To test the hypothesis advanced by Gilman and Huisman that the -158 site 5′ to the G gamma gene determines the G gamma expression after the first 4 months of life, we have examined DNA from sickle cell anemia (SS) patients from Africa and beta-thalassemic homozygotes from Algeria. We find that the Xmnl site is strongly linked to the Senegal haplotype among SS patients, to haplotype IX (most probably identical to the Senegal haplotype), and to haplotype III among the Algerian thalassemics. Thalassemics with haplotypes I/I and V/V have no Xmnl site and low G gamma expression. In contrast, beta-thalassemia- associated haplotype II (also characterized by high G gamma expression) fails to exhibit the Xmnl site. We conclude that, although highly correlated, the -158 C----T substitution does not perfectly predict the presence of high G gamma expression. These findings also exclude the possibility that the Xmnl site is solely involved in the determination of high G gamma expression and suggest that either several different site substitutions in the area 5′ to the gamma gene might have the same effect or that, alternatively, the Xmnl site and its surrounding area is not involved in G gamma expression and may be only in linkage disequilibrium with a controlling sequence elsewhere.