We have determined the S, Se, Cu and La contents through a complete stratigraphic section of the Bushveld Complex. The principle aim was to determine which phases controlled these elements. S, Se and Cu show positive correlations, but these elements do not correlate with La. In most cases, the concentration of S, Se and Cu in rocks containing greater than 800 ppm S can be modeled by segregation of a Fe–Ni–Cu sulfide liquid from a fractionating magma. As the magma evolved, Se and Cu were depleted by the continual segregation of sulfide liquid and the S/Se and S/Cu of the rocks increased. The Se/Cu ratio is higher in the more evolved rocks, which suggests that Se has a slightly lower partition coefficient than Cu into sulfide liquid (1,200 versus 1,700). The Lower and lower Critical Zone of the complex contains on average only 99 ppm S. The low S content of these rocks has led some authors to suggest that these rocks do not contain cumulate sulfides, despite the fact that they are moderately enriched in PGE. These samples fall along the same trend as the S-rich samples on the S-versus-Se plot and the S/La and Se/La ratios are greater than the initial magmas suggesting that despite the low S contents cumulate sulfides are present. Three models may be suggested in order to explain the low S content in the Lower and Critical Zone rocks: (a) the sulfides that were present have migrated away from the cumulate pile into the footwall or center of the intrusion; (b) the magma was saturated in sulfides at depth and during transport some sulfides lagged in embayments; (c) the rocks have lost both S and Se at high temperature. The first two models have important implications for exploration.