To determine the accuracy of school-based childhood immunization records and to describe the effects of their use on estimates of community-wide immunization coverage, we verified the immunizations to 72 months of age for children born in 1986 to residents in Dallas County, TX, and in Minnesota. Verified immunizations were compared with those documented in the school record. Major transcription errors accounted for fewer than 1% of discrepancies between school and provider records. For 99 subjects with 987 verified immunizations in Minnesota, age-appropriate immunization coverage estimated from the school records was within two percent of actual coverage. For 86 subjects with 981 verified immunizations in Dallas County, age-appropriate immunization coverage from the school records underestimated actual coverage by as much as 21%. The primary factor explaining the underestimate in Dallas was incomplete school immunization records for 33 (38%) subjects and 126 (13%) immunizations. Selective recording of immunizations related to the minimum state requirements in Texas contributed to incomplete school records in Dallas County. Verification of the completeness of records selected to estimate immunization coverage is essential if the estimates are used to monitor trends or to make public policy decisions.