The term ‘cap carbonate’ is commonly used to describe carbonate units associated with glacigenic deposits in Neoproterozoic successions. Attempts to use carbonate units as stratigraphic markers have been counfounded by inconsistent identification of ‘cap carbonates’ and a somewhat broad use of the term. Systematic sedimentological and geochemical analysis of carbonate rocks (mostly dolomite) associated with glacigenic deposits from the Neoproterozoic succession of the Kimberley region, north-western Australia, shows that it is possible to characterize such units by their specific mineralogical, sedimentological, petrographic, geochemical and stratigraphic features. Hence, it is possible to differentiate true ‘cap carbonates’ from other carbonate units that are associated with glacigenic deposits. In the Kimberley successions two broad carbonate types are identified that reflect two stratigraphically distinct depositional realms. Carbonate rocks from the Egan Formation and Boonall Dolomite (the youngest carbonate units in the succession) are characterized by sedimentary components and features that are consistent with deposition on shallow platforms or shelves, analogous to Phanerozoic warm-water carbonate platform deposits. In contrast, dolomite from the Walsh, Landrigan and Moonlight Valley Tillites preserves a suite of sedimentary and geochemical characteristics that are distinctly different from Phanerozoic-like carbonate rocks; they are thin (ca 6 m), laterally persistent units of thinly laminated dolomicrite/dolomicrospar recording δ13C fluctuations from −1‰ to −5‰. These latter features are consistent with a ‘Marinoan-style cap-carbonate’ rock described from other Neoproterozoic successions. The similarity and broad distribution of these rocks in Australia, when considered within the context of genetic models suggesting a global oceanographic–atmospheric event, support their use as a lithostratigraphic marker horizon for the start of the Ediacaran Period at ca 635 Ma.