To assess the scale of the possible exposure by the breast-fed infant to potentially harmful substances in breast milk, methodologically robust studies are essential. Many studies in this field, however, do not report details of crucial issues such as recruitment and milk sampling. The aims of the study reported here were to develop robust methods for the study of contaminants in breast milk, and to develop a framework for future research and population monitoring. Three cohorts of women and babies were recruited by midwives from five sites in northern England. Cohort 1 (cross-sectional, n = 322) were asked to provide two milk samples, one at one week following birth and one at a subsequent time point. Cohort 2 (longitudinal, n = 54) were asked to provide five samples at specified time points over the first 12-16 weeks after birth. Cohort 3 (convenience, n = 18), mothers of babies in the Special Care Unit, were asked to donate surplus breast milk. A novel method of analysing fat concentration in small volumes was developed and tested. A randomly selected set of samples from different donors and stages of lactation was screened for organochlorine pesticide residues, polychlorinated biphenyls, dioxins/furans, phthalates and heavy metals. A total of 453 samples were donated. Cohort 3 was the least successful route of recruitment. Cohorts 1 and 2 combined were most representative of the population. Sample collection, transport and storage procedures, and the collection of data on life style and diet, were robust and acceptable to women. Midwifery involvement in recruitment was an essential component. This study offers a framework both for the conduct of future research studies, and for the establishment of regional and national monitoring mechanisms for contaminants in breast milk. Similar work on contaminants in formula as fed to babies is needed to inform risk assessment methods.