This study was done to determine the accuracy of fetal ultrasound (US) predicting the estimated date of delivery (EDD) in a primarily Afro-Caribbean population in Antigua and Barbuda. A total of 206 women had retrospective review of fetal ultrasound tests done between July 1994 and January 1996. The EDD based on last menstrual period (LMP) was calculated for 104 women with dates and compared with EDD calculated from ultrasound test. These were then compared with actual date of birth from maternity records. For women with unreliable menstrual data (102 women), ultrasound EDD was compared with actual date of birth. Ultrasound EDD demonstrated a trend toward being more accurate than menstrual history EDD, being 10.0 ± 9.4 (mean ± standard deviation) days off versus 13.3 ± 15 days, p = 0.057, (n = 104), CI: -0.1, 6.73 days. Ultrasound EDD was within ten days of delivery 60% of the time versus 57% for menstrual history EDD. The average error in estimating EDD was + 1.7 days for ultrasound and + 3.7 days for menstrual history. In women without menstrual history data (n = 102), the ultrasound EDD was just as accurate in both the second trimester, 11.8 ± 9.5 versus 11.4 ± 10.7 days off (not statistically significant) and the third trimester, 10.0 ± 8.0 versus 8.1 ± 7.0 days off (not statistically significant). Fetal ultrasound is marginally better at predicting the date of birth compared with menstrual history but the difference does not justify routine use for that purpose. When menstrual history is unknown, ultrasound EDD is just as accurate as when menstrual history is known, making it a very useful test.