Mechanisms of concentration of the platinmn-group and other chalcophile elements iu several platinum-group element deposits in layered intrusions are evaluated in light of recently published data on element partitioning between immiscible sulfide and silicate melts. The similarity of the partition coefficieut D (super sulfide melt/silicate melt) of Ir to that of Pd indicates that the observed fractionation of these elements in nature resulted from a process other than or in addition to sulfide liquid immiscibility. In experiments designed to test the possibility that this fractionation was due to the presence of Os-Ir alloy, sulfide and silicate melts were equilibrated with Ir metal at 1,450 degrees C, 8 kbars total pressure, and under sulfur and oxygen fugacity conditions appropriate for nature. The Ir content of the sulfide liquid was found to be > or = 3.7 wt percent and that of the coexisting silicate melt inferred to be > or = 1,000 ppb. These concentrations are far in excess of the Ir contents of natural sulfides or rocks and indicate that Os-Ir alloys are not stable in the presence of sulfide or silicate melts in nature.Using our preferred values of D (super sulfide melt/silicate melt) for the chalcophile elements and their estimated concentrations in the mantle, limits are determined for the possible ranges of concentrations and relative abundances of these elements in immiscible sulfide liquid formed in magma derived by partial melting of the mantle. The ratios of highly chalcophile to moderately chalcophile elements and of metal to sulfur of several deposits, including the J-M reef (Stillwater Complex), the UG-2 (Bushveld Complex), and the Lac des Isles Complex, are inconsistent with concentration solely by segregation of sulfide liquid. In contrast, the Merensky reef fits the geochemical criteria for formation by accumulation of magmatic sulfide. Stratigraphic variations in chalcophile element concentrations through the UG-2 and the Main Sulfide zone of the Great Dyke also indicate that metals must have been redistributed on at least the meter scale. Our analysis implies either different relative mobilities of the chalcophile elements relative to one another in a process postdating initial sulfide liquid deposition or concentration by a different mechanism.