Passive sampling devices accumulate chemicals continuously from water and can provide time weighted average (TWA) concentrations of pollutants over the exposure period. Hence, they offer a number of advantages over other conventional monitoring techniques such as spot or grab sampling. The diffusive gradient in thin film (DGT) and the Chemcatcher passive samplers can be used to provide TWA concentrations of labile metals, but the approaches to their calibration differ. DGT uses diffusion coefficients of metals in the hydrogel layer, whereas Chemcatcher uses metal specific uptake rates, with both sets of values obtained under controlled laboratory conditions with constant aqueous metal concentrations. However, little is known of how such samplers respond to fluctuating concentrations. We evaluated the responsiveness of these two passive sampling devices to rapidly changing concentrations of Cd, Cu, Ni, Pb and Zn in natural freshwater, over a relatively short deployment time. Maximum metal concentrations in water were varied between 70 and 140 microg L(-1). Experiments were carried out in a tank with a rotating carousel system and filled with Meuse river water, allowing a degree of control over experimental conditions while using natural river water. Fluctuating concentrations were obtained by stepwise addition of standard solutions of the metals. The reliability and accuracy of the TWA concentrations measured by the samplers were assessed by comparison with concentrations of the metals in spot samples of water taken regularly over the deployment period. The spot samples of water were either unfiltered (total), filtered (0.45 microm) or ultrafiltered (5 kDa). Predictive speciation modelling using the visual MINTEQ programme was also undertaken. There was reasonable agreement between the TWA concentrations of Cd and Ni obtained with Chemcatcher and DGT and the total Cd and Ni concentrations measured in repeated unfiltered spot samples. For elements (i.e. Cu, Pb, Zn) that associate to a significant degree with suspended solids, colloids or dissolved organic carbon, or form complexes with large organic ligands, optimum agreement was with the filtered or ultrafiltered fractions and with the predicted inorganic and inorganic-fulvic acid associated fractions. While Chemcatcher-based TWA concentration ranges for Cu and Zn were in best agreement with the total filtered fraction, there was lack of agreement for Pb. The combined use of DGT devices with open pore (OP) and restricted pore (RP) gels allowed the labile fraction of metal associated with large organic ligands or DOC to be differentiated and quantified, since this is available to DGT OP but unable to diffuse into the DGT RP. This evaluation of the two sampling devices clearly demonstrated their ability to react reliably to transient peaks in concentration of metal pollutants in water and indicated where future efforts are needed to improve calibration data. Such samplers may prove valuable in responding to the monitoring requirements of the European Union's Water Framework Directive.