S CIENCE’ S C OMPASS ● REVIEW

REVIEW: ECOLOGY Economic Reasons for Conserving Wild Nature Andrew Balmford,1* Aaron Bruner,2 Philip Cooper,3 Robert Costanza,4† Stephen Farber,5 Rhys E. Green,1,6 Martin Jenkins,7 Paul Jefferiss,6 Valma Jessamy,3 Joah Madden,1 Kat Munro,1 Norman Myers,8 Shahid Naeem,9 Jouni Paavola,3 Matthew Rayment,6 Sergio Rosendo,3 Joan Roughgarden,10 Kate Trumper,1 R. Kerry Turner3

matched estimates of the marginal values of On the eve of the World Summit on Sustainable Development, it is timely to assess goods and services delivered by a biome progress over the 10 years since its predecessor in Rio de Janeiro. Loss and degradation when relatively intact, and when converted of remaining natural habitats has continued largely unabated. However, evidence has to typical forms of human use. To ensure been accumulating that such systems generate marked economic benefits, which the we did not neglect private benefits of con- available data suggest exceed those obtained from continued habitat conversion. We version, studies were only included if they estimate that the overall benefit:cost ratio of an effective global program for the covered the most important marketed conservation of remaining wild nature is at least 100:1. goods, as well as one or more nonmarketed services delivering local social or global benefits. We cross-validated figures for in- umans benefit from wild nature (1)in per hectare for each of 17 services across 16 dividual goods and services with other es- very many ways: aesthetically and cul- biomes and then extrapolating to the globe by timates from similar places. Finally, we Hturally; via the provision of ecological multiplying by each biome’s area, the Costanza checked that the comparisons across differ- services such as climate regulation, soil forma- team estimated the aggregated annual value of ent states of a biome used the same valua- tion, and nutrient cycling; and from the direct nature’s services (updated to 2000 US$) to lie tion techniques for particular goods and harvest of wild species for food, fuel, fibers, in the range of $18 trillion to $61 trillion (1012), services. Our survey uncovered only five and pharmaceuticals (2). In the face of increas- around a rough average of ϳ$38 trillion. These examples that met all these criteria. Here, ing human pressures on the environment, these figures are of similar size to global gross na- we summarize their findings, with all fig- benefits should act as powerful incentives to tional product (GNP), but have been criticized ures expressed as net present values (NPVs, conserve nature, yet evaluating them has by some in the economic community (5–9). in 2000 US$ haϪ1), and using the discount proved difficult because they are mostly not One problem is that such macroeconomic rates considered by the authors [see Fig. 1 captured by conventional, market-based eco- extrapolations are inconsistent with microeco- and supplemental online material (10) for nomic activity and analysis. nomic theory: extrapolation from the margin to further details]. In 1997, Costanza et al. published a synthe- a global total should incorporate knowledge Two studies quantified net marginal ben- sis (3) of more than 100 attempts to value about the shape of the demand curve (3, 5–8). In efits of different human uses of tropical forest ecosystem goods and services using a range of practice, it is very likely that per-unit demand areas. Kumari compared the values obtained techniques including hedonic pricing, contin- for nonsubstitutable services escalates rapidly as from timber plus a suite of nontimber forest gent valuation, and replacement cost methods supply diminishes, so that simple grossing-up of products (NTFPs), as well as the values of (4). Using case studies to derive average values marginal values (as is also done in calculating water supply and regulation, recreation, and GNP from prices) will probably underestimate the maintenance of carbon stocks and endan- true total values. On the other hand, high local gered species, for forests under a range of 1Conservation Biology Group, Department of Zoology, values of services such as tourism may not be management regimes in Selangor, Malaysia 2 University of Cambridge, Cambridge CB2 3EJ, UK. Cen- maintained if extrapolated worldwide. In addi- (11). Compared with two methods of re- ter for Applied Biodiversity Science at Conservation In- ternational, 1919 M Street, NW, Suite 600, Washington, tion, while some policy decisions are made us- duced-impact logging, high-intensity, unsus- DC 20036, USA. 3Centre for Social and Economic Re- ing macroeconomic indicators, many others are tainable logging was associated with greater search on the Global Environment (CSERGE), School of made at the margin, and so are