The association between breastfeeding and wheezing, lung function and atopy was evaluated in the International Study of Asthma and Allergy in Childhood (ISAAC) Phase II. Cross-sectional studies were performed in 27 centres in 20 countries. Information on disease and exposure factors was collected by parental questionnaires. Data from 54,000 randomly selected school children (aged 8-12 yrs, 31,759 with skin prick testing) and a stratified subsample (n = 4,888) were used for testing the correlation of breastfeeding with bronchial hyperreactivity and lung function. Random effect models for meta-analysis were applied to calculate combined odds ratios (ORs). Any breastfeeding was associated with less wheeze both in affluent (adjusted OR (OR(adj)) 0.87, 95% confidence interval (CI) 0.78-0.97) and nonaffluent countries (OR(adj) 0.80, 95% CI 0.68-0.94). Further analyses revealed that this was true only for nonatopic wheeze in nonaffluent countries (OR(adj) 0.69, 95% CI 0.53-0.90). Breastfeeding was not associated with atopic wheeze and objective measures of allergy in both affluent and nonaffluent countries. In contrast, breastfeeding was associated with higher predicted forced expiratory volume in one second in affluent countries only (mean ratio 1.11, 95% CI 1.02-1.20). Breastfeeding is associated with protection against nonatopic wheeze, which becomes particularly evident in nonaffluent countries. Overall, breastfeeding was not related to any measure of allergy. These findings may explain some of the controversy regarding breastfeeding, since the direction of the association with breastfeeding depends on the predominating wheeze phenotype (e.g. atopic, nonatopic).