Sulphur Isotope Database:An extensive sulphur isotope database was used in the construction of Main Text Figure 1. These data were compiled from primary references 1-37 , although in some cases age constraints for certain units were obtained from other sources [38][39][40][41] or estimated. For Main Text Figure 1b, we calculated the cumulative average ∆ 33 S value for all samples up to each year of publication. The average values reported here thus include some contribution from sulphate minerals. However, the overwhelming majority of the data are sulphide mineral phases, and given that most sulphate data are characterized by negative ∆ 33 S values their inclusion will largely attenuate the pattern emphasized here, rendering our argument conservative. Furthermore, the overall effect of their inclusion is quite small (Fig. S1a,b). We also examined the effect of filtering data from analyses made via secondary ion mass spectrometry (SIMS) and analyses of macroscopic pyrite textures, which are not likely to be representative of bulk weathering crust, but again note that the basic pattern remains unchanged (Fig. S1a,b).The mean ∆ 33 S values shown in Fig. 1 are presented as unweighted averages. We consider this approach qualitatively justified, in that the database is composed largely of fine-grained siliciclastic sedimentary rocks that are relatively organic carbon and sulphur rich. Such rocks will not only be, on average, the most sulphur-enriched phases of the crust, but they will also likely hold the majority of the weathering sulphur reservoir at Earth's surface. Nonetheless, it is important to consider the possibility that there is a bias related to systematic relationships between the sulphur content of sedimentary rocks and their ∆ 33 S value. For this purpose, we compiled the available sulphur concentration data for the units within the isotope database and performed concentration-weighted statistical analyses.There does not appear to be any systematic relationship between rare sulphur isotope composition and sulphur content within the database (Fig. S2a). In addition, the cumulative unweighted average is nearly indistinguishable from the concentration-weighted average within this subset of the database (Fig. S2a,b). When the evolution of the overall ∆ 33 S value of the database is concentration-weighted, it shows a generally similar pattern through time to that of the overall mean values (Fig. S2b). Indeed, the mean ∆ 33 S value when weighted by concentration is more positive than the unweighted mean for a given population from the database. We stress the caveat that many of the published rare sulphur isotope datasets do not contain sulphur concentration data, so unfortunately this analysis cannot be performed on the entire database. However, the coupled isotope and concentration data represent well over 500 samples that range in sulphur content over three orders of magnitude. These considerations lead us to conclude that the pattern expressed in the overall mean ∆ 33 S values through time is robust and is ...