The platinum-group element (PGE) contents of the upper portions of the Bushveld Complex were investigated with three questions in mind: (a) In natural systems does magnetite concentrate Os, Ir, Ru and Rh (IPGE), as has been observed in experimental systems? (b) Is there a Au-Pd-enriched layer present such as observed in the upper parts of the Skaergaard intrusion? (c) Can changes in metal ratios be used for prospecting for PGE deposits?In the sulfide-poor Main Magnetite Layer of the eastern Bushveld, Ir and Rh are enriched relative to Pt, Pd and Au and this could be because magnetite preferentially concentrated Ir and Rh over Pt, Pd and Au. However, ir. most other magnetite layers no enrichment was observed. This could be because most magnetite layers contain approximately 1% sulfides and the PGE budget is dominated by the sulfides. These sulfides obscured the effects of magnetite collecting IPGE, because sulfides collect all the PGE and the partition coefficients for the PGE into a sulfide liquid are much greater than the partition coefficient for IPGE into magnetite.The weighted average of the platinum-group elements (PGE) and Au over the 2000 m of sampled stratigraphy is Au 2.1 ppb, Pd 1.7 ppb, Pt 1.7 ppb, Rh 0.18 ppb, Ru<0.5, Ir 0.16 ppb, Os<0.5 ppb. Compared to the marginal rocks (presumed initial liquids) of the Bushveld Complex the PGE and Au are severely depleted. Only one sample (a leuccnorite in the first cyclic unit) contained Pt and Pd at economic grade (Pt 2 ppm, Pd 2 ppm). The overall depletion of the PGE in the Upper Zone (despite the presence of 1% to 3% sulfides) could be the result of the PGE having been stripped from the magma by early sulfide liquid which had already settled out of the magma to form the world famous platinum reefs lower in the magma chamber. In addition to the overall depletion of PGE, PGE/S ratios decrease up section indicating that the sulfide fraction is poorer in PGE up section. This is interpreted to be the result of continued depletion of the silicate liquid as sulfides constantly settle out of the silicate liquid. There appears to be little prospect of a Pd-reef type deposit in the Upper Zone.Comparison of the composition of the marginal rocks of the Bushveld with a weighted average for the complete 6 km of the Bushveld cumulates shows that the cumulate pile is much richer in compatible elements (Ir, Rh, Cr) and poorer in incompatible elements (Sm and Hf) than the marginal rocks. Two possible solutions to this are: (a) The magma emplaced into the chamber was a crystal mush and thus more mafic than marginal rocks to the intrusion or (b) fractionated magma has been removed from the cumulate sequence and either irrupted or intruded the country rocks as granites and granophyres. (C) 2004 Elsevier B.V. All rights reserved.