PURPOSE: Understanding the natural length of human pregnancy is central to clinical care. However, variability in the reference methods to assign gestational age (GA) confound our understanding of pregnancy length. Assignation from ultrasound measurement of fetal crown-rump length (CRL) has superseded that based on last menstrual period (LMP). Our aim was to estimate gestational length based on LMP, ultrasound CRL, and implantation that were known, compared to pregnancy duration assigned by day of ovulation. METHODS: Prospective study in 143 women trying to conceive. In 71 ongoing pregnancies, gestational length was estimated from LMP, CRL at 10-14 weeks, ovulation, and implantation day. For each method of GA assignment, the distribution in observed gestational length was derived and both agreement and correlation between the methods determined. RESULTS: Median ovulation and implantation days were 16 and 27, respectively. The gestational length based on LMP, CRL, implantation, and ovulation was similar: 279, 278, 276.5 and 276.5 days, respectively. The distributions for observed gestational length were widest where GA was assigned from CRL and LMP and narrowest when assigned from implantation and ovulation day. The strongest correlation for gestational length assessment was between ovulation and implantation (r = 0.98) and weakest between CRL and LMP (r = 0.88). CONCLUSIONS: The most accurate method of predicting gestational length is ovulation day, and this agrees closely with implantation day. Prediction of gestational length from CRL and known LMP are both inferior to ovulation and implantation day. This information could have important implications on the routine assignment of gestational age.