the system,the origin of which is unclear.The pulsed emission in the flare’s tail

comes from a fraction of the plasma that initially remains confined by the magneticfield lines anchored to the neutron star. Asthe star rotates, it produces a ‘lighthouse’effect, resulting in the periodic oscillations in brightness.

On a timescale of days after the initialflare, X-ray9 and radio emissions6,7 originat-ing from SGR 1860�20 were also observed.Gaensler et al.6 (page 1104) report that theexpanding radio nebula had a luminositymore than 500 times greater than the onlyother similar object detected (after the flareof August 1998). In some ways the afterglowwas similar to that of long �-ray bursts(GRBs). GRBs are another class of bright �-ray flashes that are known to originate fromgalaxies on the edge of the visible Universe.Long GRBs last for tens of seconds and arefollowed by a dimmer emission tail lastingweeks to months. Cameron et al.7 (page1112), however, point out general problemsin reconciling the spectrum and light curveof the SGR 1860�20 radio emission withstandard models of long GRBs.

The most striking property of the SGR 1860�20 flare is its extraordinary lumi-nosity. This raises the question of whetherthis giant flare and its two predecessors1,2 arerelated to a class of mysterious short GRBs,detected in large numbers by BATSE (the

news and views

NATURE | VOL 434 | 28 APRIL 2005 | www.nature.com/nature 1075

Burst and Transient Source Experiment), asoft �-ray detector that flew on board NASA’sCompton observatory in the 1990s.No after-glow emission has so far been identified fromthese GRBs. However, if we rescale the prop-erties of the SGR 1806–20 flare to a distanceof several megaparsecs (around a thousandtimes farther away), an instrument such asBATSE would indeed have seen only the ini-tial bright spike of the event, with a timescaleof several hundred milliseconds and a spec-trum consisting almost entirely of hard pho-tons. Such a description matches that of theshort GRBs detected. Hurley et al.3 (page1098) estimate that the rate of giant flaresthat should have been detectable by BATSE isabout 30 per year, which could account forup to 40% of the short GRBs actually found.

There are three ways to identify the pres-ence of SGR flares in the BATSE catalogue of short bursts. First, they should be muchcloser than cosmological GRBs,and thereforeassociated with bright galaxies. The positionin the sky is known with the required accu-racy for only five short GRBs; but no brighthost galaxy can be associated with any ofthese five10, limiting the possible proportionof SGR flares to less than 20% in the BATSEshort-burst catalogue. Second, SGRs shouldproduce a periodic signal in the 200 secondsfollowing the burst; a search for such a signalhas so far been unsuccessful. Third, SGR candidates in the BATSE catalogue can be

A certain flareDavide Lazzati

Giant flashes from soft �-ray repeaters are spectacular but rare events — only three have ever been observed in our Galaxy. The suspicion is thatwe have been missing some from farther afield.

On 27 December 2004, virtually all of the �-ray detectors in orbit were triggered by the brightest flash of

�-rays ever seen. Two similar flares had pre-viously been detected from different sourcesof the same class during 30 years of observa-tions1,2 — on 5 March 1979 and 27 August1998. The 2004 flare, however, must beregarded as unique: it outshone both the preceding events by two orders of magni-tude, releasing in its first fraction of a secondas much energy as the Sun releases in a quarter of a million years. Five papers3–7 inthis issue provide an observational overviewof this exceptional event.

The source of the outburst is known asSGR 1806�20, a ‘soft �-ray repeater’ (SGR)lying in our Galaxy at an estimated distanceof 15 kiloparsecs (almost 50,000 light years)from the Solar System (Fig. 1). An SGR is anextremely highly magnetized neutron star8,or ‘magnetar’, that produces recurrent burstsof low-energy (‘soft’) �-rays — that is,high-energy photons, or electromagneticradiation. The flare from SGR 1806�20 wascharacterized by an initial spike that lastedless than a second and contained most ofthe energy of the burst3,4 as well as the high-est-energy, or ‘hardest’, photons. This spikeof the flare was followed by an exponentialtail with a duration of some 400 seconds,oscillating with the period (7.56 seconds) at which SGR 1806�20 is known — frommeasurements of its much dimmer X-rayemission during quiescence — to rotate.

