A detailed study of the distribution of Pd, Ir, and Au in the sulfide component of contact ore layers at Lunnon shoot has shown consistent differences between the matrix and massive layers and significant lateral variation within them. Palladium values in sulfides increase upward in ore sections, and when average values are compared with those predicted by the relationship between Pd and Ni in mined ore (Ross and Keays, 1979), loss of Pd is indicated. The Pd-Au-Cu-enriched footwall sulfide stringers are believed to be the principal repository of remobilized Pd. Iridium has a more uniform distribution and there is no evidence of important loss or gain from contact ore, even though massive layer sulfides contain more Ir at nearly all sample locations. Average values of Au are higher in matrix layer sulfides and some loss of Au from contact ore into footwall stringers, and perhaps more distant sites, is likely. If it is assumed that the distribution of these metals, and of Ni and Cu, was uniform during emplacement of sulfide melt, their present distribution can be attributed to the following processes. Redistribution of Ir could have occurred during igneous cooling in the presence of a high thermal gradient at the base of the komatiitic volcanic pile and could have been accompanied by some movement of Pd in areas of active stress. Most redistribution of Pd is attributed to stress during the existence of metamorphic monosulfide solid solution, and significant redistribution of Cu may also have been induced by stress. The circulation of latestage sulfur-bearing solutions through contact ore layers (Seccombe et al., 1981) could have transferred Pd, Au, and Cu from contact sulfides (particularly from the more permeable massive layer) to footwall stringers and resulted in the close association of anomalous values of Au and Cu. Transfer of about 5 percent of the S in the matrix layer to the massive layer by these solutions would account for the consistently higher Ni content of the matrix layer and development of secondary pyrite in the massive layer.Analyses of sulfide mineral separates from several shoots have consistently shown that pentlandite is the dominant host for Pd; Ir is more uniformly distributed with higher values in pyrite; and chalcopyrite (and perhaps pyrite) is the most important host for Au. Concentrates of secondary pyrite appear to have normal values, which suggests some movement of Ir during its development. The precious metals appear to be uniformly distributed through sulfide mineral hosts, but observed variation in ore sections and layers requires that their abundance in the preferred sulfide host must vary considerably.Further work has strengthened the conclusion of Ross and Keays (1979) that olivine in the original komatiitic magmas was the principal host for Ir and the liquid component was the principal host for Pd. Olivine control lines for Pd and Ir in primary melts predict values at lower MgO in good agreement with those in hanging-wall basalts at Kambalda. At higher MgO they agree closely with the most recent estimate for pyrolite and a chondritic Pd:Ir ratio of about unity. These predicted values result in a substantial increase in Pd:Ir with decreasing MgO. Results from Kambalda and other Ni sulfide deposits indicate that the Pd:Ir ratio of the sulfide melt may approximate that of the coexisting silicate melt even though absolute values may vary widely. This variation exceeds that expected from the relative volumes of silicate and sulfide melt and may result from differences in factors such as f (sub s 2 ) and f (sub o 2 ) when immiscible sulfides formed. Measurements of Pd and Ir offer scope for defining the MgO contents of primary silicate melts associated with immiscible sulfides and may allow a more precise definition of f (sub s 2 ) (and perhaps f (sub o 2 ) ) during formation of immiscible sulfide melts.