Mitochondrial DNA (mtDNA) is functionally conservative, encoding basically the same genes in all eukaryotes, yet mitochondrial genome organization and expression show an amazing diversity, and mtDNAs are highly variable in size and in potential information content. This review focuses on certain novel features of mitochondrial genome diversity which have emerged from recent comparative studies of mtDNA and which indicate that, in the course of evolution, mtDNAs have undergone massive sequence rearrangements, have lost (and (or) acquired) large stretches of sequence between and within genes, and have sustained, in at least some cases, an unusually high degree of primary sequence divergence. It is suggested that much of the mitochondrial genome diversity that presently exists can be accounted for by a variable (and in some cases, quite rapid) rate of evolution of mtDNA. An endosymbiotic origin of mitochondria can be accommodated in this scenario if it is assumed that the "protomitochondrion," although eubacterial in essential elements of its translation system (e.g., ribosomal RNA), was unlike modern eubacteria in containing introns and extensive intergenic "spacer" sequences in its genome. Differing evolutionary pressures on the "free-living" and "intracellular" progenitors of eubacteria and mitochondria, respectively, could have resulted in the subsequent elimination of these "extra" sequences at different rates and to different extents, leading in the case of humans and other mammals to a mitochondrial genome having little or no noncoding sequences. Viewed from this perspective, most of the introns and intergenic sequences in some larger mtDNAs are considered to be retained primitive traits, vestiges of genome organization in the universal "progenote" from which a very early divergence of the eukaryotic nuclear, eubacterial, and archaebacterial lineages has been postulated. In all but the more conservative mitochondrial genomes, rapid primary sequence divergence will have tended to obscure the phylogenetic ancestry of most mtDNA-encoded macromolecules. This could explain why animal and fungal mitochondrial tRNAs, and plant mitochondrial 5S rRNA, are not typically eubacterial. A variable rate of mtDNA evolution is indicated by the fact that plant mitochondrial ribosomes are less "atypical" than the same components in other mitochondria, with plant (wheat) mitochondrial small-subunit rRNA showing much clearer indications of a eubacterial origin than its counterparts in animal (human, mouse) and fungal (yeast) mitochondrial ribosomes. Thus, the plant mitochondrial genome may be evolving considerably less rapidly than its counterparts in fungi and (especially) vertebrate animals. Our present state of knowledge about the structure, function, and evolution of mtDNA demonstrates both the power of, and essential need for, broadly based comparative studies in this field.