The glycosylated hemoglobins (HA1) were measured in the blood of normal nonpregnant (n = 50) and pregnant (n = 29) volunteers and pregnant diabetic patients (n = 21). HA1 in normal pregnancy (6.7% +/- 1.03%) did not differ significantly from values in nonpregnancy (7.5% +/- 0.73%) and did not vary with the stage of gestation. HA1 was increased in diabetic pregnancy (8.4+ +/- 2.15%) and was positively correlated with serial fasting blood sugar (FBS) mean values in samples drawn up to 16 weeks (r = 0.57) prior to the measurement of HA1, although the highest correlation (r = 0.73, p less than 0.001) was with FBS levels over the prior 8 weeks. A large standard error of the estimate (26.9 mg/100 ml) showed HA1 to be insensitive as an indicator of prior FBS. There was a large overlap of HA1 from known diabetic patients with the normal range for pregnancy. There was no correlation of third-trimester HA1 and newborn birth weight. It was concluded that (1) normal pregnancy does not alter HA1; (2) HA1 is proportional to prior mean FBS; (3) marked elevations of HA1 (greater than 10%) reliably predict poor diabetic control, but HA1 is not useful for fine control of FBS; (4) HA1 is not useful as a screen for diabetes mellitus; (5) HA1 is not predictive of newborn birth weight.