Page 46

The Sixth Extinction:
How Large, Where, and When?

Stuart L. Pimm
Thomas M. Brooks
Department of Ecology and Evolutionary Biology University of Tennessee, Knoxville, TN 37996-1610

The scientific consensus is that if current rates of species extinction continue, the fraction of species lost will be comparable to that of the five major extinction events in Earth's geological past (Leakey and Lewin 1996). Unlike the past episode—the famous one exterminated the dinosaurs—this sixth extinction is driven by the dominance of one species, humans (Ehrlich and Ehrlich 1981). The powerful ethical (Norton 1988) and economic (Costanza and others 1996) reasons why we should prevent this scenario are well known (Myers 2000). Less clear are the details. How many species will we lose? Will these losses occur across the globe, or are some areas more vulnerable than others? How quickly will species disappear: do we have years, decades, or centuries to mitigate our current actions?

Those are the questions we will address here. They are circumscribed in one obvious way: we count species, in part, because it is easy to do so. How, then, might our answers apply to other levels of biodiversity? The utility of the term biodiversity stems from the recognition that there is variety in life between individuals in a given population, between populations of a given species, and between species (Wilson 1992, 2000).

Killing individuals does not necessarily kill a population, exterminating a population does not necessarily eliminate the species, the species its genus, and so on. What happens when we reverse this logic (Raup 1979)? When we exterminate, for example, 10% of all species, we will likely exterminate far more than 10% of all populations. Some species will survive our depredations but with severely pruned populations. As Hughes and others (1997, 2000) point out, many important justifications for protecting biodiversity emerge from populations, not species.

Page 47

In what follows, whatever statistics we estimate for species must be substantial underestimates of the effects on populations.

Some species have many populations, others few. Similarly, some genera have many species, others few; and so on through the taxonomic hierarchy of families, orders, and classes. This hierarchy is the sometimes imperfect surrogates for evolutionary lineages of increasing depth. Random species kills often fall on genera with many species, so generic diversity will survive. A benevolent species killer might select some of the buntings (Emberiza), sandpipers (Calidris), and greenbuls (Phyllastrephus). This action might remove no genera, but only species (whose loss would be mourned only by us connoisseurs of subtle differences in their shades of brown and green). Humanity, however, can be malevolent. Elsewhere, we show that we have already lost more genera of birds and mammals than one expects to lose by chance on the basis of random species losses (Russell and others 1998). So, perversely, the impacts we estimate for species also underestimate the impacts we might expect on the diversity of higher taxonomic categories.

Species Losses Past and Present

The vast majority of the species that have ever lived are now extinct. So the question “How many species are going extinct?” has to be rephrased: “How much faster are species going extinct than one would expect?” The contrast is one of rate.

Thirteen studies of the fossil record show that species persist for one to a few million years (May and others 1995). We know the names of about 1.55 million species (May, this volume), so each year we would expect one or at most a few species to expire. Within small subsets of species, we would expect to wait longer to see just one extinction—about a century for the 10,000 species of birds, for example. Calculating extinction rates as ‘extinctions per species per year’ provides a convenient frame of reference for calculating human impact (Pimm and others 1995). We know the names of only a small fraction of the planet's species (May 2000), and so by design, this measure does not depend on our knowing them all.

The fossil-record estimate of roughly a million-year life span for a species is suspect in two obvious ways. First, most kinds of species are absent from the record while invertebrates with hard shells (mollusks and brachiopods) dominate. Second, rare species are likely to be missed entirely (McKinney and others 1996). So how typical is this estimate of the rare vertebrates that form the core of our subsequent discussion? An important clue comes from the constraint that natural extinction rates cannot greatly exceed natural speciation rates (Pimm and others 1995). If it were otherwise, there would be no species in the group in question for us to observe.

The common model of speciation assumes an interbreeding population, and then a barrier that splits it allowing the daughter populations to diverge evolutionarily. Taxonomists pass judgment on whether this divergence is sufficient to have formed species. Alternatively, the barrier might later dissolve and the two populations, by not interbreeding, unequivocally demonstrate their distinctiveness. The distinctiveness of two populations in the latter alternative (sympatry) informs

Page 48

the taxonomic judgments about the former (allopatry). In this model, barriers make species and geological knowledge allows us to date the barriers (Rosenzweig 1995). On the average, species-making barriers should form half a species's lifetime in the past, for some species are near their births, and others are near their deaths.

In North America, the presence of many pairs of similar bird species in forests on either side of the central prairies suggests the Late Pleistocene glaciation only 10,000 years ago as the species-making barrier. This high speciation rate might have been a fortuitous baby boom in species, with current high extinction rates a natural pruning of evolutionary exuberance. In fact, the suggestion itself is wrong. Klicka and Zink (1997) use molecular data to show that for 35 such species pairs the average divergence time is 2.45 million years. Increasing numbers of similar studies will likely flesh out many other details, but overall they support the million-year life span as a conservative estimate for species in general.

