news and views

These authors analysed the skeletal understanding of the migration history of populations. Slowly, we are realizing that remains of 33 modern Amerindians from different American populations and what the ancestry of the Americas is as complex early historic times, excavated from the tip kind of evolutionary forces might have and as difficult to trace as that of other of the Baja peninsula in Mexico. Surpris- influenced them10. human lineages around the world. ingly,the craniofacial features of these Baja Given these limitations, the findings of Tom D. Dillehay is in the Department of Amerindians show closer affinity to the González-José et al. do not allow us to draw Anthropology, University of Kentucky, Lexington, Palaeoamerican skulls than to other mod- firm conclusions about the relationship Kentucky 40506, USA. ern Amerindian remains. The Baja Amer- between the ancient Palaeoamericans and e-mail: [email protected] indian and Palaeoamerican skulls have the later Baja Amerindians. But the impor- 1. Dixon, E. J. Quat. Sci. Rev. 20, 277–299 (2001). similar long and narrow braincases and tance of this and other studies7,11 is that they 2. Dalton, R. Nature 422, 10–12 (2002). relatively short, narrow faces, implying a suggest a different view of the origins and 3. Neves, W. A. & Pucciarelli, H. M. J. Hum. Evol. 21, 261–273 (1991). common ancestry with the ancient inhabi- interactions of early human populations in 4. Dillehay,T.D.The Settlement of the Americas: A New Prehistory tants of south Asia and the Pacific Rim. the Americas.What we really want to know is (Basic Books, New York, 2001). González-José et al. confirm that modern what took place within and between these 5. González-José, R. et al. Nature 425, 62–65 (2003). Amerindian skulls from other areas are populations, how they changed over time, 6. Turner, C. G. Mem. Calif. Acad. Sci. 27, 123–158 (2002). 7. Meltzer,D.J.Mem. Calif. Acad. Sci 27, 27–58 (2002). similar to ancient northeast Asian remains. and how quickly they changed. These issues 8. Lahr, M. M. The Evolution of Modern Human Diversity Their new data add to accumulating can be resolved only by obtaining more (Cambridge Univ. Press, 1998). evidence of morphological differences skeletal data11 and by combining them 9. Neves, W. A. et al. Homo 50, 258–263 (1999). between early humans from different areas with regional archaeological records, which 10.Steele, D. G. & Powell, J. F. Mem. Calif. Acad. Sci. 27, 8,9 93–122 (2002). of the Americas . should provide information on the social 11.Jantz, R. E. & Owsley, D. W. Nevada Hist. Soc. Q. 40, The authors consider several potential and cultural histories of the different 62–84 (1997). explanations to account for the presence of Palaeoamerican traits in the Baja Amer- indian skulls, but they suggest that the best Earth science explanation is that the Palaeoamericans were the direct ancestors of the Baja Amerindians.After the Ice Age,the increased Just add water aridity could have geographically isolated Albrecht W. Hofmann the founding Palaeoamerican population in the Baja area, and limited its gene flow with A new model could explain why Earth’s upper mantle is depleted of other modern Amerindian groups. many trace elements. At a certain depth, minerals might release water, Do the new findings tell us anything creating a molten filter that traps trace elements in the mantle beneath. more about when the first humans arrived in the Americas? The authors do not fully andwiched between Earth’s thin crust gases and other ‘incompatible’ elements. discuss the chronological implications of and its metallic core lies a layer of The primitive layer serves as a reservoir for their work, but their interpretation of Spressurized rock at high temperature these elements (which were depleted from shared ancestry between the Palaeoameri- — the mantle. Convection in this layer the upper mantle when Earth’s crust was cans and the Baja Amerindians might best drives plate tectonics and sea-floor spread- formed) and it is occasionally sampled by fit a model of Palaeoamerican arrival about ing, but we know little about the pattern of deep-mantle plumes (Fig.1a). 11,000–12,000 years ago. There is no direct circulation. Indeed, current thinking about Over the past several years, however, seis- evidence to support this view, but if the mantle dynamics is in a state of turmoil. As mic tomography has given us increasingly Palaeoamericans had arrived 15,000 years we cannot observe convection directly, we detailed images of apparently cold ‘slabs’ ago or earlier, the Baja population would must piece together indirect evidence from (characterized by fast seismic velocities) have remained isolated for much longer. seismology, geochemistry, mineral physics, descending into the deep mantle right This seems unlikely, given the rate of fluid dynamics and numerical simulations through the 660-km boundary, effectively population growth and movement that of convection. But the evidence is contradic- cutting to shreds the simple picture of the probably occurred after initial coloniza- tory and has led to at least two