Ecological applications of stable isotope analysis rely on different producers having distinct isotopic ratios to trace energy and nutrient transfer to consumers. Carbon (C) and nitrogen (N) are the usual elements analysed. We tested the hypothesis that producers unable to be separated using C and N would be separated by sulphur (S), by reviewing estuarine and marine food web studies using all three elements (total of 836 pairwise comparisons between producers). S had a wider range of values across all producers than C and N (S: 34.4, C: 23.3, N: 18.7 per thousand ), and a higher mean difference among producers (S: 9.3, C: 6.5, N: 3.3 per thousand ). We varied from 1 to 10 per thousand the distance producers must be apart to be considered separate. For each of these gap distances, S-separated producers tied on C and N in 40% or more of cases. Comparing the three elements individually, S had fewer tied pairs of producers for any gap distance than C or N. However, S also has higher within-producer variability. Statistical tests on simulated data showed that this higher variability caused S to be less effective than C for analysing differences among mean producer values, yet mixing models showed that S had the smallest confidence intervals around mean estimates of source contributions to consumers. We also examined the spatial and temporal scales over which S isotope signatures of the saltmarsh plant Spartina alterniflora varied. Differences between samples taken within tens of metres were smallest, but between samples hundreds of metres apart were as different as samples thousands of kilometres apart. The time between samples being taken did not influence S signatures. Overall, the use of S is recommended because it has a high probability of distinguishing the contribution of different producers to aquatic food webs. When two elements are employed, the combination of S and C separates more producers than any other combination.