Constraints on the history of the Martian volatile system can be determined by using observations of the ratios 18O/16O, 17O/16O, 13C/12C, and D/H in atmospheric species and in various components of the SNC meteorites which are thought to be pieces of the Martian surface. These observations are compared with the variations expected from processes which can fractionate the different isotopes-thermal escape to space, nonthermal escape to space, formation of carbonate mineral deposits from atmospheric CO2, and condensation of CO2 or H2O ice from the atmosphere. All of the available observations and models are synthesized for the first time, along with other observations of the Martian surface and atmosphere, in order to provide a coherent view of the evolution of both CO2 and H2O over geologic time. While no scenarios for atmospheric evolution can be absolutely and uniquely ruled out, a strongly favored scenario does exist. In this favored model, fractionation of oxygen by escape to space is diluted by the exchange with a nonatmospheric reservoir; it is unlikely that there is sufficient CO2 adsorbed in the regolith to serve this function, so the polar water-ice deposits must be exchanging with the atmospheric water over geologic time. Exchange of oxygen between atmospheric CO2 and H2O probably provides the observed fractionation between those species. While the observed fractionation of D/H suggests a significant loss of water to space, the timescale on which this loss occurred is uncertain; the results can be reconciled with the oxygen results by invoking time-dependent models (which are consistent with what we know about the Mars climate). Additional observations and numerical modeling efforts which would help eliminate some of the uncertainties in the results are suggested. © 1991.