A PCB input−output balance study was performed on five cows in which they were fed a natural background contaminated diet and kept under typical U.K. winter husbandry conditions for 15 weeks. Initially an intensive study was performed over five days to investigate short-term variability in PCB fluxes. This was followed by a long-term (14 week) study to again assess variability and the appropriateness of a steady-state assumption for modeling purposes. Measurements of feed intake, milk, and milk fat output were made regularly, and feces output was estimated on two occasions using a recognized dosing and recovery technique. It was found that PCB concentrations in milk dropped by an average of 25% through the lactation, while PCB intakes rose as the cows voluntarily ate more silage later in the lactation. No difference was found in PCB concentrations from the morning and afternoon milkings. True steady-state conditions were not attained at any time, and therefore models which assume steady-state conditions to predict milk and meat concentrations should be used with care. It was evident that changes in the body fat depot throughout the lactation cycle play an important role in the storage and release of PCBs. Input−output balances (i.e. milk and fecal output fluxes divided by the dietary intake flux) of over 100% were found for conservative, or unmetabolized, PCB congeners (e.g. PCB 138). However, this dropped to approximately 20% for readily metabolized congeners (e.g. PCB 33). Variability in congener-specific carry-over rates (CORs, defined as the output flux in milk divided by the dietary intake flux) was small between cows sampled at the same time. Elevated levels of certain PCB congeners were found in the liver of a cow which was slaughtered at the end of the study.