The characteristics of the SGR 1860�20flare can be explained as the outcome of areadjustment of the huge magnetic field —up to 1015 times stronger than that at Earth’ssurface — anchored to what is a relativelyyoung (about 5,000-year-old) neutron star8.Such a readjustment releases a sizeable fraction of the internal energy of the field,stored in a hot ‘plasma’ of radiation and electron–positron pairs, and generates thebright initial spike of the flare.

The flux of photons in the spike ofSGR 1860�20 was so large that it saturatedmost detectors3, making it difficult to characterize its properties. Terasawa et al.5

(page 1110), however, report an oscillatorymodulation,with a period of around 60 milli-seconds, in the number of photons detectedin the spike. They suggest that the periodi-city of these ‘humps’ in the flare’s profileindicates repeated injections of energy into

Figure 1 Site unseen. A wide-field view of the area around SGR 1806–20 (at the centre of the white circle) before the colossal �-frequency flare of 27 December 2004, from radio frequencymeasurements. At this point SGR 1806–20 was still ‘radio quiet’; an intense radio nebula emanatingfrom the neutron star was only observed days after the first �-ray burst6,7.

UN

IV.H

AW

AII

28.4 n&v 1075 MH 22/4/05 5:19 pm Page 1075

Nature Publishing Group© 2005

© 2005 Nature Publishing Group

identified by their spectrum. SGR flares typi-cally emit a thermal ‘black-body’ spectrum(indicative of an optically thick medium),whereas GRBs are characterized by broadpower-law spectra (emitted by optically thinmaterial) with the radiation spread overmany orders of magnitude in frequency. TheSGR 1806–20 flare had an average tempera-ture of 2�109 kelvin, which should be easily identifiable among non-thermal GRBspectra; but a search for high-temperaturethermal spectra performed over about 100BATSE short GRBs has been unsuccessful.

These factors constrain the percentage ofpossible SGR flares among short GRBs to 5%at most, close to an order of magnitude lessthan expected.So,where are the extragalacticSGR flares?

Several factors may explain the apparentdiscrepancy. First, the distance to SGR 1806–20 could be inaccurate. If the flare were atonly half the distance assumed, its luminos-ity would be reduced by a factor of four,limiting the volume of space in which extra-galactic SGRs could be detected by a factor of eight. A possible smaller distance to SGR 1806–20 has been proposed. An upperlimit of 9.8 kiloparsecs, two-thirds of theprevious working assumption11 of 15 kilo-parsecs, is suggested by Cameron et al.7.

Another factor in the calculation is therate of occurrence of giant flares in ourGalaxy. SGR 1806–20 is the only flare of itsscale to be seen in 30 years of observations,and this is our best estimate of how often theyoccur. Of course, we may merely have beenlucky, and the true rate could be lower: wemight just happen to live in an epoch whenone of these flares went off randomly.

Whatever the answers to these questions,the hunt for extragalactic flares is on. In theiranalysis, Palmer et al.4 (page 1107) use datafrom NASA’s Swift satellite, which waslaunched in 2004 expressly to solve the GRBmystery,and the rapid follow-up capabilitiesof this mission should facilitate further dis-coveries. The importance of such detectionsfor our understanding of an extreme phe-nomenon that is represented so far only bythe example of 27 December 2004 cannot beoverestimated. ■

Davide Lazzati is at JILA, University of Colorado at Boulder, 440 UCB, Boulder, Colorado 80309-0440, USA.e-mail: lazzati@colorado.edu1. Mazets, E. P. et al. Nature 282, 587–589 (1979).

2. Hurley, K. et al. Nature 397, 41–43 (1998).

3. Hurley, K. et al. Nature 434, 1098–1103 (2005).

4. Palmer, D. M. et al. Nature 434, 1107–1109 (2005).

5. Terasawa, T. et al. Nature 434, 1110–1111 (2005).

6. Gaensler, B. M. et al. Nature 434, 1104–1106 (2005).

7. Cameron, P. B. et al. Nature 434, 1112–1115 (2005).

8. Thompson, C. & Duncan, R. C. Mon. Not. R. Astron. Soc. 275,

255–300 (1995).

