Study of commercially produced volcanic CO 2 gas associated with the Colorado Plateau, USA, has revealed substantial new information about the noble gas isotopic composition and elemental abundance pattern of the mantle. Combined with published data from midocean ridge basalts, it is now clear that the convecting mantle has a maximum 20 Ne/ 22 Ne isotopic composition, indistinguishable from that attributed to solar wind-implanted (SWI) neon in meteorites. This is distinct from the higher 20 Ne/ 22 Ne isotopic value expected for solar nebula gases. The non-radiogenic xenon isotopic composition of the well gases shows that 20 per cent of the mantle Xe is 'solar-like' in origin, but cannot resolve the small isotopic difference between the trapped meteorite 'Q'-component and solar Xe. The mantle primordial 20 Ne/ 132 Xe is approximately 1400 and is comparable with the upper end of that observed in meteorites. Previous work using the terrestrial 129 I-129 Xe mass balance demands that almost 99 per cent of the Xe (and therefore other noble gases) has been lost from the accreting solids and that Pu-I closure age models have shown this to have occurred in the first ca 100 Ma of the Earth's history. The highest concentrations of Q-Xe and solar wind-implanted (SWI)-Ne measured in meteorites allow for this loss and these high-abundance samples have a Ne/Xe ratio range compatible with the 'recycled-aircorrected' terrestrial mantle. These observations do not support models in which the terrestrial mantle acquired its volatiles from the primary capture of solar nebula gases and, in turn, strongly suggest that the primary terrestrial atmosphere, before isotopic fractionation, is most probably derived from degassed trapped volatiles in accreting material.By contrast, the non-radiogenic argon, krypton and 80 per cent of the xenon in the convecting mantle have the same isotopic composition and elemental abundance pattern as that found in seawater with a small sedimentary Kr and Xe admix. These mantle heavy noble gases are dominated by recycling of air dissolved in seawater back into the mantle. Numerical simulations suggest that plumes sampling the core-mantle boundary would be enriched in seawater-derived noble gases compared with the convecting mantle, and therefore have substantially lower 40 Ar/ 36 Ar. This is compatible with observation. The subduction process is not a complete barrier to volatile return to the mantle.