conflicting mantle convecting in two nearly isolated tion and then after the Ice Age when the views about mantle movement. On page 39 layers. If cold ‘slabs’ descend into the deep climate warmed. of this issue, Bercovici and Karato1 propose mantle, there must be a corresponding But could the similarities between the a new model that might resolve this conflict. upward flow of deep-mantle material to ancient Palaeoamericans and the later Baja They suggest that water dissolved in the shallow levels (Fig. 1b). No matter what Amerindians instead reflect the influence of mantle might create a thin layer of melt at a specific form the exchange across the 660-km other evolutionary forces, such as gene flow depth of 400 km, causing an unexpected boundary takes, in this ‘whole-mantle’ or natural selection and convergent adapta- pattern of circulation in the mantle. model of mantle convection,it would within tion of different populations to similar local The two conflicting models for mantle a few hundred million years destroy any environments? Answering this question will convection (Fig. 1a, b, overleaf) are usually compositional layering that had possibly depend upon finding more isolated prehis- described as ‘layered’ convection (suppor- been inherited from early in Earth’s history. toric populations showing ancient Palaeo- ted by geochemists) and ‘whole-mantle’ Meanwhile, the geochemical arguments american traits, and then establishing convection (supported by seismologists). for a separate deep reservoir have not disap- whether parallel evolutionary forces were Geochemists have long insisted on the two- peared. Primordial noble gases are still acting on them and whether they were layered model, in which the mantle consists preferentially associated with ‘hot spots’2,at derived from a single ancestry. But this will of a relatively primitive layer below a depth of least some of which seem to come from deep- be a difficult task. Human remains from the 660 km — containing primordial noble mantle plumes3. And much of the upper end of the Ice Age are scarce and often frag- gases, trapped 4.5 billion years ago when the mantle remains highly depleted of incom- mentary, so we have only a vague notion Earth formed — and an upper layer that is patible trace elements, including the heat- of the skeletal characteristics of the ancient highly depleted of heat-producing elements producing thorium,uranium and potassium Palaeoamericans. And we have a poor (uranium, thorium and potassium), noble — also suggesting the presence of a less

24 © 2003 Nature Publishing Group NATURE | VOL 425 | 4 SEPTEMBER 2003 | www.nature.com/nature news and views depleted reservoir deep within the mantle4. Can these conflicting data be reconciled? c Some scientists have proposed that a rela- Depleted tively small, seismically ‘invisible’ primitive ab upper Ocean Continent Ocean Continent reservoir, which is slightly denser than ordi- mantle nary mantle material, is trapped in the deep 0 Olivine 400 km Melt- Wadsleyite + H O layer mantle5,6. Others suggest that the seismically 2 410 km filter

Plume anomalous, so-called D layer just above the 660 Slow Transition core–mantle boundary can satisfy the geo- rising zone 7 mantle chemical requirements . Yet another theory 660 km is that the geochemical heterogeneities are dispersed throughout the mantle but are preferentially sampled by hot-spot magmas, Depth (km) 2,900 which therefore deliver a biased record of Liquid outer core their source chemistry8. Bercovici and Karato1 propose a new model for mantle convection that might Figure 1 Models of mantle circulation. a, Geochemists support a two-layered model in which the upper reconcile the geochemists and the seismolo- layer of mantle, depleted in ‘incompatible’ elements, is separated from the primitive lower mantle by gists. Experimental work has shown that, a boundary at a depth of 660 km. b, Seismologists disagree. They support a ‘whole-mantle’ model of at high pressure, the mantle minerals circulation with exchange between subducting crust and upwelling plumes. c, Bercovici and Karato1 wadsleyite and ringwoodite have surprising- now propose a model that might satisfy both camps. They suggest that at depths of 410 km the mineral ly high solubilities for water9. High-pressure wadsleyite releases water and is transformed into olivine. This creates a thin layer of molten silicate forms of these minerals occur in the depth that acts as a filter, removing ‘incompatible’ trace elements from slowly uprising mantle. range of 410–660 km — the so-called mantle transition zone — and Bercovici and Karato but I am not so sure. The difficulty comes Another concern is that the model postulate that this zone has a high water in reconciling Bercovici and Karato’s hypo- depends on the assumption that the mantle content (0.2–2.0% by weight). They suggest thesis with our view of how the mid-ocean transition zone contains enough water to that when material in the transition zone ridge forms. This part of Earth’s crust is cause the postulated melting.There is at least rises through slow convective motion to thought