9. Mereghetti, S. et al. Astrophys. J. Lett. (submitted); preprint

available at www.arxiv.org/astro-ph/0502577 (2005).

10. Nakar, E., Gal-Yam, A., Piran, T. & Fox, D. B. preprint available

at www.arxiv.org/astro-ph/0502148 (2005).

11. Corbel, S. & Eikenberry, S. S. Astron. Astrophys. 419, 191–201

(2004).

tree the A–P–C hypothesis. Under A–P–C,humans are more closely related to the fruit-fly Drosophila melanogaster than either is tothe nematode roundworm Caenorhabditiselegans5,6 (Fig.1).

In contrast, the new trees1–3,7 suggest thatthe basic division in animals is between theProtostomia and Deuterostomia (a distinc-tion based on the origin of the mouth duringembryo formation). Humans are deutero-stomes, but because flies and nematodes areboth protostomes they are more closelyrelated to each other than either is tohumans. The Protostomia can be dividedinto two ‘superphyla’: Ecdysozoa (animalsthat undergo ecdysis or moulting, includingflies and nematodes) and Lophotrochozoa(animals with a feeding structure called thelophophore, including snails and earth-worms).We call this tree the L–E–D hypoth-esis (Fig.1). Importantly, in this new tree, thecoelom must have arisen more than once, orhave been lost from some phyla.

Molecular analyses have been divided in

news and views

1076 NATURE | VOL 434 | 28 APRIL 2005 | www.nature.com/nature

Evolutionary biology

Animal roots and shootsMartin Jones and Mark Blaxter

DNA sequence data from neglected animal groups support acontroversial hypothesis of deep evolutionary history. Inferring thathistory using only whole-genome sequences can evidently be misleading.

Despite the comforting certainty oftextbooks and 150 years of argument,the true relationships of the major

groups (phyla) of animals remain conten-tious. In the late 1990s, a series of controver-sial papers used molecular evidence topropose a radical rearrangement of animalphyla1–3. Subsequently, analyses of whole-genome sequences from a few speciesshowed strong, apparently conclusive, sup-port for an older view4–6. Philippe et al., writ-ing in Molecular Biology and Evolution7, nowprovide evidence from expanded data setsthat supports the newer evolutionary tree,and also show why whole-genome data setscan lead phylogeneticists seriously astray.

Traditional trees group together phyla ofbilaterally symmetrical animals that possessa body cavity lined with mesodermal tissue,the coelom (for example, the human pleuralcavity), as Coelomata. Those without a truecoelom are classified as Acoelomata (nocoelom) and Pseudocoelomata (a body cavity not lined by mesoderm). We call this

Figure 1 Animals on trees: the two main hypotheses of therelationships between animalphyla. a, The Acoelomata–Pseudocoelomata–Coelomata(A–P–C) phylogeny is supportedby whole-genome studies,although complete genomes areavailable for only three animalphyla. In this scheme, flies(Arthropoda) and humans(Vertebrata) are more closelyrelated to each other as membersof the Coelomata than either is tonematodes (Pseudocoelomata)5,6.Based on morphology (there is no genome sequence), theAcoelomata are presumed to have separated from otheranimals before the divergence of the Pseudocoelomata andCoelomata. b, The newphylogeny, Lophotrochozoa–Ecdysozoa–Deuterostomia(L–E–D)1–3,7: using expressed-sequence-tag data, Philippe et al.7

were able to include 12 animalphyla. Here, flies and nematodesare both members of theprotostome group Ecdysozoa,distinct from the deuterostomehumans.

Platyhelminthes

Annelida

Mollusca

Tardigrada

Nematoda

Arthropoda

Vertebrata

Urochordata

Cephalochordata

Echinodermata

Ctenophora

Cnidaria

Choanoflagellata

Fungi

A

P

C

D

E

L

Acoelomata–Pseudocoelomata–Coelomata

Lophotrochozoa–Ecdysozoa–

Deuterostomia

a b

28.4 n&v 1075 MH 22/4/05 5:19 pm Page 1076

Nature Publishing Group© 2005

© 2005 Nature Publishing Group