The allergy preventive effect of extensively (N) and partially (PH) hydrolysed cows' milk formulas compared with a regular formula (RM) was assessed in 155 infants with a family history of allergy. No cows' milk was given during the first nine months of life and no egg and fish up to 12 months of age. Breast feeding mothers avoided the same foods. At weaning the infants were randomised to one of the formula groups. The cumulative incidence of atopic symptoms at 18 months was 51, 64, and 84% in the N, PH, and RM groups, respectively. From 6 to 18 months there were significantly less cumulative atopic symptoms in the N group compared with the RM group, and significantly less than the PH group up to 6 (N = 25%; PH = 46%) and 9 months (N = 34%, PH = 58%). At 9 months significantly fewer infants in the N group (10%) than in the PH group (33%) had a positive skin prick test to eggs. The findings support an allergy preventive effect of an extensively hydrolysed formula, but not of a partially hydrolysed formula, during the first 18 months of life of high risk infants.