The source of Mg and the hydrogeologic mechanism by which it is circulated are critical issues for understanding replacement dolomitization because of the extremely large amount of mass exchanged (primarily Mg for Ca) and the volume of parent water fluxed in the process. Mass balance arguments based on dolomite stoichiometry, equilibrium chemistry and basin-scale volumetric data provide the foundation for quantitative analyses of specific dolomite occurrences. When applied to the Western Canada Sedimentary Basin (WCSB), these principles suggest that regional-scale seepage reflux: dolomitization probably accounts for the majority of the dolostones. Moreover, the analyses demonstrate why closed-system dolomitization models, such as burial compaction of shale waters, or cannibalization of Mg-bearing strata, such as older dolostones or shales, cannot account for regional dolomite occurrences. The occurrence of ＇＇hydrothermal dolomite＇＇ (cement and/or replacive phases), which is a localized phenomenon compared to the regional-scale dolostones discussed here, could have cannibalized older reflux-related dolostones in a burial environment, although their origin and age remains to be demonstrated. Upper Devonian dolostones constitute approximately 50% of the volume (1.24 x 10(5) km(3)) of contemporaneous shallow-marine carbonates in the WCSB requiring that approximately 4 x 10(16) kg of Mg be incorporated in the process. Basin-scale mass balance estimates indicate that at least 3 to 9 x 10(16) kg of Mg were contained in the residual evaporitic seawater brines associated with Upper Devonian evaporite deposits, whereas only about 1 x 10(14) kg was associated with connate seawater in all the Paleozoic shales in basin. These results warrant a quantitative analysis of a regionalscale reflux flow system for the Late Devonian as well as abandonment of the shale compaction theory as model for regional-scale dolomitization. A volumetric flow model, i.e., a dimensional analysis of Darcy＇s Law, for the Late Devonian to early Mississippian indicates that regional-scale, shallow-burial (probably less than 500 m) reflux is a viable mechanism for regional-scale dolomitization. The model is consistent with the regional distribution of dolostones and primary evaporites as well as the early diagenetic, burial and thermal histories of the host strata. It also explains the chemical composition of modern connate brines in these rocks.