The relationship between hyperglycemia, measured by glycosylated hemoglobin at the initial examination, and the four-year incidence and progression of diabetic retinopathy was examined in a population-based study in Wisconsin. Younger- (n = 891) and older-onset (n = 987) persons participating in baseline and follow-up examinations were included. Glycosylated hemoglobin was measured by microcolumn. Retinopathy was determined from stereoscopic fundus photographs. In the younger-onset group, comparing the highest with the lowest quartile of glycosylated hemoglobin, the relative risk for developing any diabetic retinopathy was 1.9; for proliferative retinopathy, 21.8; and for progression, 4.0. Among older-onset persons taking insulin, the corresponding relative risks were 1.9, 4.0, and 2.1. Among older-onset persons not taking insulin, relative risks were 4.0 for any retinopathy and 6.2 for progression. A positive relationship between incidence and progression of retinopathy and glycosylated hemoglobin remained after controlling for duration of diabetes, age, sex, and baseline retinopathy. These data suggest a strong and consistent relationship between hyperglycemia and incidence and progression of retinopathy.