Humans have seven alcohol dehydrogenase genes (ADH) falling into five classes. Three out of the seven genes (ADH1A, ADH1B and ADH1C) belonging to Class I are expressed primarily in liver and code the main enzymes catalyzing ethanol oxidization. The three genes are tandemly arrayed within the ADH cluster on chromosome 4 and have very high nucleotide similarity to each other (exons: >90%; introns: >70%), suggesting the genes have been generated by duplication event(s). One explanation for maintaining similarity of such clustered genes is homogenization via gene conversion(s). Alternatively, recency of the duplications or some other functional constraints might explain the high similarities among the genes. To test for gene conversion, we sequenced introns 2, 3, and 8 of all three Class I genes (total>15.0 kb) for five non-human primates--four great apes and one Old World Monkey (OWM)--and compared them with those of humans. The phylogenetic analysis shows each intron sequence clusters strongly within each gene, giving no evidence for gene conversion(s). Several lines of evidence indicate that the first split was between ADH1C and the gene that gave rise to ADH1A and ADH1B. We also analyzed cDNA sequences of the three genes that have been previously reported in mouse and Catarrhines (OWMs, chimpanzee, and humans) and found that the synonymous and non-synonymous substitution (dN/dS) ratios in all pairs are less than 1 representing purifying selection. This suggests that purifying selection is more important than gene conversion(s) in maintaining the overall sequence similarity among the Class I genes. We speculate that the highly conserved sequences on the three duplicated genes in primates have been achieved essentially by maintaining stability of the hetero-dimer formation that might have been related to dietary adaptation in primate evolution.