Humanity's Impact on Species' Lifetimes

The expectation that one should wait a century to observe an extinction among a sample of 10,000 species is rudely rejected by birds. In recent history (the last 2,000 years), the 10,000 bird species have suffered an average of one or a few extinctions each year (Steadman 1997). Humanity has decreased the average species lifetime and consequently increased the extinction rate by a factor of several hundred (Pimm and others 1995). We know birds well, and the details are informative.

Most of the bird extinctions have been in islands in the Pacific (Steadman 1995). The extinctions represented by stuffed skins in museums, collected within the last century or so, are a small fraction of the total. We know of many more species only as bones from archaeological samples. These species persist up to, but not through, the layers indicating the island's human colonization. The archaeological samples are inevitably incomplete. On the basis of what fraction of today's species the samples include, we estimate that they have found only half the extinct species (Pimm and others 1994). In addition, few of the 700 Pacific islands large enough and isolated enough to host unique species have been explored by archaeologists. Once again, we must correct the body count to reflect the incompleteness of the sampling. Statistical corrections from known species and surveyed islands suggest that the Polynesian colonization of the Pacific exterminated at least 1,000 species of birds. Locally, as in Hawai'i, the Polynesians exterminated over 90% of bird faunas (Pimm and others 1994).

Conclusion 1. Over the last few thousand years, humans have eliminated over 10% of the world's bird species and locally over 90% of them. Double-digit extinction percentages are part of our history, not merely a prediction about our future.

Page 49

The obvious question is whether birds are exceptionally wimpy. They are well-known and so provide unusual details, but are they just extinction-prone? The answer is an emphatic no. Those who argue, like Simon (1986), that the only current extinctions among the 1.55 million named species are a few species of birds and mammals each year are simply ignorant of the facts. The data prove that extinctions are much more comprehensive. The examples are extraordinarily diverse, including animals and plants, invertebrate and vertebrate animals, species on islands and those on continents, desert species and rain-forest species, and aquatic and terrestrial species (Pimm and others 1995).

Statisticians know that their craft depends on samples and the inferences made from them. Reliable inferences require that samples be representative. Reading the list of examples in the previous paragraph, it is hard to imagine a more representative selection of samples. (Those who deny the generality of high extinction rates frequently use economic statistics based on samples—sometimes very small samples—of the numbers in question. The uncritical faith in statistics in one field and the denial of their existence in others is incongruous.) The high rates of extinction in so many different groups lead to our second conclusion.

Conclusion 2. Surveys of many groups of plants and animals uncover global rates of extinction at least several hundred times the rate expected on the basis of the geological record. These groups are diverse in their natural histories and evolutionary origins. With high statistical confidence, they are typical of the many groups of plants and animals about which we know too little to document their extinction.

That is an ecologically surprising conclusion. Would it not be more reasonable for some kinds of species, such as birds versus beetles, or some kinds of places, such as forests versus deserts, to display concentrations of extinctions? Certainly, islands are home to many of the groups of species that are endangered. Yet extinction centers are found on continents, too, so there is nothing unique about islands. There are some differences between taxa. In North America, The Nature Conservancy has surveyed 18 groups of animals and plants to calculate the fraction that are on the verge of extinction (TNC 1996). Only butterflies are less vulnerable than birds. Proportionally, freshwater fishes, amphibians, crayfish, and freshwater mussels have 3–7 times more species at risk.

Despite these differences in places and taxa, we find high extinction rates in almost every group of species and in almost every kind of place. This “ecologically surprising conclusion” suggests that general ecological principles that work across all groups lead to substantial fractions of their constituent species becoming vulnerable to extinction. We suggest that there are three such principles:

• Many species have very small range sizes, relative to the average range size. (In other words, the statistical distribution of range size is highly right-skewed). Among birds, 30% of all terrestrial species have ranges smaller than 50,000 km2 (Stattersfield and others 1998)—an area half the size of Tennessee—whereas the average size is some 40 times larger.

Page 50

• Species that have small ranges are typically less abundant within those ranges than are species that have large ranges (Gaston 1994).

• Species with small ranges are often geographically concentrated (ICBP 1992). We call these areas of concentration hot spots of endemism (Myers 1988; Reid 1998).

Low numbers make a species vulnerable to disasters. So, too, do small geographic ranges; human impacts destroy habitats locally (Manne and others 1999). Nature has put her eggs—species with small geographic ranges that typically have relatively low densities—into a few baskets, the hot spots. The pattern is general because the ecological principles that generate it are ubiquitous. These features, in turn, might be derived from deeper ecological causes, and indeed, ecologists seek such explanations. Whatever the underlying causes of the patterns, their consequences are obvious.

Conclusion 3. Many species are rare and local and so at particular risk from humanity's impact. Such species are not spread evenly; extinctions will be geographically clumped, like broken eggs in a dropped basket.

