Accurate estimates of the incidence of abnormal glucose tolerance during pregnancy are virtually nonexistent. Screening select populations of women with risk factors for the condition and the nonrandom, nonpopulation-based nature of most studies have given rise to wide variances in reported incidence. We analyzed data from the states of Mississippi and Washington and from the National Natality and Fetal Mortality Surveys conducted in 1980 in an attempt to provide more accurate population-based estimates of the incidence of gestational diabetes mellitus (GDM). In the national surveys GDM was noted (screening and diagnostic criteria were unavailable) as a complication in 0.38% of all sampled pregnancies; overt (type I and type II) diabetes was noted in 0.78%. Mean maternal age for the GDM group was 28.4 yr; 85% were white (81% controls) and 15% non-white (19% controls). Prepregnancy weights were higher in the GDM group by an average of 20 Ib. However, mean weight gain was less in this group than in controls (23 versus 29 Ib). Perinatal mortality was noted in approximately 2.8% (1.3% in controls) of the offspring in GDM-complicated pregnancies and congenital malformations in 6.4% (7.9% in controls). Methodologic problems were encountered and included lack of screening and diagnostic criteria, underreporting, and underrecording.