We compared 609 haplotypes carrying the idiopathic hemochromatosis allele with 475 control haplotypes. Four haplotypes were more frequent in hemochromatosis: A3, B7 (actually A3, CW., B7, Bfs, DR2); A3, B14 (actually A3, CW., B14, BfF, DRW6); A11, B35; and A11, B5. The linkage disequilibrium for A3, B7 and A3, B14 (and probably also for A11, B5) was undeniably stronger in hemochromatosis than in controls. Two haplotypes--A3, B12 and A3, B15--were more frequent in hemochromatosis, without linkage disequilibrium. Four haplotypes in linkage disequilibrium in hemochromatosis--i.e., A2, B12; A1, B8; A9, B7; and A29, B12--were also found to have the same frequency and strength of linkage in controls. The dual observation (1) that haplotypes carrying A3 without either B7 or B14 were highly significantly more frequent in hemochromatosis than in controls and (2) that haplotypes carrying B7 or B14 but not A3 had the same frequency in hemochromatosis and controls led to the formal conclusion that only A3 is an independent marker for the hemochromatosis allele, B7 and B14 being involved only owing to the haplotypic mode of marking; the hemochromatosis allele can thus be mapped closer to locus A than to locus B. Our findings fit well with the hypothesis that the hemochromatosis mutation was a rare if not unique event that produced an ancestral HLA marking that was subsequently modified by recombinations and geographical scattering due to migrations.