The aggregation of range-restricted species is the feature common to all the examples of high extinction rates listed earlier. Fish in East African lakes, freshwater mussels in the Mississippi drainage, mammals in Australia, flowering plants in the Cape Province of South Africa, and just about everything on oceanic islands—are all examples of aggregations of range-restricted species and very high extinction rates. Where there are not aggregations of range-restricted species, extinction rates will be low. There have been few bird extinctions in eastern North America—an example to which we will return.

There could be two classes of exceptions to the common pattern: aggregations of range-restricted species that do not suffer high extinction rates, and extinctions of widely distributed species. Salamanders constitute an example of the first. Some 20% of the world's salamanders are found in the mountains of the eastern United States, but few are threatened. The reason could be simply that the nature of the terrain protected it from logging or that salamanders can survive well in the moist, deciduous second growth typical of the region. Some species aggregations are just lucky. Other amphibians illustrate the second class: species appear to be in decline worldwide (Berger and others 1998).

Such exceptions apart, the concentration of extinctions in hot spots for species with small ranges has two consequences for policies to prevent extinction:

Policy consequence 1. The history of areas that do not have concentrations of range-restricted species (cold spots) does not inform us in any simple way about the likely fate of concentrations (hot spots).

Eastern North America is an illustration. After European settlement in the 1600s, most of the forest was cleared, although not simultaneously. There were few extinctions—only four species of birds, for example. That does not mean that

Page 51

clearing other comparable areas elsewhere will have correspondingly small impacts. For birds, eastern North America is a cold spot: of its 160 forest species, only 25 are found only there (Pimm and Askins 1995). Clearing a roughly equal area of forest in insular Southeast Asia would exterminate nearly 600 species of birds (Brooks and others 1997).

Policy consequence 2. The fraction of species that will go extinct will depend critically on whether we lose or protect aggregations of range-restricted species.

The good news is that vulnerable species are concentrated, so saving them requires relatively little area. The bad news is that many of these areas have rapidly growing human populations and are in less-developed countries that have sparse resources to protect them (Balmford and Long 1995). Combining those statements leads to

Conclusion 4. How large the sixth extinction will be is still a matter of human choice, not of predestination.

Where are the Hot Spots of Endemism?

Myers informally identified 18 hot spots (Myers 1988, 1990). More recently, there have been many efforts worldwide to identify these key areas formally. There are now sophisticated algorithms for picking the smallest subset of locations that encompass all or some specified fraction of the species that one must protect (Pressey and others 1993). Some of these provide important exercises in method development (Csuti and others 1997). Others, such as the work of Lombard (1995) in the Cape Province of South Africa, inform practical decisions about where to establish nature reserves in this extraordinarily rich (and threatened) plant hot spot (Pimm and Lawton 1998).

There are several limitations. The most severe is that only a small fraction of the planet's species are named (May 2000), and we have range distributions for only a tiny fraction of them. Stork (1997) found that the great majority of insects are known from only one specimen each, and so only one location. Worse, there are complications even for the species we do know well.

Areas rich in species are typically not those rich in range-restricted species (Prendegast and others 1993; Curnutt and others 1993). Equivalently, areas that have similar numbers of species can differ greatly in their numbers of range-restricted species. The Hawaiian Islands, eastern North America, and Great Britain have broadly similar numbers of forest-living bird species (about 150); the percentages of species restricted to those areas are 100%, 17%, and less than 1% respectively. Nor are areas rich in range-restricted species in one group always rich in another: eastern North America is a hot spot for salamanders but not for birds. Recent work in Uganda suggests that this lack of correspondence might not matter, because key areas for each species group still represent other groups

Page 52

remarkably well (Howard and others 1998). Nevertheless, we have much to learn about the geography of hot spots (Pimm and Lawton 1998).

Policy consequence 3. We cannot protect hot spots if we do not know where they are.

When comprehensive data are available, the algorithms to select areas for protection typically choose samples widely scattered across the study region. The size of the samples is set by the resolution of the range maps and is usually arbitrary. (An exception is Lombard's work [above] where the areas are set by the mosaic of different land uses and ownership.) Obviously, we can apply such methods to an ever-diminishing spatial scale. Two individuals of every species require remarkably little space. Thus, even with comprehensive data on species ranges, we must ask the ecological question: How much space must be set aside to protect species?

The question has a political answer: worldwide, about 5% of the land has been set aside for protection. The allocation of this to small and large areas also has a political answer. In the Americas, from Florida (US) southward through Mexico, Central America, and South America, only 21 national parks are larger than 10,000 km2—roughly a square of 1 degree of latitude and longitude on each side (Mayer and Pimm 1998).

Policy consequence 4. Even if we know where to protect species, we must determine how much area is necessary.

How Much Area for How Many Species?