more appropri- private benefits through timber harvesting (at Environmental Sciences, University of East Anglia, Nor- ately informed by marginal rather than total least at high discount rates and over one wich NR4 7TJ, UK. 4Center for Environmental Science, Biology Department and Institute for Ecological Eco- valuations (9). harvesting cycle), but reduced social and nomics, University of Maryland, Box 38, Solomons, MD Another problem with the original estimate global benefits (through loss of NTFPs, flood 20688, USA. 5Graduate School of Public and Internation- is that landscapes can yield substantial (albeit protection, carbon stocks, and endangered al Affairs, University of Pittsburgh, Pittsburgh, PA 15260, rather different) flows of goods and services species). Summed together, the total econom- USA. 6The Royal Society for the Protection of Birds, The Lodge, Sandy, Bedfordshire SG19 2DL, UK. 7UN Environ- after, as well as before, conversion by humans ic value (TEV) of forest was some 14% ment Programme–World Conservation Monitoring Cen- (which is of course why people convert them). greater when placed under more sustainable tre (UNEP-WCMC), 219 Huntingdon Road, Cambridge A clearer picture of the value of retaining habitat management (at ϳ$13,000 compared with CB3 ODL, UK. 8Green College, Woodstock Road, Oxford in relatively undisturbed condition might there- $11,200 haϪ1). OX2 6HG, UK; and Upper Meadow, Old Road, Heading- fore be obtained by estimating not the gross A study from Mount Cameroon, Cam- ton, Oxford OX3 8SZ, UK. 9Department of Zoology, University of Washington, 24 Kincaid Hall, Box 351800, values of the benefits provided by natural bi- eroon, comparing low-impact logging with Seattle, WA 98195–1800, USA.10Department of Biolog- omes, but rather the difference in benefit flows more extreme land-use change again found ical Sciences, Stanford University, Stanford, CA 94305, between relatively intact and converted versions that private benefits favor conversion, this USA. of those biomes. time to small-scale agriculture (12). Howev- *To whom correspondence should be addressed.E- er, a second alternative to retaining the forest, mail: [email protected] Net Marginal Benefits conversion to oil palm and rubber planta- †Address after Sept.2002: Gund Institute of Ecolog- ical Economics, The University of Vermont, Burling- To address these concerns, we reviewed tions, in fact yielded negative private benefits ton, VT 05405, USA. more than 300 case studies, searching for once the effect of market distortions was

950 9 AUGUST 2002 VOL 297 SCIENCE www.sciencemag.org S CIENCE’ S C OMPASS removed. Social benefits from NTFPs, sedi- did sustainable fishing (15). The social ben- vices outweighs the marketed marginal ben- mentation control, and flood prevention efits of sustainable exploitation, arising from efits of conversion, often by a considerable were highest under sustainable forestry, as coastal protection and tourism, were also lost amount. Across the four biomes studied, were global benefits from carbon storage upon dynamiting reefs. As a consequence, mean losses in TEV due to conversion run and a range of option, bequest, and exis- the TEV of retaining an essentially intact at roughly one-half of the TEV of relatively tence values. Overall, the TEV of sustain- reef was almost 75% higher than that of intact systems (mean ϭ 54.9%; SE ϭ able forestry was 18% greater than that of destructive fishing (at ϳ$3300 compared 13.4%; n ϭ 4). This is certainly not to say small-scale farming (ϳ$2570 compared with $870 haϪ1). that conversion has never been economical- with $2110 haϪ1), whereas plantations had One clear message from our survey is ly beneficial; in most instances, past clear- a negative TEV. the paucity of empirical data on the central ance of forests and wetlands for prime Three other biomes yielded single studies question of the changes in delivery of agricultural land and other forms of devel- meeting our criteria. Analysis of a mangrove goods and services arising from the conver- opment probably benefited society as a system in Thailand revealed that conversion sion of natural habitats for human use. For whole. But unless the present case studies for aquaculture made or the range of ser- sense in terms of vices and biomes short-term private examined in the lit- benefits, but not once erature are extreme- external costs were ly unrepresentative factored in (13). The (and we know of no global benefits of reason why this carbon sequestration should be the case), were considered to be our synthesis indi- similar in intact and cates that at present, degraded systems. conversion of re- However, the sub- maining habitat for stantial social bene- agriculture, aquac- fits associated with ulture, or forestry the