Most epidemiological studies of diabetic retinopathy have been based on clinic populations. This produces a bias for the more severe cases and later stages of the disease. To avoid this bias, 1567 Micronesian adults of Nauru (82% of total adult population) were examined. Diabetic retinopathy was classified by both the World Health Organization (WHO) criteria and the Airlie House reference photographs. Results were subject to multiple logistic regression model analysis. Diabetic retinopathy was present in 7% of the general population and in 24% of diabetics. Of the signs, microaneurysms and microhaemorrhages were the most important, being present alone in 19.5%; and in 68.1% when associated with exudates. Proliferative retinopathy was present in 4.5% of cases. Exudates alone were found in two cases (2.2%). Duration of diabetes was the strongest predictor variable and increasing two-hour plasma glucose levels significantly increased the risk of developing retinopathy.