Global extinction is driven by the fate of the hot spots (Myers 1988). As the area of these hot spots shrinks because of habitat loss and fragmentation, how many species do we lose? One way to approach the question is simply to count the numbers of threatened and endangered species. That is the approach taken by the “Red Data Books” (Baillie and Groombridge 1996). But only for a few well-known groups of species is such information available (Pimm and others 1995). Fortunately, we can estimate losses of species by considering the amount of habitat that is being destroyed.

Exhaustive surveys of species in progressively larger areas of continuous habitat show that the larger the area surveyed is, the more species there will be. These surveys make it possible to deduce a mathematical relationship between species and area. Surveys of archipelagoes show the same relationship but with fewer species for an area of given size than in areas of continuous habitat. The derivation of a power function from first principles by Preston (1962) has led to the nearly universal acceptance of a form S = cAz for this relationship, where S is species number, A is area, and c and z are constants (Rosenzweig 1995). Typical values of z for increasingly large subsets of continuous habitat are about 0.15; values for areas between islands within an archipelago are about 0.25 (Rosenzweig 1995).

Page 53

We can use this relationship to derive mathematically the species loss after fragmentation of a once-continuous habitat area, Atotal, initially holding Stotal species that are found only in this habitat (figure 1A). When we destroy the habitat,

Figure 1
Typical species-area relationships. Larger areas (A) have more species than
smaller ones (B, C), and areas that have been long isolated—such as islands—have
proportionately fewer species (C) than do equal sized areas that are nested within
continuous habitat (B).

Page 54

leaving only an archipelago of fragments, the z value necessary to estimate the number of species that survive, Sfragment,, in a fragment of area Afragment is the “archipelago value” of 0.25 (figure 1C). In graphical terms, our number of species extinctions is represented by the drop from Stotal (figure 1A) to Sfragment (figure 1C).

We have calibrated this approach for three areas. For eastern North America, a region that has long been deforested and is a cold spot for bird diversity, we find that this recipe exactly predicts the number of bird extinctions (four) that have occurred (Pimm and Askins 1995). For two recently deforested hot spots, insular Southeast Asia (Brooks and others 1997) and the Atlantic forests of South America (Brooks and Balmford 1996), the recipe accurately predicts the numbers of bird species threatened with extinction in the medium term. The recipe is silent, however, about how long the still-surviving but probably doomed species will last. That leads us to our last question.

How Long Does it Take to Lose Species?

There are many ways to answer the question. Extensive modern experience shows that populations numbering in the thousands have risks of extinction observable within human lifetimes. Populations numbering in the tens and hundreds frequently become extinct. Computer and mathematical models provide the theoretical underpinnings of such observations and inform the management of particular species (Pimm 1991).

An entirely different tack comes from looking at the large national parks that are the flagships of their nations' conservation policies. Our experience in advising management about the endangered species in one of these, Everglades National Park in Florida, is that even at this scale, protecting such species requires constant vigilance (Mayer and Pimm 1998). Similar results across similarly large areas have been found elsewhere (Brash 1987; Daniels and others 1990; Diamond 1972; Newmark 1996; Soulé and others 1979; Terborgh 1975).

Between the management of particular endangered species and that of large parks are studies of fragmented habitats. It is on these that we shall concentrate. We can estimate the time that it takes for small patches of natural habitat to lose their species in at least two ways.

The simple way is to find a freshly isolated fragment and then to watch and wait. That is the approach being taken by the exemplary Biological Dynamics of Forest Fragments project in the Brazilian Amazon (Bierregaard and others 1992) and studies of islands isolated by rising waters after the damming of the Lago Gurí, Venezuela (Terborgh and others 1997). The only problem with the approach is that we might not have time to watch and wait. We would like the answers now, not in the future when it is too late to use them.

An alternative approach is to study old fragments of various ages. This approach relies on serendipity; but given the near ubiquity of habitat fragmentation, some fragments, somewhere, will surely provide something close to an ideal experiment. It is also less direct.

Page 55

Historical collections can provide lists of the total species pool, Stotal, in the prefragmentation area, Atotal (figure 1A). Such records rarely distinguish the particular subset that now remains as a fragment, from the once-continuous habitat that surrounded it. We can estimate the number of species in such prefragmentation subsets, Soriginal, using the species-area relationship for the “continuous habitat value” of z = 0.15 (figure 1B). Similarly, we can estimate the number that will eventually remain after fragmentation, Sfragment, using the ‘archipelago value’ of z= 0.25 (figure 1C). Here, we are interested in the species loss from a particular subset, Afragment, (“local extinctions” or extirpations), rather than from the entire original area, Atotal (“global extinctions”). Graphically, the eventual species loss is represented by the drop from Soriginal (figure 1A) to Sfragment (figure 1C).

