The protective effect of breast-feeding in allergy prophylaxis remains controversial despite many epidemiologic studies. One reason for confusion could be heterogeneity of breast milk with regard to its protective capacity. To clarify this issue, we studied 57 mother-infant pairs where breast-feeding was the sole source of infant nutrition. Family history and infant symptom scores suggestive of allergic disease were noted. The presence and quantity in breast milk of a prominent food allergen (beta-lactoglobulin), total IgA, and IgA antibodies to whole cow's milk and casein were measured serially. Eleven infants had symptom scores highly suggestive of allergic disease. The breast milk from mothers of these 11 infants was found to have lower total IgA (p less than 0.01) and IgA antibodies to both whole cow's milk and casein (p less than 0.001 and p less than 0.005, respectively) than milk from mothers whose infants had few or no symptoms. Beta-lactoglobulin was detectable in 45% of breast-milk specimens at concentrations up to 6.4 ng/ml and persisted up to 3 days after maternal dietary milk exclusion but was unrelated to antibody levels or symptom scores. Inadequate quantities of maternal IgA antibodies to food allergens appear to play a permissive role in development of infantile allergic disease in breast-fed infants and are more closely associated with symptoms than parental atopic history.