original man- often does not make grove cover—from sense from the per- timber, charcoal, spective of global NTFPs, offshore sustainability. fisheries, and storm protection—fell to Continuing Losses almost zero follow- These results there- ing conversion. Sum- fore provide a clear ming all measured and compelling eco- goods and services, nomic case, alongside the TEV of intact sociocultural and mangroves exceeded moral arguments (16– that of shrimp farm- 18), for us to strength- ing by around 70% en attempts to con- (ϳ$60,400 com- serve what remains of pared with $16,700 natural ecosystems. haϪ1). Yet, when we sum- van Vuuren and marized available es- Roy (14) reported that timates of recent draining freshwater Fig. 1. The marginal benefits of retaining and converting natural habitats, expressed as NPV (in trends in the global Ϫ1 ␦ marshes in one of 2000 US$ ha ) calculated using the discount rates ( ) and time horizons presented.Values of status of natural habi- measured goods and services delivered when habitats are relatively intact and when converted Canada’s most pro- are plotted as green and black columns, respectively.[From ( 11–15); see (10) for further tats and free-ranging ductive agricultural details.] vertebrate popula- areas yielded net pri- tions, we found that vate benefits (in large part because of substan- 10 of the largely natural biomes (including although key data are again disturbingly scarce, tial drainage subsidies). However, social bene- rangelands, temperate forests, rivers and they show that rates of conversion are high fits of retaining wetlands, arising from sustain- lakes, and most marine systems) in across most biomes (10). able hunting, angling, and trapping, greatly ex- Costanza et al. (3), we found no studies that We included in our survey any estimate of ceeded agricultural gains. Consequently, for all met all of our criteria. For the four biomes global trend in habitat cover based on a series three marsh types considered, TEVs were high- which were analyzed, only a handful of which began in 1970 or later and included a er when the wetlands remained intact, exceed- well-established ecosystem services were period of at least 5 years after the 1992 ing figures for conversion by a mean of around considered, and some particularly valuable United Nations Conference on Environment 60% (ϳ$8800 compared with $3700 haϪ1). services, such as nutrient cycling, waste and Development in Rio de Janeiro. We sup- Finally, a synthesis of economic studies treatment, and the provision of cultural val- plemented this with biome-specific indices examining Philippine reef exploitation dem- ues, were not examined at all. based on time-series data on populations of onstrated that despite high initial benefits, Despite the limited data, our review also wild vertebrates, derived from the World destructive techniques such as blast fishing suggests a second broad finding: in every Wildlife Fund (WWF) 2000 Living Planet had a far lower NPV of private benefits than case examined, the loss of nonmarketed ser- Index (LPI) and UN Food and Agricultural

www.sciencemag.org SCIENCE VOL 297 9 AUGUST 2002 951 S CIENCE’ S C OMPASS Organization (FAO) fisheries data (19, 20). standing to gain immediate private benefits should in due course result in public and For three biomes, we found two estimates from land-use change. Hence, conserving rel- private decision-makers acting to reduce con- derived by different methods and from either atively intact habitats will often require com- version of remaining habitats worldwide. largely or wholly independent data. In each pensatory mechanisms to mitigate the impact More immediately, given concerns about the case, the two estimates were remarkably sim- of private, local benefits foregone, especially practicalities of exploiting natural resources ilar (10), and so were averaged to yield single in developing countries. We see the develop- sustainably, one of the most important strat- estimates of rates of change. Data such as ment of market instruments that capture at a egies to safeguard relatively intact ecosys- these, quantifying trends in areal coverage private level the social and global values of tems is the maintenance of remaining habitats and in populations, in some ways provide a relatively undisturbed ecosystems—for in- in protected areas. This costs money, and more tractable measure of the scale of the stance, through carbon or biodiversity credits predictably, our current undervaluation of na- ongoing crisis facing nature than do estimates or through premium pricing for sustainably ture is reflected in marked underinvestment of extinction rates, which are harder to doc- harvested wild-caught fish or timber (22, in reserves. To the best of our knowledge, the ument and more