Galactokinase (GALK) deficiency is an autosomal recessive disorder, which causes cataract formation in children not maintained on a lactose-free diet. We characterized the human GALK gene by screening a Japanese genomic DNA phage library, and found that several nucleotides in the 5'-untranslated region and introns 1,2, and 5 in our GALK genomic analysis differed from published data. A 20-bp tandem repeat was found in three places in intron 5, which were considered insertion sequences. We identified five novel mutations in seven unrelated Japanese patients with GALK deficiency. There were three missense mutations and two deletions. All three missense mutations (R256W, T344M, and G349S) occurred at CpG dinucleotides, and the T344M and G349S mutations occurred in the conserved region. The three missense mutations led to a drastic reduction in GALK activity when individual mutant cDNAs were expressed in a mammalian cell system. These findings indicated that these missense mutations caused GALK deficiency. The two deletions, of 410delG and 509-510delGT, occurred at the nucleotide repeats GGGGGG and GTGTGT, respectively, and resulted in in-frame nonsense codons at amino acids 163 and 201. These mutations arose by slipped strand mispairing. All five mutations occurred at hot spots in the CpG dinucleotide for missense mutations and in short direct repeats for deletions. These five mutations in Japanese have not yet been identified in Caucasians. We speculate that the origin of GALK mutations in Japanese is different from that in Caucasians.