Pii: S0962-8924(99)01663-3

Homology of segmentation and phylogeny Before considering the evolution of segmentation, we first need to specify what we mean by a ‘segment’. True segmentation, or metamerism, is usually considered to be the repetition along the anterior–posterior axis of a structural unit that comprises a suite of characters involving the entire body1. Animals and plants, which evolved multicellularity independently, possess simple serial repetition of structure, and serial repetition is also a feature of some animal appendages, such as insect antennae, suggesting that it evolved multiple times in many contexts. Traditionally, however, it is the body segments of arthropods, annelids and chordates that have been accorded special significance as examples of true metamerism. Historically, intuitive ideas concerning the evolutionary origins of segmentation in arthropods, annelids and chordates have strongly influenced our picture of the evolutionary relationships among the bilaterally symmetrical metazoans (the bilaterians). At the close of the 19th century, the segmentation observed in these three phyla was commonly held to be homologous, that is, derived from a segmented common ancestor. Thus, segmentation was often used to unite these groups within a single clade. However, the protostome–deuterostome distinction2, made at the beginning of the 20th century, asserted that most bilaterian phyla are more closely related to either chordates or annelids plus arthropods than these two groups are to each other. Since then, the deep phylogenetic separation of chordates and annelids plus arthropods has been retained and confirmed, leading many to regard the segmentation in these two groups as having evolved independently3. Similarly, true segmentation traditionally has been regarded as a shared, derived character of annelids and arthropods, uniting these phyla in a clade to the exclusion of unsegmented phyla, such as the molluscs4. However, recent analyses of morphological4 and molecular5,6 data have suggested that these two segmented phyla are actually more closely related to several unsegmented phyla than they are to each other (Fig. 1). If this most recent version of metazoan phylogeny7 is correct, then it represents a direct challenge to the supposed shared, derived characters that previously united annelids and arthropods, segmentation being primary among them. Essentially, we are now faced with three different hypotheses for the evolution of segmentation (Fig. 1). While it is true that mere phylogenetic separation does Millennium issue