Addressing the issue of “how long” requires more information. We can determine through survey work the number of species surviving now, Snow, at any time t after fragmentation. This value should be somewhere between Soriginal and the final number, Sfragment. From those numbers, we can derive a ‘relaxation index’ (I), a ratio of the proportion of extinctions yet to occur after time t to the proportion that will eventually occur:

Immediately after fragmentation, I will equal 1.0, and it should eventually decline to zero. The final step is to assume a particular form for how I declines with time. As a first approximation, we assume that the decline in species is exponential (Diamond 1972) and therefore that we can characterize it by a fixed time to lose half its species (figure 2). (If the fragment loses 50% of its species in x years, it will lose half of what remains—25% of the total—in another x years, half of what remains (12.5%) of the total in the next x years, and so on.) Thus,

where k is a decay constant and t is the time since the fragment was isolated. When I = 0.5, the fragment has lost half the species that it stands to lose, so t equals the half-life.

Elsewhere, we present data on birds in five rainforest fragments near Kakamega, western Kenya, that we collected over 1996 (Brooks and others 1999). Those data are the results of 8 months of bird surveys through mist-netting, spot counts, and extensive observation; a thorough literature review; an assessment of large quantities of forest-cover data in the form of aerial photographs dating back to 1948, satellite imagery, and anecdotal reports; and a survey of the historical bird specimens in most major museums. For each of the five fragments, we know Atotal, Afragment, Stotal, Snow, and t. From these we can estimate Soriginal and Sfragment and then use equation 1 to estimate I.

In figure 3 we plot the proportion of species still expected to be lost, I, against their times since isolation, t. If the declines in species numbers are all exponential with exactly the same half-lives, these points would fall along the same curve.

Page 56

Figure 2
Exponential loss of species from fragmented forest. The number of species in an area
of once continuous forest (Soriginal) declines through the number (Snow) at the time (t) when
a survey was conducted to the number that will eventually survive (Sfragment) We can estimate
Soriginal and Sfragment using the method of figure 1. Because the decay is exponential,
we can characterize it by a half-life, the time taken to lose 50% of the species that will
eventually be lost.

To a rough approximation, they do so, and their calculated half-times are all broadly similar at between 25 and 75 years—around 50 years.

The long technical details have a short conclusion. Of the species that fragments are going to lose, they lose half in about 50 years. In a century, they will lose 75% of those species.

Conclusion 5. Isolated habitat fragments (certainly fragments of tropical rain forest) will have suffered most of their extinctions by 100 years after isolation.

Page 57

How do these results compare with other studies? Historical data on forest fragments are rare (Laurance and others 1997; Turner 1996), but a few studies do provide dated information on bird communities in fragmented tropical forests (Aleixo and Vielliard 1995; Christiansen and Pitter 1997; Corlett and Turner 1997; Diamond and others 1987; Kattan and others 1994; Robinson 1999; Renjifo 1998). Table 1 summarizes the data from those studies, giving the time between the historical and contemporary surveys (t) and the historical (Shistorical) and contemporary (Snow) numbers of bird species. Assuming a half-life of 50 years, we predict the future equilibrium numbers of species (Sfragment). Future resurveys of the sites could provide a third point in time along the relaxation curve (figure 2) and therefore test the predictions. Their value now is in suggesting how many more species the sites stand to lose.

How do our results extend globally? We know that over 10% of the world's roughly 10,000 bird species are threatened with extinction, with habitat loss and fragmentation as the main causes (Collar and others 1994). We therefore predict that about 500 of these bird species will go extinct in the next 50 years, producing an extinction rate of 1,000 extinctions per million species per year. The

Figure 3
How long does it take to lose birds from Kakamega's habitat fragments? We
plot a relaxation index (I), which indicates how close a fragment is to suffering so
many extinctions that it reaches a new, lower equilibrium of species numbers, against
the time (t) since isolation of each fragment. The solid lines indicate exponential decay
from the fragments with the shortest (lower line) and longest (upper line) half-lives.

Page 58

rates for other groups of species will likely be higher in that they have much greater rates of current endangerment (TNC 1996).

There are sources of uncertainty in these estimates. For example, they assume that habitat destruction will freeze at current levels. Tropical deforestation, in particular, is continuing and accelerating. The worst-case scenario is that we retain only the 5% of the world's tropical forests in protected areas—an event that will happen within 50 years at current rates (Myers 1992). Species-to-area relationships would predict that some 50% of the world's roughly 5,000 forest birds (and millions of other forest animals and plants) would go extinct eventually. Our results above suggest that half the 50% (1,250 species) will be lost before the end of the 21st century, giving an extinction rate of 1,250 extinctions per million species per year.

How do those results compare with other estimates of the magnitude and speed of the extinction? In table 2, we summarize estimates of current global extinction rates produced by seven methods, along with the background rate. The similarity of current rates and their difference by 3–4 orders of magnitude from background rates is striking.

There is a consequence:

Page 59

Policy Consequence 5. To prevent species extinctions in fragmented habitats we must act immediately, for after a narrow window of only a century, it will be too late.

