We have examined the UG1 and UG2 chromitites, the Bastard Reef, and ultramafic rocks in the footwall of the Merensky Reef that are possibly correlatives to the Pseudoreef, in drill core at Impala platinum mine. The UG2 consists of a 66 cm main seam and, ca 5 m above the UG2, four subsidiary seams with a combined thickness of 30 cm. The UG1 consists of a 112 cm main seam and, in its hanging- and foot wall, 50 subsidiary seams/stringers with a combined thickness of 91 cm. The UG2 main seam contains 2111 ppb Pt, 1214 ppb Pd, 362 ppb Ru, and 98 ppb Ir, with the subsidiary seams containing broadly similar PGE levels. The UG1 main seam contains 628 ppb Pt, 265 ppb Ru, and 45 ppb Ir over 112 cm, but only 13 ppb Pd. The subsidiary seams have variable compositions broadly overlapping with the main seam. Well defined positive correlations between Pt and Pd with the IPGE suggest that in the chromitites the PGE were originally concentrated by a sulfide liquid. In both the UG2 and the UG1 chromitites the highest PGE contents occur at the top and bottom of the seams, analogous to other localities in the Bushveld Complex. Such element distribution patterns resemble those in certain dykes interpreted to result from flow differentiation, suggesting that the seams formed via injection of chromite slurries into their semi- consolidated footwall cumulates, possibly during sagging of the centre of the complex in response to emptying of staging chambers at depth. Down-dip sliding of the semi-consolidated cumulates was also responsible for the phase layering of the chromitites, as represented by the sharp top and bottom contacts of the seams, and the extremely well defined correlation between PGE and chromitites. One of the last processes during cumulate solidification involved removal of some of the sulfur from the chromitites by late magmatic and/or hydrothermal fluids. Pd and Cu have also been mobilized locally (particularly in the UG1 chromitite), but the other PGE appear to have been largely immobile.