Leaching losses of sulfate-sulfur (S) through mole drains at 450 mm depth in two adjacent paddocks were measured for three years. Sulfur was applied in autumn at the rate of 50 and 30 kg S ha-1 in the first and second years respectively, as single superphosphate to one paddock and as elemental sulfur to the other. The concentration in the drainage was mostly between 5 and 10 g S m-3 from the superphosphate-treated paddock, and between 1 and 5 g S m-3 from the elemental S treated paddock. Simultaneous collection and analysis of suction cup and soil samples from 250 mm depth on a number of occasions provided data from which a sulfate adsorption isotherm was constructed. Soil sampling suggested 60% of the superphosphate was immobilized within a week of application. After that net mineralization of sulfate occurred. Sulfate S drainage concentrations were successfully simulated for most of the drainage periods, by using a transfer function derived from chloride and bromide leaching data from the same site and the measured sulfate adsorption isotherm. Oxidation of elemental S, net mineralization, rainfall addition, and herbage uptake of S were incorporated into the model through their effects on the sulfate concentration of the solution entering the soil's transport volume. This meant the convolution integral in the transfer function equation had to be evaluated numerically. Inadequate description of the dynamics of mineralization appeared to be the main weakness of the model, rather than the simplifying assumption that the soil's transport volume could be treated as a well mixed system.