Three hundred new samples of marine evaporite sulfate, of world-wide distribution, were analyzed for δ34S, and 60 of these also for δ18O in the sulfate ion. Detailed δ34S age curves for Tertiary—Cretaceous, Permian—Pennsylvanian, Devonian, Cambrian and Proterozoic times document large variations in δ34S. A summary curve forδ18O also shows definite variations, some at different times than δ34S, and always smaller. The measured δ34S and δ18O correspond to variations in these isotopes in sulfate of the world ocean surface. The variations of δ18O are controlled by input and output fluxes of sulfur in the ocean, three of which are the same that control δ34S: deposition and erosion of sulfate, and deposition of sulfide. Erosion of sulfide differs in its effect on the S and O systems. δ18O in the sulfate does not seem to be measurably affected by equilibration with either seawater or with subsurface waters after crystallization. In principle, the simultaneous application of both δ34S and δ18O age curves should help reduce the number of assumptions in calculations of the cycles of sulfur and oxygen through geological time, and a new model involving symmetrical fluxes is introduced here to take advantage of the oxygen data. However, all previously published models as well as this one lead to anomalies, such as unreasonable calcium or oxygen depletions in the ocean—atmosphere system. In addition, most models are incapable of reproducing the sharp rises of the δ34S curve in the late Proterozoic, the Devonian and the Triassic which would be the result of unreasonably fast net sulfide deposition. This fast depletion could result from an ocean that has not always been mixed (as previously assumed in all model calculations).