Human milk samples (n = 300) were collected during a 3-week period from 10 healthy mothers and from 10 atopic mothers, all with healthy, solely breast-fed infants. The milk samples were analysed by an enzyme-linked immunosorbent assay (ELISA) for the content of bovine beta-lactoglobulin (BLG). In a cross-over design the atopic and non-atopic mothers alternated their intake of milk between homogenized and unhomogenized milk each week. On day 7, in each week, consecutive milk samples were taken before and 4, 8, 12 and 24 hr after a single ingestion of 500 ml of homogenized or unhomogenized milk. Detectable amounts of BLG (0.9-150 micrograms/l, median value 4.2 micrograms/l) were measured in 19/20 of the mothers (95%), in 9 of 10 atopic mothers and in all 10 of 10 non-atopic mothers. No correlation was found between the type of milk preparation (homogenized or unhomogenized) and the presence of BLG or the level of BLG in human milk. A great intra-individual and inter-individual variation of BLG level was found, and no relationship was observed between BLG levels and atopic status of the mothers. The interval between ingestion of 500 ml of milk and the maximal concentration of BLG on milk-free diet varied between 4 and 24 hr, median value 8 hr. The presence of BLG in human milk is a common finding in both atopic and non-atopic mothers.