When a mechanistic food web model was parameterized for the Arctic marine ecosystem, it predicted PCB concentrations in zooplankton and fish that were two orders of magnitude lower than measured. PCB concentrations measured in zooplankton and fish were within the laboratory's accredited quality assured criteria, and were comparable to other Arctic regions. Although on a different scale, the predicted and measured PCB concentrations were highly correlated. As sensitivity analyses indicated water concentrations as the most sensitive parameter for the model output, dissolved water concentrations were predicted using measured zooplankton and air PCB concentrations, and empirical and mechanistic models. The food web model and the empirical relationship between bioaccumulation factor and octanol-water partitioning coefficient predicted mean dissolved water concentrations of 28 and 29 pg/L P PCB 6 (PCB-28, -52, -105, -118, -138, -153), respectively. Mean dissolved water concentration predicted from measured air concentrations in 1996 was 7.6 pg/L P PCB 5 (PCB-28 was not analysed). Mean dissolved water concentration measured in Barents Sea water sampled simultaneously as the biota in 1999 was 0.3 pg/L P PCB 6 . The dissolved water concentrations predicted from zooplankton PCB concentrations were comparable to water concentrations measured in 1996, whereas the dissolved water concentrations measured in 1999 were comparable to measurements from 2001. If the present high empirically derived bioaccumulation factors (log BAF 7.3-9.0) were realistic, this suggests that bioaccumulation in Arctic zooplankton is more efficient that previously assumed. The present study illustrates and discusses some of the difficulties encountered when different approaches to study environmental distribution of contaminants are compared. D