difficult to link to monetary 23)—as a crucial step toward sustainability. world spends (in 2000 US$) ϳ$6.5 billion values. Third, the private benefits of conversion each year on the existing reserve network Overall, we found that five of the six biomes are often exaggerated by intervention fail- (28). Yet, half of this is spent in the United measured have experienced net losses since the ures. In the Cameroon study, for example, States alone. Globally, despite increased ex- Rio summit, with the mean rate of change forests were cleared for plantations because penditure since the Rio Summit by both across all measured biomes running at –1.2% of private benefits arising from government international institutions and private founda- per year, or –11.4% over the decade (Fig. 2) tax incentives and subsidies (12). The same is tions, available resources for existing re- (10). Hence, the capacity of natural systems to true for the Canadian wetland example (14), serves fall far short of those needed to meet deliver goods and services upon which we de- as well as for many other wetlands across the basic management objectives (29). Moreover, pend is decreasing markedly. Costing the over- United States and Europe (24). While over terrestrial and marine reserves currently cov- all value of these losses is fraught with the the short term these programs may be rational er only around 7.9% and 0.5% of Earth’s land problems of extrapolation and data availability with respect to public or private policy ob- and sea area, respectively (30, 31), well be- already discussed. Nevertheless, it is sobering jectives, over the longer term many result in low the minimum safe standard considered to calculate that if the aggregate figures of necessary for the task of maintaining wild Costanza et al. (3) and our estimate of the nature into the future (32–34). proportion of TEV lost through habitat change To estimate the resources needed to meet are roughly representative, a single year’s hab- this shortfall on land, we reworked recent itat conversion costs the human enterprise, in calculations (28, 35) of the costs of properly net terms, of the order of $250 billion that year, managing existing terrestrial protected areas and every year into the future (10). Why then is and expanding the network to cover around widespread habitat loss still happening, and 15% of land area in each region. We found what can we do about it? that a globally effective network would re- quire an approximate annual outlay of be- Reasons for Continued Conversion tween ϳ$20 billion and $28 billion [includ- In economic terms, our case studies illustrate ing payments to meet private opportunity three broad, interrelated reasons why the costs imposed by existing and new reserves, planet is continuing to lose natural ecosys- spread out over 10 and 30 years, respectively tems despite their overall benefits to society (10)]. New work derived from the costs of (21). First, there are often failures of infor- existing marine reserves suggests that an mation. For many services, there is a lack of Fig. 2. Recent global estimates of the annual equivalent initiative for the world’s seas, this valuations of their provision by natural sys- rate of change in area or the abundance of time covering 30% of total area (34, 36), tems, and particularly of changes in this pro- associated vertebrate populations for six bi- would cost at most ϳ$23 billion/year in re- vision as human impacts increase. Although omes.Note that the biomes that have declined current costs, plus ϳ$6 billion/year (over 30 this is an understandable reflection of sub- deliver valuable ecosystem services (3).*Values years) in start-up costs (10). The estimated plotted are the mean of habitat and popula- stantial technical difficulties, we believe that tion-based estimates; †Little confidence can be mean total cost of an effective, global reserve future work needs to compare delivery of attached to this value (10). program on land and at sea is some $45 multiple services across a range of competing billion/year. This sum dwarfs the current $6.5 land uses if it is to better inform policy both economic inefficiency and the erosion of billion annual reserve budget, yet could be decisions. Our examples show that even natural services. Globally, the subset of sub- readily met by redirecting less than 5% of when only a few ecosystem services are con- sidies which are both economically and eco- existing perverse subsidies (25, 26). The cru- sidered, their loss upon conversion typically logically perverse totals between $950 billion cial question is whether this is a price worth outweighs any gains in marketed benefits. and $1950 billion each year [depending on paying. Second, these findings highlight the fun- whether the hidden