Many marine species, including mussels in the Mytilus edulis species group (i.e. M. edulis L., M. galloprovincialis Lamarck, and M. trossulus Gould), have an antitropical distribution pattern, with closely related taxa occurring in high latitudes of the northern and southern hemispheres but being absent from the tropics. We tested four hypotheses to explain the timing and route of transequatorial migration by species with antitropical distributions. These hypotheses yield dierent predictions for the phylogenetic relationship of southern hemisphere taxa relative to their northern counter-parts. The three Mytilus species were used to test these hypotheses since they exhibit a typical antitropical distribution and representative taxa occur in both the Paci®c and Atlantic. Two types of mtDNA lineages were found among populations of mussels collected from the southern hemisphere between 1988 and 1996; over 90% of the mtDNA lineages formed a distinct subclade which, on average, had 1.4% divergence from haplotypes found exclusively in northern Atlantic populations of M. galloprovincialis. These data indicate that southern hemisphere mussels arose from a migration event from the northern hemisphere during the Pleistocene via an Atlantic route. The remainder of the southern hemisphere lineages (<10%) were very closely related to mtDNA haplotypes found in both M. edulis and M. galloprovincialis in the northern hemisphere, suggesting a second, more recent migration to the southern hemisphere. There was no evidence that southern hemisphere mussels arose from Paci®c populations of mussels.

Summary. Since the discovery of the coelacanth, L a t i m e r i a c h a l u m n a e , more than 50 years ago, paleontologists and comparative morphologists have debated whether coelacanths or lungfishes, two groups of lobe-finned fishes, are the closest living relatives of land vertebrates (Tetrapoda). Previously, Meyer and Wilson (1990) determined partial DNA sequences from two conservative mitochondrial genes and found support for a close relationship of lungfishes to tetrapods. We present additional D N A sequences from the 12S r R N A mitochondrial gene for three species of the two lineages of lungfishes that were not represented in the first study: Protopterus a n n e c t e n s a n d Protopterus aethiopicus from Africa and N e o c e r a t o d u s forsteri (kindly provided by B. Hedges and L. Maxson) from Australia. This extended data set tends to group the two lepidosirenid lungfish lineages (Lepidosiren and Protopterus) with N e o c e r a t o d u s as their sister group. All lungfishes seem to be more closely related to tetrapods than the coelacanth is. This result appears to rule out the possibility that the coelacanth lineage gave rise to land vertebrates. The common ancestor of lungfishes and tetrapods might have possessed multiple morphological traits that are shared by lungfishes and tetrapods [Meyer and Wilson (1990) listed 14 such traits]. Those traits that seem to link L a t i m e r i a and tetrapods are arguably due to convergent evolution or reversals and not to common descent. In this way, the molecular tree facilitates an evolutionary interpretation of the morphological differences among the living forms. We recommended that the extinct groups of lobe-finned fishes be placed onto the molecular tree that has lungfishes and not the coelacanth more closely related to tetrapods. The placement of fossils would help to furOffprint requests to: Axel Meyer ther interpret the sequence of morphological events and innovations associated with the origin of tetrapods but appears to be problematic because the quality of fossils is not always high enough, and differences among paleontologists in the interpretation of the fossils have stood in the way of a consensus opinion for the branching order among lobefinned fishes. Marshall and Schultze (1992) criticized the morphological analysis presented by Meyer and Wilson (1990) and suggest that 13 of the 14 morphological traits that support the sister group relationship of lungfishes and tetrapods are not shared derived characters. Here we present further alternative viewpoints to the ones of Marshall and Schultze (1992) from the paleontological literature. We argue that all available information (paleontological, neontological, and molecular data) and rigorous cladistic methodology should be used when relating fossils and extant taxa in a phylogenetic framework. K e y w o r d s : Polymerase chain reaction --12S rRNA --Coelacanth --L a t i m e r i a c h a l u m n a e -- Ray-finned fishes --Lungfishes --L e p i d o s i r e n --Protopterus --N e o c e r a t o d u s ...