born from August 1988 through March 1989 to mothers living in low-income Census tracts of the city of Baltimore. RESULTS: Only one-third of children had vaccination cards; based on medical records, these children had higher up-to-date coverage at 24 months of age than did children without cards. For individual vaccines, only two-thirds of parents could provide information to calculate coverage rates; however, almost all provided enough information to estimate coverage for the primary series. For each vaccine and the series, parental recall estimates were at least 17 percentage points higher than estimates from medical records. For children without vaccination cards whose parents could not provide coverage information, up-to-date rates based on medical records were consistently lower than for children with cards or with parents who provided coverage information. CONCLUSIONS: Population-based vaccine coverage surveys that rely on vaccination cards or parental recall or both may overestimate vaccination coverage.