The hypothesis that early exposure to cow's milk or lack of breast-feeding predisposes to type 1 diabetes remains controversial. We aimed to determine prospectively the relationship of, first, duration of exclusive breast-feeding and total duration of breast-feeding, and second, introduction of cow's milk protein as infant formula, cow's milk, or dairy products, to the development of islet antibodies in early life. Some 317 children with a first-degree relative with type 1 diabetes were followed prospectively from birth for 29 months (4-73). Mothers kept a home diary and answered infant feeding questionnaires at 6-month intervals. No systematic feeding advice was given. Insulin autoantibodies (normal range <5.5%), anti-GAD antibodies (<5.0 U), and anti-IA2 antibodies (<3.0 U) were measured at 6-month intervals. Cox proportional hazards model of survival analysis detected no significant difference between children who did not develop islet antibodies (225 of 317 [71%]), children with one islet antibody raised once (52 of 317 [16.4%]), children with one antibody raised repeatedly (18 of 317 [5.7%]), or children with two or more antibodies raised (22 of 317 [6.9%]), in terms of duration of exclusive breast-feeding, total duration of breast-feeding, or introduction of cow's milk-based infant formulas, cow's milk, or dairy products (relative risk: 0.91-1.09). Four of the children with two or more islet antibodies developed type 1 diabetes. We conclude that there is no prospective association between duration of breast-feeding or introduction of cow's milk and the development of islet autoimmunity in high-risk children.