Rates of reported eczema during early childhood were studied in 3 national cohorts of children born in 1946, 1958, and 1970. Overall rates rose from 5.1% in children born in 1946, to 7.3% in those born in 1958, to 12.2% in the 1970 cohort. In the 1958 and 1970 cohorts there was a positive association between eczema and breastfeeding. This relationship remained significant after allowing for parental history of allergy and socioeconomic status. Social classes I and II children born in 1946 were less likely to be reported as having eczema, compared with children from lower social classes, whereas children born into higher social classes in 1958 and 1970 had higher rates. These findings may reflect secular changes in the diagnosis of eczema or may represent a real increase in the disorder. The positive association with breastfeeding in the more recent cohorts suggests a new environmental agent may be crossing in breast-milk. The agent(s) may well be in other infant foods, since the rate of reported eczema in non-breastfed children rose from 5.7% in the 1946 and 1958 cohorts to 11.1% of children born in 1970.