This paper is based on 465 new analyses of Ni, Cu, S and PGE from the 19 chromitite horizons between the LG-1 and UMG-2 from 6 sectors around the Bushveld Complex, along with microprobe analyses of representative samples of 41 chromites. Two trends in chromite composition, A and B, are distinguished on a plot of cation% Mg/(Mg + Fe2+) versus Cr/(Cr+Al). Trend A, that has a negative slope, is close to that predicted as the result of the reciprocal exchange substitution of Cr and Fe2+ for Mg and Al between spinel and liquid affecting the Mg-Fe2+ spinel-liquid Kd2. Trend B, that has a positive slope and is defined primarily by the LG-5 to MG-2 chromitites, is the result of the progressive increase in the activity of Al2O3 as a result of the fractional crystallization of orthopyroxene. Overall, the average PGE concentrations in massive chromitite increase upward. The LG-1 to LG-4 chromities have low (Pt+Pd)/(Rh+Ru+Ir+Os) ratios (0·1 to 0·3), above which there is an abrupt jump to higher ratios in the LG-5 (0·9 to 10) and all overlying chromitites (also documented by Scoon and Teigler). The Pt/Ru and Pd/Ru ratios are very variable, but the Ru/Ir, Ru/Rh and Ru/Os ratios of all chromitites are relatively constant, indicating that Pt and Pd respond to different concentration mechanisms to the other PGE. Rh, Ru, Ir and Os were likely concentrated by chromite itself, probably as grains of laurite and alloys incorporated in growing chromite crystals, but the bulk of the Pt, Pd along with lesser proportions of the other PGE were concentrated by sulphide liquid. Most chromitites now have very low contents of S, but mineragraphic and chemical data support the suggestion of Naldrett and Lehmann that vacancies in chromite forming above 900°C were filled by Fe2+ derived from the destruction of interstitial sulphide liquid. Data on En composition through the Bushveld CriticalZone, indicate that the LG-1 to LG-4 chromitites formed at a stage when influxes of magma into the chamber were rapid and primitive, and overrode the effect of fractional crystallization, whereas above this, fractionation mostly overrode influxes of new magma. Irvine's model of mixing of resident magma with influxes of more primitive magma is invoked as the origin of the chromitite horizons. It is shown, using the equation for sulphur solubility and the programme MELTS, that influxes and mixing of fresh primitive magma from depth with that in the chamber (i.e. as envisaged for the LG-1 to LG-4) would not have caused sulphide immiscibility along with chromitite crystallisation, but that influxes and mixing of slower-ascending magma, that fractionated en route, could give rise to sulphide liquid segregating along with the chromitite (i.e. the scenario for the LG-5 and overlying chromitites). The modelling also shows that the more fractionated the magma in the chamber becomes, the more sulphide will form, accounting for the overall upward increase in Pt and Pd above the LG-5.This paper is based on 465 new analyses of Ni, Cu, S and PGE from the 19 chromitite horizons between the LG-1 and UMG-2 from 6 sectors around the Bushveld Complex, along with microprobe analyses of representative samples of 41 chromites. Two trends in chromite composition, A and B, are distinguished on a plot of cation% Mg/(Mg + Fe2+) versus Cr/(Cr+Al). Trend A, that has a negative slope, is close to that predicted as the result of the reciprocal exchange substitution of Cr and Fe2+ for Mg and Al between spinel and liquid affecting the Mg-Fe2+ spinel-liquid Kd2. Trend B, that has a positive slope and is defined primarily by the LG-5 to MG-2 chromitites, is the result of the progressive increase in the activity of Al2O3 as a result of the fractional crystallization of orthopyroxene. Overall, the average PGE concentrations in massive chromitite increase upward. The LG-1 to LG-4 chromities have low (Pt+Pd)/(Rh+Ru+Ir+Os) ratios (0·1 to 0·3), above which there is an abrupt jump to higher ratios in the LG-5 (0·9 to 10) and all overlying chromitites (also documented by Scoon and Teigler). The Pt/Ru and Pd/Ru ratios are very variable, but the Ru/Ir, Ru/Rh and Ru/Os ratios of all chromitites are relatively constant, indicating that Pt and Pd respond to different concentration mechanisms to the other PGE. Rh, Ru, Ir and Os were likely concentrated by chromite itself, probably as grains of laurite and alloys incorporated in growing chromite crystals, but the bulk of the Pt, Pd along with lesser proportions of the other PGE were concentrated by sulphide liquid. Most chromitites now have very low contents of S, but mineragraphic and chemical data support the suggestion of Naldrett and Lehmann that vacancies in chromite forming above 900°C were filled by Fe2+ derived from the destruction of interstitial sulphide liquid. Data on En composition through the Bushveld CriticalZone, indicate that the LG-1 to LG-4 chromitites formed at a stage when influxes of magma into the chamber were rapid and primitive, and overrode the effect of fractional crystallization, whereas above this, fractionation mostly overrode influxes of new magma. Irvine's model of mixing of resident magma with influxes of more primitive magma is invoked as the origin of the chromitite horizons. It is shown, using the equation for sulphur solubility and the programme MELTS, that influxes and mixing of fresh primitive magma from depth with that in the chamber (i.e. as envisaged for the LG-1 to LG-4) would not have caused sulphide immiscibility along with chromitite crystallisation, but that influxes and mixing of slower-ascending magma, that fractionated en route, could give rise to sulphide liquid segregating along with the chromitite (i.e. the scenario for the LG-5 and overlying chromitites). The modelling also shows that the more fractionated the magma in the chamber becomes, the more sulphide will form, accounting for the overall upward increase in Pt and Pd above the LG-5.