The Cardium Formation is a shallow marine clastic unit of upper Middle Turonian to Lower Coniacian age that was deposited in a ramp setting on the western side of the Western Canada foreland basin. The regional subsurface and outcrop stratigraphy of the Cardium is well-known in northern and central Alberta, but is poorly-known in southern Alberta. A correlation grid of >1200 well logs allowed allomember-bounding erosion surfaces E1 to E7 of the Cardium Formation to be traced across southern Alberta and into northern Montana. Twenty-five outcrop sections in Alberta and Montana were correlated to this stratigraphic framework. Two new bounding unconformities, designated E5.2 and E5.5 were recognized. Both surfaces onlap onto surface E5 towards the NW, showing that the original E5 surface is a composite of three unconformities. The lower Raven River allomember below E5 contains the most basinally-extensive sandstone and is composed of eastward-inclined clinothems indicative of long-term regression. The lower Raven River contains ammonites representative of four Upper Turonian Zones. The overlying upper Raven River and lower and upper Dismal Rat allomembers become progressively muddier upward and form a retrogradational stacking pattern, indicative of long-term transgression. Ammonites and bivalves locate the Turonian-Coniacian boundary just above the E5.5 surface. The E6.5 surface, originally recognized in the Willesden Green area, has been traced over the entire study area, confirming its importance as a major bounding discontinuity. The E7 surface has major erosional relief (>35 m in places) across southern Alberta. Overall, Cardium allomembers between E1 and E5 have a progradational stacking pattern. Allomembers between E5 and E6 are retrogradational whereas those between E6 and E7 are again progradational. Correlation of the Cardium in well logs from Alberta to outcrop of coeval rocks of the Marias River Formation at Ferdig and Deer Creek in Montana, supplemented by biostratigraphic information, suggests that the upper 8–16 m of the Ferdig Member is of Early Coniacian age. The new correlations imply that the top of the Carlile Formation, defined in subsurface, is not the Turonian-Coniacian boundary, as indicated by previous authors, but instead is equivalent to the E7 surface, which defines the top of the Cardium alloformation, and is of Early Coniacian age.