Although some authors explain determinants of breastfeeding that occur in a dose-response manner as evidence of causality, we argue that dose-response relationships are not proof of a biological relationship between the variables. The relationship between maternal smoking and breastfeeding and maternal obesity and breastfeeding are similar: increasing levels of smoking or obesity are associated with shorter duration of breastfeeding. However, maternal infant feeding intention is a strong predictor of breastfeeding duration.In this paper we present data on another variable with a clear dose-response relationship with breastfeeding duration--maternal diet--as a case study to argue that a dose-response relationship does not imply causality.Secondary analysis of cross-sectional survey.Nationally representative sample of 3544 singleton infants in Longitudinal Study of Australian Children, 2004.Maternal intake of fruit and vegetables; breastfeeding duration.Adjusted Odds Ratios (AOR) for breastfeeding at 6 months for daily maternal fruit and vegetable intake (reference group=no fruit/vegetables): 1 serve AOR 1.6 (95%CI 1.0, 2.6); 2 serves AOR 2.3 (1.5, 3.5); and 6 or more serves AOR 4.4 (2.8, 6.8).Although higher maternal intake of fruit/vegetables is associated with longer breastfeeding duration, this is not a biological causal relationship. There are possible biological explanations for altered milk supply in women who smoke or are obese, but not for fruit/vegetable intake. We call for a broader understanding of the social determinants of infant feeding and suggest that all breastfeeding studies measure maternal infant feeding intention as an important determinant.