Finally, we can peer into the dim, more distant future for biodiversity. Prospective extinction rates vary greatly from group to group: 11% of bird species are currently threatened with extinction on the basis of our actions to date (Baillie and Groombridge 1996). Birds appear relatively resistant to extinction. Perhaps one-fourth of all mammal species and even higher proportions of some other groups are now on their way to extinction (Baillie and Groombridge 1996).

If the destruction of natural communities that is now underway throughout the world continues at expected rates, many more species might be similarly doomed to extinction by the end of the next century. The rich tropical forests might contain as many as two-thirds of all the planet's species (Raven 1988). The loss of these forests is rapid and accelerating. We might lose all their species. Suppose that we save 5% of the forests in parks—the average global value for all protected areas—and effectively guard them from destructive incursions for the future. Our species-to-area calculations predict that we would eventually lose half the forest species—one-third of all the planet's species. Experience with tropical forests suggests that saving 5% of them will require considerable effort.

We can now give answers to the questions that we posed at the outset. How soon will extinctions occur? Very soon: we can expect to see widespread extinctions in fragmented habitats within 50 years, with the extinction rate about 1,000–10,000 times greater than background rates. Where will extinctions strike hardest? In the hot spots of biodiversity in the tropics. How many species will be lost if current trends continue? Somewhere between one-third and two-thirds of all species—easily making this event as large as the planet's previous five mass extinctions.

Page 60

Acknowledgments

This study was funded by National Geographic Society Research Award 5542–95, a Pew Fellowship in Conservation to SLP, and an American Museum of Natural History Collection Study Grant to TMB. J. Akiwumi, J. Baraza, R. Fox, R. Honea, M. Ibrahim, L. Isavwa, M. Mwangi, K. Orvis, R. Peplies, and J. Robinson helped with forest cover data. In Kakamega, J. Barnes, R. Barnes, L. Bennun, D. Gitau, T. Imboma and his family, C. Jackson, J. Kageche Kihuria, M. Kahindi, S. Karimi, L. Lens, D. Muthui, J. Odanga, D. Onsembe, N. Sagita, J. Tobias, E. Waiyaki, C. Wilder, and the rest of the staff of the Ornithology Department, NMK, and the Kenya Wildlife Service and Forest Department staff were crucial to fieldwork. T., D., and G. Cheeseman, the late G.R. Cunningham-van Someren, M. Flieg, M. Lynch. D. Turner, and D.A. Zimmerman provided further data, as did many museum staff, in particular our hosts P. Sweet, D. Willard, R. Paynter, and P. Angle; and S. Conyne. Thanks to R. May, P. Raven, D. Vázquez, C. Wilder, and an anonymous reviewer for comments on the manuscript.

References

Aleixo A, Vielliard JME. 1995. Composição e dinâmica da avifaunda da Mata de Santa Genebra, Campinas, São Paulo, Brasil. Revta bras Zool 12:493–511.

Baillie J, Groombridge B. 1996. 1996 IUCN red list of threatened animals. Gland Switzerland: The IUCN Species Survival Commission.

Balmford A, Long A. 1994. Avian endemism and forest loss. Nature 372:623–4.

Berger L, Speare R, Daszak P, Green DE, Cunningham AA, Goggin CL, Slocombe R, Ragan MA, Hyatt AD, McDonald KR, Hines HB, Lips KR, Marantelli G, Parkes H. 1998. Chytridiomycosis causes amphibian mortality associated with population declines in the rainforests of Australia and Central America. Proc Natl Acad Sci USA 95:9031–6.

Bierregaard RO Jr, Lovejoy TE, Kapos V, Santos AA dos, Hutchings RW. 1992. The biological dynamics of tropical rainforest fragments. Bioscience 42:859–66.

Brash AR. 1987. The history of avian extinction and forest conversion on Puerto Rico. Biol Cons. 39:97–111.

Brooks T, Balmford A. 1996. Atlantic forest extinctions. Nature 380:115.

Brooks TM, Pimm SL, Collar NJ. 1997. Deforestation predicts the number of threatened birds in insular South-East Asia. Cons Biol 11:382–394.

Brooks TM, Pimm SL, Oyugi JO. 1999. The time-lag between deforestation and bird extinction in tropical forest fragments. Cons Biol 13:1140–50.

Christiansen MB, Pitter E. 1997. Species loss in a forest bird community near Lagoa Santa in south-eastern Brazil. Biol Cons 80:23–32.

Collar NJ, Crosby MJ, Stattersfield AJ. 1994. Birds to watch 2. Cambridge UK: BirdLife International.

Corlett RT, Turner IM. 1997. Long-term survival in tropical forest remnants in Singapore and Hong Kong. In: Laurance WL, Bierregaard RO Jr (eds). Tropical forest remnants. Chicago IL: Univ Chicago Pr. p 333–45

Costanza R, d'Arge R, de Groot R, Farber S, Grasso M, Hannon B, Limburg K, Naeem S, O'Neill RV, Pareuelo J, Raskin RG, Sutton P, van den Belt M. 1996. The value of the world's ecosystem services and natural capital. Nature 387:253–60.