to be derived from the upper man- circumstantial evidence that the mantle is depths of less than 410 km, wadsleyite is tle, whereas ocean islands are made from rather severely dehydrated during subduc- transformed into the mineral olivine and deeper mantle plumes. These different tion12, and estimates based on trace-element liberates water (olivine has a much lower source materials can be traced by the com- ratios13 limit the water content of the primi- solubility for water than wadsleyite). position of strontium isotopes in the basalt tive silicate mantle to less than about 0.05% Adding water to almost any rock will rock — mid-ocean ridge basalts have lower — much lower than Bercovici and Karato’s lower its melting point by several hundred 87Sr/86Sr ratios than ocean island basalts.The estimate of 0.1–1.5% for the transition zone. degrees. The authors suggest that this ‘dehy- 87Sr isotope is created by the very slow decay But these measurements may not necessarily dration melting’ might create a thin (about of 87Rb (with a half-life of about 47 billion be applicable to the transition zone in the 10-km) layer of molten silicate just above the years), so the isotopic differences between new model. 410-km level,which would retain most of the the upper mantle and the deeper mantle Bercovici and Karato’s hypothesis is water as well as all those incompatible trace take a very long time to develop, typically intriguing,but wouldn’t such a mid-mantle elements (including uranium and thorium) 1–2 billion years. melt layer have been detected long ago by that do not easily fit into the crystal struc- In the Bercovici–Karato model,the upper seismologists? Perhaps it was14,15,but the tures of the available solid minerals.The melt mantle is continuously replenished by slowly evidence is patchy. Features less than 10 km layer is trapped at 410 km, because melt is rising deep-mantle material. The melt layer thick are easily missed in seismic records denser than solid at very high pressures, at 410 km removes water and incompatible and will have to be carefully searched for but it is ultimately entrained, refrozen and elements, but this does not change the iso- in the future. returned to the transition zone by descend- topic composition of the rising dehydrated Albrecht W. Hofmann is at the Max-Planck-Institut ing cold, subducted slabs. This return flow solid. According to the isotopic differences, für Chemie, Postfach 3060, 55020 Mainz, keeps the thickness of the molten silicate to fit Bercovici and Karato’s model, the deep Germany. approximately constant. So the melt layer mantle rises at a very slow speed of about e-mail: [email protected] acts as a filter,removing elements from rising 1 mm per year or even less. So it would take 1. Bercovici, D. & Karato, S. Nature 425, 39–44 (2003). mantle material and keeping the upper layer more than 400 million years to traverse the 2. Graham, D. W. Rev. Mineral. Geochem. 47, 247–317 (2002). in a depleted state (Fig. 1c). Deep-mantle 400-km distance towards the melting region 3. Montelli, R. et al. Geophys. J. Int. (in the press). 4. Hofmann, A. W. Nature 385, 219–229 (1997). plumes are exempt from this filtering process beneath ocean ridges. But upwelling rates in 5. Davaille, A. Nature 402, 756–760 (1999). because their higher temperature lowers the immediate vicinity of mid-ocean ridges 6. Kellogg, L. H. et al. Science 283, 1881–1884 (1999). the solubility for water in wadsleyite, are significantly greater, and this would tend 7. Tolstikhin, I. N. & Hofmann, A. W. Geochim. Cosmochim. Acta but increases solubility in olivine, thus to minimize the time available for isotopic 66, A779 (2002). 8. Morgan, J. P. & Morgan, W. J. Earth Planet. Sci. Lett. 170, preventing the formation of a melt layer at a differences to develop. Still, opinions differ 215–239 (1999). depth of 410 km. rather widely as to how great the isotopic 9. Williams, Q. & Hemley, R. J. Annu. Rev. Earth Planet. Sci. 29, The new model has not yet been tested by difference actually is between average 365–418 (2001). numerical simulations of mantle convec- mantle samples from mid-ocean-ridge 10.Hofmann, A. W. & White, W. M. Earth Planet. Sci. Lett. 57, 421–436 (1982). tion. But this might prove to be difficult or basalts and from plumes. Moreover, it is 11.Christensen, U. R. & Hofmann, A. W. J. Geophys. Res. 99, impossible to do in the near future because likely that plume-source material does not 19867–19884 (1994). the melt layer is very thin and its mechanical represent average lower mantle;instead,its 12.Dixon, J. E. et al. Nature 420, 385–389 (2002). behaviour is complex. But does the model composition is probably biased towards 13.Michael, P. J. Earth Planet. Sci. Lett. 131, 301–320 (1995). 14.Revenaugh, J. & Sipkin, S. Nature 369, 474–476 (1994). satisfy the geochemical constraints? Many segregated subducted former oceanic 15.Vinnik, L. & Farra, V. Geophys. Res. Lett. 29, 10,11 isotope geochemists will say that it does not, crust . doi: 10.1029/2001GL014064 (2002).

25 NATURE | VOL 425 | 4 SEPTEMBER 2003 | www.nature.com/nature © 2003 Nature Publishing Group