The dry herbage yield increase (response) of subterranean clover (Trifolium subterraneum L.)-based pasture (>85% clover) to applications of different sources of sulfur (S) was compared in 7 field experiments on very sandy soils in the > 650 mm annual average rainfall areas of south-western Australia where S deficiency of clover is common when pastures grow rapidly during spring (August–November). The sources compared were single superphosphate, finely grained and coarsely grained gypsum from deposits in south-western Australia, and elemental S. All sources were broadcast (topdressed) once only onto each plot, 3 weeks after pasture emerged at the start of the first growing season. In each subsequent year, fresh fertiliser-S as single superphosphate was applied 3 weeks after pasture emerged to nil-S plots previously not treated with S since the start of the experiment. This was to determine the residual value of sources applied at the start of the experiment in each subsequent year relative to superphosphate freshly-applied in each subsequent year. In addition, superphosphate was also applied 6, 12 and 16 weeks after emergence of pasture in each year, using nil-S plots not previously treated with S since the start of the experiment. Pasture responses to applied S are usually larger after mid-August, so applying S later may match plant demand increasing the effectiveness of S for pasture production and may also reduce leaching losses of the applied S.At the same site, yield increases to applied S varied greatly, from 0 to 300%, at different harvests in the same or different years. These variations in yield responses to applied S are attributed to the net effect of mineralisation of different amounts of S from soil organic matter, dissolution of S from fertilisers, and different amounts of leaching losses of S from soil by rainfall. Within each year at each site, yield increases were mostly larger in spring (September–November) than in autumn (June–August). In the year of application, single superphosphate was equally or more effective than the other sources. In years when large responses to S occurred, applying single superphosphate later in the year was more effective than applying single superphosphate 3 weeks after pasture emerged (standard practice), so within each year the most recently applied single superphosphate treatment was the most effective S source. All sources generally had negligible residual value, so S needed to be applied each year to ensure S deficiency did not reduce pasture production.