Csuti B, Polasky S, Williams PH, Pressey RL, Camm JD, Kershaw M, Kiester AR, Downs B, Hamilton R, Huso M, Sahr K. 1997. A comparison of reserve selection algorithms using data on terrestrial vertebrates in Oregon. Biol Cons 80:83–97.

Curnutt J, Lockwood J, Luh H-K, Nott P. Russell G. 1993. Hotspots and species diversity. Nature 367:326–7.

Page 61

Daniels R JR, Joshi NV, Gadgil M. 1990. Changes in the bird fauna of Uttara Kannada, India, in relation to changes in land use over the past century. Biol Cons 52:37–48.

Diamond JM. 1972. Biogeographic kinetics: estimation of relaxation times for avifaunas of southwest Pacific islands. Proc Natl Acad Sci USA 69:3199–3203.

Diamond J M, Bishop KD, van Balen SV. 1987. Bird survival in an isolated Javan woodland: island or mirror? Cons Biol 1: 132–42.

Ehrlich PR, Ehrlich AH. 1981. Extinction: the causes and consequences of the disappearance of species. New York NY: Random.

Ehrlich PR. 1994. Energy use and biodiversity loss. Phil Trans R Soc Lond B 344:99–104.

Gaston KJ. 1994. Rarity. London UK: Chapman & Hall.

Howard PC, Viskanic P, Davenport TRB, Kigenyi FW, Baltzer M, Dickinson CJ, Lwanga JS, Matthews RA, Balmford A. 1998. Complementarity and the use of indicator groups for reserve selection in Uganda. Nature 394:472–5.

Hughes JB, Daily GC, Ehrlich PR. 1997. Population diversity: its extent and extinction. Science 280:689–92.

Hughes JB, Daily GC, Ehrlich PR. 2000. The loss of population diversity and why it matters. In: Raven PH, Williams T (eds). Nature and human society: the quest for a sustainable world. Washington DC: National Academy Press. p 71–83.

Kattan GH, Alvarez-Lopez H, Giraldo M. 1994. Forest fragmentation and bird extinctions: San Antonio eighty years later. Cons Biol 8:138–146.

Klicka J, Zink RM. 1997. The importance of recent ice ages in speciation: a failed paradigm. Science 277:1666–9.

Laurance WF, Bierregaard Jr RO, Gascon C, Didham RK, Smith AP, Lynam AJ, Viana VM, Lovejoy TE, Sieving KE, Sites Jr JW, Andersen M, Tocher MD, Kramer EA, Restrepo C, Moritz C. 1997. Tropical forest fragmentation: synthesis of a diverse and dynamic discipline. In: Laurence WL, Bierregaard Jr RO (eds). Tropical forest fragments. Chicago IL: Univ Chicago Pr. p 502–14.

Leakey R, Lewin R. 1996. The Sixth Extinction. New York NY: Doubleday.

Lombard AT. 1995. The problems with multi-species conservation: do hot spots, ideal reserves and existing reserves coincide? S Afr J Zool 30:145–63.

Mace GM. 1994. Classifying threatened species: means and ends. Phil Trans R Soc Lond B 344:91–7.

Manne LL, Brooks TM, Pimm SL. 1999. Relative risk of extinction of passerine birds on continents and islands. Nature 399:258–61.

May RM. 2000. The dimensions of life on earth. In: Raven PH, Williams T (eds). Nature and human society: the quest for a sustainable world. Washington DC: National Academy Press. p 30–45.

May RM, Lawton JH, Stork NE. 1995. Assessing extinction rates. In: Lawton JH, May RM (eds). Extinction Rates. Oxford UK: Oxford Univ Pr. p. 1–24.

Mayer AL, Pimm SL. 1998. Integrating endangered species protection and ecosystem management: the Cape Sable Seaside-sparrow as a case study. In: Mace GM, Balmford A, Ginsberg JR (eds). Conservation in a changing world. Cambridge UK: Cambridge Univ Pr. p 53–68.

McKinney ML, Lockwood JL, Daniel F. 1996. Does ecosystem and evolutionary stability include rare species? Palaeogeogr Palaeoclim Palaeoecol 127:191–207.

Myers N. 2000. The meaning of biodiversity loss. In: Raven PH, Williams T (eds). Nature and human society: the quest for a sustainable world. Washington DC: National Academy Press. p 63–70.

Myers N. 1979. The sinking ark: a new look at the problem of disappearing species. London UK: Pergammon Pr.

Myers N. 1988. Threatened biotas: “hot spots” in tropical forests. Environmentalist 8:1–20.

Myers N. 1990. The biodiversity challenge: expanded hot-spots analysis. Environmentalist 10:243–56.

Myers N. 1992. The primary source: tropical forests and our future. New York NY: WW Norton.

Newmark WD. 1996. Insularization of Tanzanian parks and the local extinction of large mammals. Cons Biol 10:1549–56.

