Petrological models relating the different rock types constituting the 1.85 Ga Sudbury Igneous Complex are constrained with extensive new geochemical data. We show that the main mass felsic norite, transition zone quartz gabbro, and granophyre have similar ratios of the highly incompatible trace elements (e.g., La/Sm = 4.5-7, La/Nb = 2.8-4.2, Th/Zr = 0.04-0.05) and that these variations are consistent with the crystallization and differentiation of magma types largely (>80%) derived from the upper crust, with a smaller contribution from a mantle source. Although there is presently no conclusive proof that magma was generated in situ as a melt sheet produced by meteorite impact, we find no principal reason why this model should be rejected. However, we propose that a small contribution of mantle-derived picritic magma is required to explain the abundant Ni, Cu, and platinum-group elements (PGE) in the Sudbury deposits, as well as the compositions of the ultramafic inclusions (MgO = 12-36 wt %; Fo (sub 68-87) olivines with 450-3,700 ppm Ni, and abundant chrome-rich spinel), and the magnesian composition of the mafic norite (8-14 wt % MgO) and the sublayer (6-12 wt % MgO). We believe that the main mass of the Sudbury Complex achieved its present composition through incorporation of up to 20 percent mantle-derived picritic magma emplaced along crustal fractures produced by the impact event. These picritic magmas entered the melt sheet as a dense plume, vigorously mixing with it, and due to the marked compositional shift, the mixed magma formed magmatic sulfides which sank through the magma column, depleting the melt in Ni, Cu, and PGE. Since both the felsic norite and granophyre have indistinguishable ratios of the incompatible trace elements, we see no requirement to derive these units of rock by the crystallization of magmas derived from different sources. Rather, the compositional difference between the felsic norite and granophyre is attributed to the in situ differentiation of the magma. We show that the main mass has many compositional traits similar to those of most of the offset dike quartz diorites (e.g., the Parkin offset dike: La/Sm = 6.3, La/Nb = 4.5; Th/Zr = 0.05) and of embayment-related leucocratic norites from the Whistle mine (La/Sm = 6.2, La/Nb = 5.0, Th/Zr = 0.02). These rocks have compositions intermediate between the felsic norite and the granophyre, and therefore crystallized from the same magma type; arguably, the unmineralized quartz diorites provide the best possible estimate of the original magma from which the Sudbury Complex crystallized. In detail, there are subtle variations in composition within and between offset dikes, with the largest difference being between the North and South Range offsets; the North Range offset dikes cut Archean granitoids and gneisses and have elevated Sr, La/Yb, La/Sm, and Gd/Yb and low TiO 2 whereas the South Range dikes cut Early Proterozoic sediments, mafic volcanics, and intrusions, and have low Sr, La/Yb, Gd/Yb, La/Sm, and high TiO 2 . These differences may be caused by the assimilation of different country rocks during emplacement of the dike. A strongly mineralized offset dike at the Creighton mine has geochemical variations that are different when compared to the main mass, and in the case of Creighton, are more similar to the local mineralized sublayer. These data suggest that mineralized and barren quartz diorites have different geochemical compositions, and that these traits may be of value in mineral exploration.