Porosity and permeability data from five carbonate platform successions of different settings, ages, and burial depths are examined to identify overall similarities and differences in the reservoir quality of interlayered limestones and dolostones. Each succession consists mainly of limestone and dolostone, with subordinate proportions of intermediate, partly dolomitized compositions. In the three deeply buried platforms, the key features are that limestones have much lower average porosity than associated dolostones, and that limestones and dolostones show little difference in average permeability-for-given-porosity. In contrast, the shallowly buried platforms show little difference in average porosity between limestones and dolostones and also display higher average permeability-for-given-porosity in dolostones than limestones. These data suggest the following general guidelines for depositional and diagenetic controls on reservoir architecture in carbonates consisting of interlayered limestone and dolostone. (1) Reservoir compartmentalization by the formation of tight limestone barriers is largely a burial diagenetic process involving calcite cementation locally produced by chemical compaction. (2) Both the pattern of early dolomitization and the distribution of clay minerals (because they influence the localization of chemical compaction) are key factors that determine the distribution of tight limestone barriers separating flow units. Thus, the pattern of eventual burial compartmentalization follows a template that is hardwired into the stratigraphic architecture by depositional mineralogy and early diagenesis. (3) After burial (2-3 km; 1.2-1.8 mi), dolostones should not be expected to have higher permeability-for-given-porosity than associated limestones. These rules assume dolomitization to occur early relative to chemical compaction, which will commonly be the case because of combined hydrologic and mass-balance constraints. Dolomitization and concentration of clay appear linked to cycle and sequence architecture in the present examples and thus may have a useful degree of predictability, at least in terms of statistical parameters, such as net/gross and probability of flow-unit thickness distribution.