Norton B. 1988. Commodity, amenity and morality. In: Wilson EO, Peter FM (eds). Biodiversity. Washington DC: National Acad Pr. p 200–5.

Pimm, SL. 1991. The balance of nature? Chicago IL: Univ Chicago Pr.

Pimm SL, Askins RA. 1995. Forest losses predict bird extinction in eastern North America. Proc Natl Acad Sci USA 92:9343–7.

Pimm SL, Lawton JH. 1998. Planning for biodiversity. Science 2279:2068–9.

Page 62

Pimm SL, Moulton MP, Justice LJ. 1994. Bird extinctions in the central Pacific. Phil Trans R Soc Lond B 344:27–33.

Pimm SL, Russell GJ, Gittleman JL, Brooks TM. 1995. The future of biodiversity. Science 269:347–50.

Prendegast JR, Quinn RM, Lawton JH, Eversham BC, Gibbons DW. 1993. Rare species, the coincidence of diversity hot spots and coincidence strategies. Nature 365:335–7.

Pressey RL, Humphries CJ, Margules CR, Vane-Wright RI, Williams PH. 1993. Beyond opportunism: key principles for systematic reserve selection. Trends Ecol Evol 8:124–8.

Preston FW. 1962. The canonical distribution of commonness and rarity. Parts 1 and 2. Ecology 43:185–215, 410–32.

Raup DM. 1979. Size of the Permo-Triassic bottleneck and its evolutionary implications. Science 206:217–8.

Raven PH. 1988. Our diminishing tropical forests. In: Wilson EO, Peter FM (eds). Biodiversity. Washington DC: National Acad Pr. p 119–122.

Reid WV. 1992. How many species will there be? In: Whitmore TC, Sayer JA (eds). Tropical deforestation and species extinction. London UK: Chapman & Hall. p 55–73.

Reid WV. 1998. Biodiversity hotspots. Trends Ecol Evol 13:275–80.

Renjifo LM. 1998. Changes in the avifauna of an Andean region following eight decades and a half of forest fragmentation. In: North American Ornithological Conference, 6–12 April 1998. CD-ROM. St. Louis.

Robinson WD. 1999. Long-term changes in the avifauna of Barro Colorado Island, Panama, a tropical forest isolate. Cons Biol 13:85–97.

Rosenzweig ML. 1995. Species diversity in space and time. Cambridge UK: Cambridge Univ Pr.

Russell GJ, Brooks TM,. McKinney L, Anderson CG. 1998. Present and future taxonomic selectivity in bird and mammal extinctions. Cons Biol 12:1365–76.

Simon JL. 1986. Disappearing species, deforestation and data. New Sci. 110(May):60–3.

Smith FDM., May RM, Pellew R, Johnson TH, Walter KR. 1993. Estimating extinction rates. Nature 364:494–6.

Soulé ME, Wilcox BA, Holtby C. 1979. Benign neglect: a model of faunal collapse in the game reserves of East Africa. Biol Cons 15:259–72.

Steadman D. 1995. Prehistoric extinctions of Pacific island birds: biodiversity meets zooarcheology. Science 267:1123–31.

Steadman D. 1997. Human-caused extinction of birds. In: Reaka-Kudla HML, Wilson DE, Wilson EO (eds). Biodiversity II. Washington DC: Joseph Henry Pr. p 139–61.

Stork NE. 1997. Measuring global biodiversity and its decline. In: Reaka-Kudla HML, Wilson DE, Wilson EO (eds). Biodiversity II. Washington DC: Joseph Henry Pr. p 41–68.

Terborgh J. 1975. Faunal equilibria and the design of wildlife preserves. In: Golley FB, Medina E (eds). Tropical ecological systems. New York NY: Springer Verlag. p 369–80.

Terborgh J, Lopez L, Tello J, Yu D, Bruni AR. 1997 Transitory states in relaxing ecosystems of land bridge islands. In: Laurence WF, Bierregaard Jr RO (eds). Tropical forest remnants. Chicago: Univ Chicago Pr. p 256–74.

TNC [The Nature Conservancy]. 1996. TNC priorities for conservation: 1996 annual report card for U.S. plant and animal species. Arlington VA: The Nature Conservancy.

Turner IM. 1996. Species loss in fragments of tropical rain forest: a review of the evidence. J Appl Ecol 33:200–9.

Van Jaarsveld AS, Freitag S, Chown SL, Muller C, Koch S, Hull H, Bellemy C, Krüger M, Endrödy-Younga S, Mansell MW, Scholtz CH. 1998. Biodiversity assessment and conservation strategies. Science 279:2106–8.

Wilson EO. 2000. The creation of biodiversity. In: Raven PH, Williams T (eds). Nature and human society: the quest for a sustainable world. Washington DC: National Academy Press. p 22–9.

Wilson EO. 1992. The diversity of life. Boston MA: Belknap Harvard.