As part of a larger cervical cancer study, we tried to verify the Pap smear histories for 125 black women with cervical cancer. For 105 of the patients, we identified all possible providers for the five-year period before the calendar year of diagnosis. Agreement between the medical records and the patient reports was poor to fair (kappa = 0.34) for whether the patient had a Pap smear in the three-year period before diagnosis. Patients tended to report far more Pap smears than medical records confirmed. Important determinants of agreement were the number of Pap smears reported during the five-year period and the age of the patient. The older the patient and the more Pap smears reported, the larger the discrepancy between the medical record and her self-report. The medical records did not contain enough data for us to complete an investigation of the possible reasons for this disagreement. Our results suggest these implications: (1) clinicians should strongly consider performing Pap smears if they doubt a patient's screening history, and (2) Pap smear registries are required for reliable and efficient evaluations of cervical cancer control programs because neither the patient report nor medical records are adequate.