Various studies suggest that the formation of a magmatic Ni-Cu-platinum group element (PGE) sulfide deposit requires addition of S from the country rocks into the magma. Many sulfide deposits have S/Se ratios higher than primary magmas, which is attributed to the preferential addition of S from the country rocks. However, this has not been clearly demonstrated because most studies do not determine Se in the country rocks, partly because, although Se concentration in igneous rocks can be assessed by most routine analytical techniques, their detection limits are too low for measuring with precision the level of Se in sedimentary rocks. We have determined Se abundances, using thiol cotton fiber (TCF) combined with instrumental neutron activation analysis (INAA), S, and δ34S on the magmatic sulfides of the Duluth Complex and on the host-rock sediments of the Virginia Formation. Our data show that the pelitic and metapelitic rocks of the Virginia Formation are characterized by low S/Se ratios (~3,000), close to the mantle values, whereas the S-rich layer known as the bedded pyrrhotite unit has high S/Se ratios (~20,000). Consequently, most of the Virginia Formation sediments and xenoliths are unsuitable as a source of S because they have S/Se ratios and δ34S values lower than those of the mineralized rocks (<10,000). Only the bedded pyrrhotite unit, which is present at the contact with the intrusion and as partly digested xenoliths within the intrusion, has S/Se ratios and δ34S values higher than those of the magmatic sulfides and could thus be the source of the S. Furthermore, the presence of sulfides in the norites around the xenoliths suggests that the S was released from the xenolith during melting of the xenolith. Detailed sampling from the xenoliths into the norite shows that the S/Se ratio and δ34S decrease with distance from the xenolith. Our results favor a mechanism of crustal S contamination by in situ assimilation of bedded pyrrhotite unit xenoliths.