Cenozoic limestones from Hawaii and Enewetak were studied to characterise diagenesis in deep sea water. Hawaii samples were from subsea outcrops of drowned Pleistocene reefs 150-1,505 m deep (maximum age 550-600 ka). Most samples had early fibrous aragonite and high-magnesium calcite cements precipitated in shallow sea water. Partial dissolution of aragonite (including coral) and highmagnesium calcite were significant at 412 m and increased to 1,505 m. Crusts of 'stubby' sparry calcite cement (2-8 mol.% MgCO 3 ; 'lower Mg calcite') precipitated on early aragonite and high-magnesium calcite cements at 473-1,358 m. Dissolution of aragonite and high-magnesium calcite was incomplete. Aragonite and high-magnesium calcite were not neomorphosed to low-magnesium calcite ( <5 mol.% MgCO 3 ). Enewetak well samples came from 3 to 1,400 m (Holocene to Upper Eocene). Lower Miocene to Upper Eocene carbonates at 380-1,380 m near the atoll margin showed pervasive dissolution of aragonite and conversion of high-magnesium calcite fossils to low-magnesium calcite. Their lower-Mg calcite cements (380-820 m; mainly radiaxial) were associated with aragonite dissolution. The lower-Mg calcite cements and bulk limestones below 500 m had geochemistry indicating precipitation or stabilisation in sea water at 10-27°C. Data indicate Enewetak dolomitisation (1,250-1,320 m) in cold sea water during burial >1,000 m. Coralline algae showed little petrographic alteration, but Mg decreased downward from 15 to 1.5 mol.% MgCO 3 . In both areas, aragonite dissolution, alteration of high-magnesium calcite, and precipitation of lower-Mg calcite cements occurred in deep sea water (>300 m) undersaturated for aragonite, but supersaturated for low-magnesium calcite. Original high-magnesium calcite was partially dissolved in Hawaii samples, but converted to low-magnesium calcite in deep Enewetak cores, possibly due to gradual deepening at Enewetak. Dolomitisation and low-magnesium calcite dissolution occurred below the calcite saturation depth (approximately 1,000 m) in Enewetak, but not deep Hawaii samples, possibly because dolomitisation is slower. Temporal variations in carbonate saturation, especially related to pCO 2 , are interpreted as the main control on mineralogy during marine diagenesis now and in many ancient oceans.