Background. The purpose of the present study was to validate the registration of obstetric sphincter tears in 2 registers, the Medical Birth Registry of Norway [MBRN] and Patient Administration System [PAS]. Methods. A retrospective cohort study of all obstetric sphincter tears that occurred in our department in 1990–1992 and 2000–2002 was performed. The case records of all patients registered either in MBRN, PAS or the birth logs were compared with the information in the medical records, which constituted the ‘golden standard’. Results. The incidence of obstetric sphincter tears in 1990–1992 was 5.8% (774/13381), 5.6% (745/13381) had a perineal tear of third degree and 0.2% (29/13381) of fourth degree. In 2000–2002, the total incidence was 6.6% (813/12380), 5.9% (731/12380) was a third degree perineal tear and 0.7% (82/12380) fourth degree, respectively. The sensitivity and specificity of the MBRN database to detect obstetric sphincter tears was 85.3 and 99.5% in 1990–1992, and 91.8 and 99.7% in 2000–2002, respectively. The positive and negative predictive values of a MBRN-registered diagnosis of obstetric sphincter tears in 1990–1992 were 91.4 and 99.1%, while the corresponding percentages in 2000–2002 were 95.4 and 99.4%, respectively. The sensitivity and specificity of the PAS database was correspondingly 52.1 and 99.0% in 1990–1992, and 84.6 and 98.5% in 2000–2002. The positive and negative predictive values of a PAS-diagnosis of obstetric sphincter tears were 75.8 and 97.1% in 1990–1992. In 2000–2002, they were 92.7 and 98.9%, respectively. Conclusion. The validity of a diagnosis of obstetric sphincter tears, based on the MBRN, is sufficiently high to justify future large-scale epidemiologic studies based on this database, while the validity of a PAS diagnosis is lower, but improves.