Petrologists all agree that the “carbonated mantle”, i.e. peridotite with accessory carbonate, is necessary to generate CO2-bearing melts. Carbonated peridotite is also a useful theoretical concept for geochemists seeking to explain trace element enrichment of the lithospheric mantle. Melting of peridotite with addition of carbonate has been the subject of hundreds of experimental studies. Yet mantle samples from below cratons do not contain carbonate. The talk tries to reconcile the theoretical view of the carbonated mantle with the empirical observations on cratonic mantle xenoliths. We show that the natural carbonated mantle are peridotites with elevated modes of clinopyroxene, garnet and olivine, and with thin rims of calcic silicate minerals. The peridotites were affected by “carbonation freezing”, i.e. immobilization of ephemeral carbonate-rich metasomatic agent via reactions with pyroxenes. The empirically observed cratonic carbonated mantle does not store carbon.