The origin of replacement dolomites in the Western Canada Sedimentary Basin (WCSB) remains controversial and problematic. In order to meet the considerable magnesium mass-balance requirements, a wide range of groundwater-flow systems, encompassing all diagenetic settings, has been invoked to explain their formation. One of the previously proposed dolomitizing flow systems is reflux circulation. Reflux circulation occurs in response to differences in fluid density controlled by variations in salinity. The restriction and evaporation of seawater can result in the generation of dense platform-top brines. These brines are potential dolomitizing fluids that can descend into underlying pore networks under the influence of gravity. We used a numerical groundwater flow model to investigate the fluid-flow constraints of reflux dolomitization in carbonate platforms. Specifically, we investigated the sensitivity of the pattern and magnitude of fluid flux to the platform-top (seawater) boundary condition, the concentration of platform-top brines and critical hydraulic parameters (permeability and permeability anisotropy). The pre-existing model of reflux employed in the WCSB has an unrealistic uniform distribution of fluid flux, which we have termed the "constant flux" model, or CFM. In contrast to the CFM, our simulations of reflux circulation incorporate a more realistic platform-top seawater boundary condition, which is open, rather than closed, to flow. Open-topped models generate spatially dependent fluid-flux distributions and are termed here "variable-flux" models, or VFM. Results with the VFM, for an optimal parameter configuration designed to maximize lateral flow, indicate that almost all refluxing brines discharge within 30 km of the brine's source. The spatial distribution of dolomitization times calculated using the VFM demonstrate that the CFM grossly underestimates dolomitization times for most of the platform, but overestimates dolomitization times close to the brine source where fluid flux is greatest. Calculated dolomitization times with our optimal parameter configuration indicate that a region within 15 km of the brine source could be dolomitized in the 16 Ma maximum time constraint available in the WCSB during the Devonian. Regions beyond 15 km from the brine source remain undolomitized. With our VFM it is difficult to reconcile how reflux circulation could pervasively dolomitze Devonian WCSB carbonate platforms with length scales of tens to hundreds of kilometres. The differences in the flow fields and corresponding calculated dolomitization times between the VFM and CFM models clearly demonstrate the inadequacy of the pre-existing CFM. Thus, the proposal that regional-scale reflux dolomitization accounts for most of the Devonian dolostones in the WCSB is unsubstantiated. Furthermore dolomitization times calculated using the CFM cannot be applied to other dolomites of reflux origin in the geological record.