Many ethnic groups at high risk of non-insulin-dependent diabetes mellitus are hyperinsulinaemic by early adult life. This study assessed whether such hyperinsulinaemia is present at birth. Cross sectional comparisons of maternal biochemistry, umbilical cord biochemistry and neonatal anthropometry were made between one 'low risk' and three 'high risk' ethnic groups, without diabetes in pregnancy in Auckland, New Zealand. The study comprised 123 European, Polynesian (Maori and Pacific Islands) and Indian normal pregnancies. Indian mothers were the smallest, with the highest insulin and non-esterified fatty acid concentrations. Polynesian mothers were the most obese with a higher fructosamine concentration. From these pregnancies, Indian neonates were smaller, slimmer, with the highest cord triglyceride (0.6 mmol/l vs 0.4 mmol/l, p < 0.01), and lowest cord insulin concentrations (7.1 mU/l vs 8.6 mU/l (European), 9.2 mU/l (Polynesian), p < 0.05). Polynesian babies had a high cord insulin: C-peptide ratio (52.5 mU/nmol vs 44.4 mU/nmol (European), 44.1 mU/nmol (Indian), p = 0.05). Although reduced intrauterine growth may contribute to the excess of diabetes and heart disease in Indians, it cannot explain the excess of diabetes in Polynesians. Exposure to minor relative maternal hyperglycaemia in the mother and abnormal neonatal insulin handling (as demonstrated by the higher insulin: C-peptide ratio) may be of long-term significance in Polynesians.