Rationale: The analysis of the three sulfur stable isotope ratios (S-33/S-32, S-34/S-32, S-36/S-32) is routinely performed by gas-source isotope ratio mass spectrometry (IRMS) on the SF6 gaseous molecule, collecting SF5+ ions at m/z similar to 127, 128, 129 and 131. High precision and accuracy are commonly achieved owing to a lack of correction because fluorine has only one isotope and the inert nature of the SF6 molecule. The analysis of the S-36/S-32 ratio is, however, complicated by the low abundance of S-36 (similar to 0.015%) and the possible occurrence of trace amounts of fluorocarbon compounds leading to (C3F5+)-C-12 ions at m/z similar to 131, i.e. where (SF5+)-S-36 ions are collected. Methods: We used gas-source high-resolution IRMS to better characterize the nature of possible interferences, and we tested novel types of filaments in order to investigate their influence on possible interferences. Results: We confirm that the (C3F5+)-C-12 ion represents the main isobaric interference at m/z similar to 131. We also demonstrate that tungsten fluoride adducts are formed from the reaction of fluorine ions derived during fragmentation of the SF6 molecule with the hot tungsten filament. These reactions lead to the formation of e.g. WF5+, WF4+, WF3+ ions, including doubly charged ions. WF4++, in particular, leads to isobaric interference on m/z similar to 128, 129 and 131 from (WF4++)-W-180, (WF4++)-W-182 and (186) WF4++ ions, respectively. Because W-180 (0.12%) is at low abundance, its influence on delta S-33 measurements would remain negligible, but W-182 (26.5%) and W-186 (28.4%) lead to scale contraction for both delta S-34 and delta S-36. Conclusions: Rather than correcting for these interferences, or working at high mass resolution, we suggest avoiding W isobaric interferences by using other types of filaments, with initial reports on both pure Re filaments and Y2O3-coated W filaments.