A model is presented for the formation of platinum-group element ores by precipitation from sulfur-deficient, platinum-group element-rich, second-stage magmas. Numerous investigations indicate that basaltic magmas originating from undepleted or mildly depleted source regions are sulfur saturated at the time of eruption. Mantle-derived peridotites residual from midocean ridge basalt magma genesis have Pd abundances that are uniformly too high for it to have originated solely from silicate melt retained in the source region at the time of segregation; they imply the presence of an accessory immiscible sulfide melt to host the platinum-group elements. The very high sulfide melt-silicate melt partition coefficients of the platinum-group elements would result in strong fractionation of these metals into the accessory sulfide component during the melting event. The extreme platinum-group element enrichment of this accessory sulfide component is evidenced by the discovery ofplatinoid minerals in residual mantle nodules containing Pd (and Pt) abundances at the low ppb level (Keays et al., 1981).Magmas generated by low to moderate degrees of partial melting of undepleted mantle are therefore sulfur saturated at the time of segregation. These (first-stage) magmas become impoverished in platinum-group elements during the early stages of fractional crystallization because of coprecipitation of an immiscible sulfide phase (e.g., midocean ridge basalt glasses average <0.8 ppb Pd). Further melting of this depleted source leads to complete dissolution of the residual platinum-group element-rich sulfide component and production of second-stage magmas enriched in them but poor in sulfur.Boninitic magmas, generally regarded as second-stage melts generated from strongly depleted mantle sources, have enhanced platinum-group element abundances (mean Pd = 15 ppb) in agreement with the above model. During fractionation of these S-deficient magmas the concentration of platinum-group elements will build up until the onset of sulfur saturation and segregation of sulfides. These sulfides will generally have a higher platinum-group element tenor than those segregating from first-stage melts.A marked similarity in the composition of boninitic magmas and early parental magmas proposed for the Bushveld Complex (Davies et al., 1980; Sharpe, 1981), and possibly other intrusions hosting platinum-group element ores, suggests that their source regions have evolved in a similar manner. This accounts for the timing of deposition of the UG-2 and Merensky cyclic units in the upper partition of the critical zone and the unusual platinum-group element-rich nature of their ores. If this model is relevant to the Bushveld platinum-group element mineralization, then it provides a means of assessing the potential of other stratiform intrusions to host economic platinum-group element deposits.