Marrow transplants were carried out between unrelated donor-recipient pairs of dogs that were homozygous and identical for DLA-A, B, C, and D, i.e., mutually nonreactive in mixed leukocyte culture. Recipients were conditioned for transplantation by 1,200 R of total body irradiation and then treated with intermittent methotrexate for 102 days in order to prevent or delay graft-versus-host disease (GVHD). Of 13 dogs that received transplants, 4 are surviving with good grafts and no GVHD for more than 12 to 20 minutes. Nine died, 6 with GVHD between days 26 and 141, 1 with wasting on day 65, 1 with interstitial pneumonia on day 83, and 1 with graft rejection on day 23. In comparison, the survival of 17 DLA-identical littermates treated in the same manner was significantly better with 16 surviving without GVHD (P less than 0.01), while the survival of 54 DLA-nonidentical littermates was significantly worse with only two surviving without GVHD (P less than 0.025). These results are incompatible with the concept that solely the loci detected by mixed leukocyte culture and serotyping are responsible for GVHD. One or more additional loci appear to be involved. Knowledg e of this locus (loci) is important if marrow grafting between unrelated individuals is to be successful. However, results also indicate that an unrelated "compatible" marrow graft is more likely to succeed than a graft from an incompatible littermate.