subsidies of external costs Although limited data make the answer damental role of market failures in driving are also factored in (25, 26)]. Identifying and imprecise, they indicate that conservation in habitat loss. In most of the cases we studied, then working to remove these distortions reserves represents a strikingly good bargain. the major benefits associated with retaining would simultaneously reduce rates of habitat We assumed that the mean proportional loss systems more or less intact are nonmarketed loss, free up public funds for investing in of value upon conversion recorded in our externalities, accruing to society at local and sustainable resource use, and save money case studies is representative of all biomes global scales. Conversion generally makes (25–27). and services, and that previous gross per- narrow economic sense, because such exter- hectare values of those services are roughly nal benefits [or related external costs, as in Costing Conservation correct (3). If these assumptions are valid, the case of the damage caused by shrimp Tackling these underlying economic prob- then our hypothetical global reserve network farming (13)] have very little impact on those lems requires action on many levels, but would ensure the delivery of goods and ser-

952 9 AUGUST 2002 VOL 297 SCIENCE www.sciencemag.org S CIENCE’ S C OMPASS vices with an annual value (net of benefits case: our relentless conversion and degra- 20.UN Food and Agricultural Organization, The State of from conversion) of between ϳ$4400 billion dation of remaining natural habitats is erod- the World Fisheries and Aquaculture (UN FAO, Rome, and $5200 billion, depending on the level of ing overall human welfare for short-term 2000). 21. R.K.Turner et al., Ecol. Econ. 35, 7 (2000). resource use permitted within protected ar- private gain. In these circumstances, retain- 22. J.Hardner, R.Rice, Sci. Am. 286 (no.5), 71 (2002). eas, and with the lower number coming from ing as much as possible of what remains of 23. S.Scherr, A.White, D.Kaimowitz, Policy Brief: Making a network entirely composed of strictly pro- wild nature through a judicious combina- Markets For Forest Communities (Forest Trends, tected reserves [for working, see (10)]. The tion of sustainable use, conservation, and, Washington, DC, and Center for International Forest- benefit:cost ratio of a reserve system meeting where necessary, compensation for result- ry Research, Bogor, Indonesia, 2002). 24. R.Turner, T.Jones, Eds., Wetlands: Market and Inter- minimum safe standards is therefore around ing opportunity costs [as called for at the vention Failures (Four Case Studies) (Earthscan, Lon- 100:1. Rio Summit (40)] makes overwhelming don, 1991). Put another way, the case studies, the economic as well as moral sense. 25. C.P.van Beers, A.P.G.de Moor, Addicted to Subsi- service values of Costanza et al. (3), or our dies: How Governments Use Your Money to Destroy reserve costs would have to be off by a the Earth and Pamper the Rich (Institute for Research References and Notes on Public Expenditure, The Hague, Netherlands, 1.By “wild nature” we mean habitat in which biodiver- factor of 100 for the reserve program en- 1999). sity, nonbiotic components, and ecosystem function- visaged to not make economic sense. We ing are sufficiently intact that the majority of eco- 26.N.Myers, J.Kent, Perverse Subsidies (Island Press, consider errors of this size to be highly system services typically derived from such a habitat Washington, DC, 2001). unlikely, because most of our assumptions are still being sustainably and reliably supplied.Our 27. S.L.Pimm et al., Science 293, 2207 (2001). are conservative [for other sensitivity anal- usage differs from other usages, such as those adopt- 28. A.James, K.J.Gaston, A.Balmford, BioScience 51,43 ed in cultural or anthropological studies.Because our (2001). yses, see (10)]. For example, in terms of the focus is on wild nature, we excluded the cropland and 29. A.N.James, M.J.B.Green, J.R.Paine, Global Review values of services, we assume that unit urban biomes when using data from table 2 of (3). of Protected Area Budgets and Staff ( WCMC, Cam- values will not increase as supply dimin- 2. G.C.Daily, Ed., Nature’s Services (Island Press, Wash- bridge, 1999) ington, DC, 1997). ishes, that nature reserves do not increase 30.International Union for Conservation of Nature and 3.R.Costanza et al., Nature 387, 253 (1997). Natural Resources (IUCN), 1997 United Nations List 4.The hedonic price method values environmental ser- the flow of services beyond their bound- of Protected Areas ( WCMC and IUCN, Cambridge, vices by comparing market prices (e.g., for residential aries [whereas some clearly can (34, 37)], UK, and Gland, Switzerland, 1998). housing) across