news and views cal evidence has it), then the aquatic adapta- characters. In recent years, more genes in level of homoplasy12,13, but mutational decay tions of hippos and living cetaceans must be larger sequence samples have been added of the flanking region of retroposons may convergences that occurred well after the to the arsenal for estimating ungulate and make them difficult to detect in older lin- split of their respective lineages. This is cetacean phylogeny9–12. And the poor taxo- eages. This means that retroposon-based because certain primitive cetaceans (paki- nomic sampling of some of the earlier mol- estimates may not be effective for resolving cetids) have many ear4 and ankle6 structures ecular work has at least in part been lineages that go back more than 50 million typical of a land mammal, and mesonychids redressed10. All in all, because molecular years14. Cetaceans were already diversified by were fully terrestrial and adapted to running characters can vastly outnumber morpho- 53.5 million years ago, and their divergence (Fig. 1a). logical features, they often prevail in simulta- from extant artiodactyls goes back much fur- What about the relative strengths of the neous analyses of conflicting data sets. That ther15,16 than that. two lines of evidence? Morphological stud- is, in such analyses they effectively swamp A way to untie this Gordian knot may be ies of ungulate–cetacean phylogeny take in a the morphological estimates of ungulate– to seek out compatible aspects of the molec- much wider range of taxa than the molecular cetacean phylogeny. ular and morphological data sets. Measure- studies. This is because the living artio- Both morphological and molecular data ment of the hidden support and conflict dactyls and cetaceans available for molec- are vulnerable to the problem of homo- between them10 can help extract additional ular analyses represent only a few twigs of plasies — reversals to ancestral conditions or information. Also, morphological features their bushy family trees that have survived parallel changes in different lineages that can should be better analysed for some living pruning by extinction. As O’Leary and camouflage the true phylogeny. In this sense, taxa for which extensive molecular data are Geisler1 point out, 90% of ungulate genera neither approach is better than the other. For available. Such studies may be just as helpful and more than 86% of the cetacean genera instance, the ear region of the skull, tradi- as discoveries of new fossils and genes in are extinct. Taking the mesonychid and tionally considered to be a good source of resolving details of the cetaceans’ phyloge- cetacean fossil taxa into account produces a highly stable characters, shows some glaring netic tree — most particularly, the question markedly different evolutionary history homoplasies among the ungulates and of which group is their sister taxon. I (Fig. 1a). From this, O’Leary and Geisler cetaceans4,5. Moreover, the fossil record of Zhexi Luo is in the Section of Vertebrate conclude that artiodactyls are a clade to the many early divergent fossil taxa is incom- Paleontology, Carnegie Museum of Natural History, exclusion of cetaceans. Here we have a good plete, resulting in ambiguities in morpho- Pittsburgh, Pennsylvania 19213, USA. example of the principle that including data logical estimates. e-mail: [email protected] from early divergent fossil lineages can On the molecular side, DNA and protein 1. O’Leary, M. A. & Geisler, J. H. Syst. Biol. 48, 455–490 (1999). shake, and reshape, the trees of living taxa sequences have parallel and back mutations. 2. O’Leary, M. A. Cladistics 15, 315–330 (1999). 3. Van Valen, L. M. Bull. Am. Mus. Nat. Hist. 132, 1–126 based solely on molecular evidence. Even the newest studies using retroposons, (1966). However, for living taxa, use of sequen- which are the RNA-mediated insertion 4. Luo, Z. & Gingerich, P. D. Papers on Paleontology (Univ. Michigan) 31, 1–98 (1999). ces of genes and proteins is in some ways sequences interspersed in the genome, have 5. Geisler, J. H. & Luo, Z. in Evolutionary Emergence of Whales (ed. more powerful than use of morphological their limitations. Retroposons show a low Thewissen, J. G. M.) 161–212 (Plenum, New York, 1998).

Fibre optics Transparent talk

Modern telecommunications rely on optical fibres involves heating silica the transmission of light signals rods with ultra-pure glass cores to along fibre-optic cables — fast, but very high temperatures. Once how to minimize signal degradation softened, the rods can be drawn out over large distances? The secret is into fibres some tens of kilometres to eliminate contaminants that long. To achieve the necessary cause unwanted absorption. These temperatures — in excess of intruders tend to creep into the 2,000 °C — it is common to use fibres during the manufacturing torches that burn hydrogen and stages. Elsewhere in this issue oxygen: the perfect recipe for water. (Nature 404, 262–264; 2000), But the incorporation of water Gordon Thomas and colleagues into the silica rods is a more subtle reveal how the main culprit, water, matter. To investigate this process, sneaks in. Putting this knowledge Thomas and colleagues measured into practice allows the manufacture the transmission of infrared light diffuses from the outside in. process. This parameter was of dry fibres with near maximal through a small section of rod that Given that the signals in optical much higher than expected from transparency and improved had been cooled before drawing out, fibres are confined to a narrow core low-temperature diffusion, probably bandwidth. thereby ‘freezing in' the water. They region, the distribution of water in reflecting an increased mobility of The problem with water, in found that the absorption, and the rod might appear to be good hydroxyl groups in the hot, molten particular the hydroxyl (OH) group, hence the hydroxyl concentration, is news. But Thomas and colleagues state. is that it gets excited when much stronger in the outer layers of found that the drawing process — There is no doubt that irradiated by certain infrared the rod. This can be seen in the which increases the aspect ratio by absorption by hydroxyl groups wavelengths. As this band is used figure, which indicates the varying a factor of about a hundred million contributes significantly to for telecommunications, precious hydroxyl concentrations through a — lets the water in much further. transparency loss in optical fibres. signal power can be lost in vibration cross-section of the rod. (The colour They confirmed this by calculating Uncovering the physical origin of the of the water molecules. scale is logarithmic, blue the diffusion coefficient, a quantity contamination points to an obvious So where does the water come representing the highest that describes the flow of both water solution: pick a water-free heat from? One process for making concentrations.) Clearly, the water and glass during the contamination source. Karen Southwell

NATURE | VOL 404 | 16 MARCH 2000 | www.nature.com © 2000 Macmillan Magazines Ltd 237