Abstract The behaviour of the Pt-group elements (PGE) during both normal igneous processes and during the formation of PGE deposits is poorly understood. This is in part because of the limited data set available for nonmineralized rocks. Accordingly, PGE concentrations have been determined in komatiitic basalts associated with Ni-Cu sulphide deposits rich in PGE from the Proterozoic Cape Smith Fold Belt of northern Quebec. The lavas have been divided into olivine-phyric, pyroxene-phyric, and plagioclase-phyric on the basis of the phenocrysts present. Olivine-phyric lavas have Mg#s greater than 0.66 and are thought to be primary partial melts. The bulk partition coefficients of the PGE during partial melting calculated from the average of olivine-phyric basalts are 6, 2, 0.6, and 0.2 for Ir, Rh, Pt, and Pd, respectively, which indicates a decrease in compatibility of the PGE in this order. The pyroxene and plagioclase-phyric lavas are thought to have formed from the olivine-phyric lavas by crystal fractionation of olivine and chromite in the pyroxene-phyric lavas and olivine, chromite, and pyroxene in the plagioclase-phyric lavas. Ir shows a strong positive correlation with Mg#, Cr, and Ni, and this is attributed to Ir having partitioned into olivine or chromite. If olivine, chromite, and clinopyroxene were the only phases to have crystallized, then Pd should have behaved as an incompatible element; however, it does not correlate with the lithophile incompatible elements. It does, however, correlate with Rh and Pt. This is interpreted to suggest that Rh, Pt, and Pd were controlled by sulphide segregation. The removal of a small amount of sulphide along with the olivine and chromite during crystal fractionation could have resulted in the mildly compatible behavior exhibited by Rh, Pt, and Pd. The composition of the sulphides in the sulphide deposits may be modelled by assuming that the silicate magma from which they segregated was similar in composition to the spinifex-textured komatiites containing 15% MgO.