We studied the molecular phylogeny of the carabid subgenus Ohomopterus (genus Carabus) using two mtDNA regions (16SrRNA and ND5) and three nuclear DNA regions (wingless, phosphoenolpyruvate carboxykinase, and an anonymous locus). We revisited the previously reported incongruence between the distribution of mtDNA markers and morphologically defined species (Su et al., 1996, J. Mol. Evol., 43:662-671) which they attributed to "type switching", a concerted change in many morphological characters that results in the repeated evolution of a particular morphological type. Our mtDNA gene tree obtained from 44 individuals representing all 15 currently recognized species of Ohomopterus revealed that haplotypes isolated from individuals of a single "species" were frequently separated into distant clades, confirming the previous report. The three nuclear markers generally conformed better with the morphologically defined species. The phylogenetic signal in mtDNA and nuclear DNA data differed strongly and these two partitions were significantly incongruent with each other according to Farris et al.'s (1994) ILD test, although the three nuclear partitions were not homogeneous either. Our results did not support the type switching hypothesis which was proposed to fit the morphological data to the mitochondrial gene tree: incongruence of the mtDNA tree with other nuclear markers indicates that the mtDNA-based tree does not reflect species history any better than the morphological data. Incongruence of gene trees in Ohomopterus may have been promoted by the complex processes of geographic isolation and hybridization in the Japanese Archipelago that led to occasional gene flow and recombination between separated entities. It is intriguing that reticulate patterns occurred in this group, as species of Ohomopterus exhibit extremely divergent genitalic structures that represent a highly efficient reproductive isolation mechanism.