OBJECTIVE: Early ultrasound scanning estimation of gestational age is known to increase the reported preterm delivery rate (<37 completed weeks) compared with estimation by date of the last normal menstrual period, but it is unclear how this systematic difference arises. STUDY DESIGN: This study was a hospital-based study of 44,623 women who delivered a live-born or stillborn infant between January 1, 1978, and March 31, 1996, and who had both last normal menstrual period-based and early (usually at 16-18 weeks) ultrasound scan-based gestational age estimates. Cross-classification of the 2 estimates by completed weeks was used to examine the direction and magnitude of the differences between them and to compare the resulting classifications of preterm birth. RESULTS: The early ultrasound scan-based gestational age distribution was shifted uniformly to the left (ie, lower gestational age) relative to the last normal menstrual period gestational age distribution; the early ultrasound scan-based preterm delivery rate was 9.1%, which was 19.5% (n = 659 births) higher than the 7.6% rate by last normal menstrual period (P 14 days) ovulation is more frequent than early (