Catalase-peroxidases belong to Class I of the plant, fungal, bacterial peroxidase superfamily, together with yeast cytochrome c peroxidase and ascorbate peroxidases. Obviously these bifunctional enzymes arose via gene duplication of an ancestral hydroperoxidase. A 230-residues long homologous region exists in all eukaryotic members of Class I, which is present twice in both prokaryotic and archaeal catalase-peroxidases. The overall structure of eukaryotic Class I peroxidases may be retained in both halves of catalase-peroxidases, with major insertions in several loops, some of which may participate in inter-domain or inter-subunit interactions. Interspecies distances in unrooted phylogenetic trees, analysis of sequence similarities in distinct structural regions, as well as hydrophobic cluster analysis (HCA) suggest that one single tandem duplication had already occurred in the common ancestor prior to the segregation of the archaeal and eubacterial lines. The C-terminal halves of extant catalase-peroxidases clearly did not accumulate random changes, so prolonged periods of independent evolution of the duplicates can be ruled out. Fusion of both copies must have occurred still very early or even in the course of the duplication. We suggest that the sparse representatives of eukaryotic catalase-peroxidases go back to lateral gene transfer, and that, except for several fungi, only single copy hydroperoxidases occur in the eukaryotic lineage. The N-terminal halves of catalase-peroxidases, which reveal higher homology with the single-copy members of the superfamily, obviously are catalytically active, whereas the C-terminal halves of the bifunctional enzymes presumably control the access to the haem pocket and facilitate stable folding. The bifunctional nature of catalase-peroxidases can be ascribed to several unique sequence peculiarities conserved among all N-terminal halves, which most likely will affect the properties of both haem ligands.