situations which differ in the provi- and that all of a biome’s services not in- sion of those services.Contingent valuation involves 31. G.Kelleher, C.Bleakley, S.Wells, A Global Represen- cluded in the Costanza et al. survey (3) are asking respondents how much they would be pre- tative System of Marine Protected Areas ( The World worthless. On the reserve costs side, we pared to pay for a particular environmental benefit Bank, Washington, DC, 1995). (such as ensuring the survival of a species or habitat) 32.IUCN, Parks for Life: Report of the IVth World Con- assume that management costs do not de- or how much compensation they would demand for gress on National Parks and Protected Areas (IUCN, crease once local communities’ private op- its loss.The replacement cost technique quantifies Gland, Switzerland, 1993). portunity costs are met, and that expanding the cost of restoring or synthetically replacing an 33.M.E.Soule ´, M.A.Sanjayan, Science 279, 2060 ecosystem service. (1998). reserve systems yield no cost savings 5.M.Toman, Ecol. Econ. 25, 57 (1998). 34.J.Roughgarden, P.Armsworth, in Ecology: Achieve- through economies of scale or dissemina- 6. R.K.Turner, W.N.Adger, R.Brouwer, Ecol. Econ. 25, ment and Challenge, M.Press, N.Huntly, S.Levin, Eds. tion of best practice. Because all of these 61 (1998). (Blackwell Science, Oxford, 2001), pp.337–356. assumptions are biased against conserva- 7.P.Dasgupta, Human Well-Being and the Natural En- 35.A.N.James, K.J.Gaston, A.Balmford, Nature 401, vironment (Oxford Univ.Press, Oxford, 2001). 323 (1999). tion, we consider our 100:1 ratio as a low 8. P.A.L.D.Nunes, J.C.J.van den Bergh, Ecol. Econ. 39, 36.California Department of Fish and Game, NOAA’s estimate of the likely benefits of effective 203 (2001). Channel Islands National Marine Sanctuary, A Rec- conservation. 9. G.C.Daily et al., Science 289, 395 (2000). ommendation for Marine Protected Areas in the 10.For further details, see supporting online material. Channel Islands National Marine Sanctuary (California Many of the numbers reported here are unavoidably Department of Fish and Game, Santa Barbara, CA, Development and Wild Nature imprecise.To enable readers to follow our working, 2001). In advocating greatly increased funding for we generally present numbers used in calculations to 37. C.M.Roberts, J.A.Bohnsack, F.Gell, J.P.Hawkins, R. the maintenance of natural ecosystems, we three significant figures, but then round off the final Goodridge, Science 294, 1920 (2001). results in accord with their precision. 38.UN Population Division, World Population Prospects: are not arguing against development. Given 11.K.Kumari, thesis, University of East Anglia, Norwich, The 2000 Revision (UN Department of Economic and forecast increases in the human population UK (1994). Social Affairs, New York, 2001). of more than three billion by 2050 (38) and 12.G.Yaron, J. Environ. Planning Manage. 44, 85 (2001). 39.UN Development Programme (UNDP), Human Devel- 13.S.Sathirathai, Economic Valuation of Mangroves and the opment Report, 2001 (UNDP, New York, 2001). the fact that some 1.2 billion people still Roles of Local Communities in the Conservation of Nat- 40.See the text of the Convention on Biological Diversity live on less than 1 US$/day (39), develop- ural Resources: Case Study of Surat Thani, South of at www.biodiv.org. ment is clearly essential. However, current Thailand (unpublished report, Economy and Environ- 41.This review is the result of a workshop convened by ment Program for Southeast Asia, Singapore, 1998). development trajectories are self-evidently the Royal Society for the Protection of Birds (RSPB) 14. W.van Vuuren, P.Roy, Ecol. Econ. 8, 289 (1993). and sponsored by the RSPB and the UK Government’s not delivering human benefits in the way 15. A.T.White, H.P.Vogt, T.Arin, Mar. Pollut. Bull. 40, Department for Environment, Food, and Rural Affairs. that they should: income disparity world- 598 (2000). We thank N.Hockley, P.Gravestock, J.Scharlemann, 16. P.W.Taylor, Environ. Ethics 3, 197 (1981). wide is increasing and most countries are and C.Tiley for help with research, and M.Avery, R. 17.D.Ehrenfeld,in Biodiversity, E.O.Wilson, Ed.(Nation- Cowling, G.Daily, A.Gammell, D.Gibbons, J.Mc- not on track to meet the United Nations’ al Academy Press, Washington, DC, 1988), pp.212– Neely, and C.Roberts for stimulating discussions. goals for human development and poverty 216. 18.B.Norton, Environ. Values 1, 97 (1992). Supporting Online Material eradication by 2015 (39). Our findings 19.J.Loh et al., WWF Living Planet Report 2000 ( World www.sciencemag.org/cgi/content/full/1073947/DC1 show one compelling reason why this is the Wildlife Fund, Gland, Switzerland, 2000). SOM Text

www.sciencemag.org SCIENCE VOL 297 9 AUGUST 2002 953