CIESIN Reproduced, with permission, from: Reid, W. R. 1992. How many species will there be? Chapter 3 in Tropical deforestation and species extinction, ed. T. C. Whitmore and J. A. Sayer. New York: Chapman and Hall.


How many species will there be?



There is little doubt that the rate of species extinction has grown during the course of this century, and a consensus exists among scientists that a significant loss of the world's species will occur in coming decades if present trends of tropical deforestation continue. How large is the loss of species likely to be? Although the loss of species may rank among the most significant environmental problems of our time, relatively few attempts have been made to rigorously assess its likely magnitude.

The scant literature on extinction rates is due in part to data limitations and thus a recognition of the imprecision of any estimate. Moreover, it can be argued that the exact rate of extinction is not terribly important given that current extinction rates greatly exceed background rates. For instance, 60 birds and mammals are known to have become extinct between 1900 and 1950 (Fig. 3.1), whereas the background extinction rate for these two groups is only one extinction every 100 to 1000 years based on the average lifespan of a species of 1 to 10 million years (Raup, 1978). Overall, some 724 species are known to have been lost since 1600, and these recorded extinctions are no doubt only a fraction of the total (Table 3.1).

Despite the merits of this argument, the fact remains that most people and policy-makers are unlikely to assess the urgency of today's extinction crisis -- and determine the priority given to the issue -- by comparing it to background rates of extinction. Instead, priorities for action will be set on the basis of the absolute numbers of species likely to be lost over the coming decades, or on the basis of threats to certain 'flagship' species. A loss of 25% of the world's species over one human lifetime would be staggering by any measure and would readily mobilize political action to slow that loss. In contrast, a loss of 1% of the world's species over the same interval, though equally staggering from an evolutionary perspective, would probably result in the problem being placed on the back-burner of needed social and political change.

While better knowledge of extinction rates can clearly improve the design of public policies, it is equally apparent that estimates of global extinction rates are fraught with imprecision. We do not yet know how many species exist, even to within an order of magnitude (Wilson, 1988); the best data on rates of tropical deforestation are bounded by large uncertainties (WRI, 1990; Sayer and Whitmore, 1990; FAO, 1990); and essential information is lacking on endemism, forest fragmentation, and the potential for species persistence in disturbed habitats. Moreover, ecologists can never even know if extinction rate estimates are borne out. It will not be possible to determine at a future date the number of species that went extinct between now and that future date without knowledge of how many species exist today.

The serious problems associated with extinction rate estimation may explain why the field has advanced relatively slowly since the early calculations of Myers (1979) and Lovejoy (1980), but given the importance of the issue it is surprising that ecologists have not mounted a more substantial effort to estimate rates of extinction and thereby help clarify the actions needed in response. For example, consider the comparable problems and uncertainties associated with predicting rates and impacts of global warming. Policy-makers realize that considerable uncertainty surrounds predictions of climate change, yet they are willing to modify policies based on the consensus views of the scientific community. Although the uncertainty is great, decision-makers accept that the best estimates of reputable scientists are a much better basis for policy formulation than no estimate at all. Accordingly, for the past three years the United Nations Environment Programme and the World Meteorological Organization have facilitated the development of a world-wide scientific consensus on climate change impacts and response options through the IPCC (Intergovernmental Panel on Climate Change).

The issue of climate change contrasts with that of species extinction. Although many ecologists agree that the rate of loss of species in tropical forest is high (but see Simon, 1986; Mares, 1986; Lugo, 1988), the literature addressing this phenomenon is relatively small (Table 3.2). Increased research and debate on this topic would have a considerable pay-off. Efforts to clarify the magnitude of the extinction crisis and the steps that can be taken to defuse the crisis could considerably expand the financial and political support for actions to confront what is indisputably the most serious issue that the field of ecology faces, and arguably the most serious issue faced by humankind today.


The best tool available to estimate species extinction rates is the use of species-area curves. These curves represent the relationship between the number of species in a region and its area. Based on these curves, it is possible to predict the proportion of species that will become extinct in a region based upon the amount of habitat that is lost. This approach has formed the basis for almost all current estimates of species extinction rates. This technique can either be applied to specific localities in which deforestation rates and species numbers are high (Raven, 1987, 1988a, b; Myers, 1988) or can be applied on a regional (Simberloff, 1986) or global (Lovejoy, 1980; Wilson, 1988; Reid and Miller, 1989) basis.

The following sections update the analysis of Reid and Miller (1989) based on recent evidence that tropical deforestation is proceeding much more rapidly than previously thought. This is followed by a critical examination of the assumptions behind the use of the species-area technique to clarify the limits of this type of analysis, and a brief exploration of the policy options suggested by the analysis as means for lessening the magnitude of the extinction crisis.

Tropical forest habitat

Three parameters are needed to estimate species loss due to deforestation: current forest area, rate of deforestation, and the slope of the species-area curve. Tropical forests can be readily divided into closed and open forest types, based on the density of trees and the presence or absence of a continuous grass cover below the trees (FAO, 1981). Closed tropical forests support the great majority of tropical species and thus it is most meaningful to examine trends of deforestation and extinction in this one forest type. While the area of climates suitable for closed tropical moist forest is 1600 million hectares (Sommer, 1976), part of this is thought to have historically been covered by other vegetation types due to soil type and other ecological and historical factors. Simberloff (1986) estimated that prior to significant human impacts, 10% was not forested, thus the original extent of the biome was 1440 million hectares (Table 3.3).

The most recent estimates of closed tropical forest cover indicate that in the 1980s the total area was 1166 million hectares--roughly 80% of its original extent (Table 3.3). Note that 'closed forest' cover includes more vegetation types than 'closed moist forest', thus the figure of 80% loss is conservative. An estimated 15-20% of this forest area was disturbed by logging or related activities (Lanly, 1982; Reid and Miller, 1989). Land with woody vegetation that was in the fallow portion of shifting cultivation cycles is not included in the total of forest cover.

The term 'deforestation', as used by forestry statisticians, refers to the transformation of forested land to permanently cleared land or to a shifting-cultivation cycle. In a study conducted in the early 1980s, FAO estimated that annual tropical deforestation amounted to 11.4 million hectares per year (Lanly, 1982). Of this, 7.4 million hectares (65%) of closed forest were lost each year. However, three new estimates of tropical forest loss have recently been released that indicate deforestation is proceeding much more rapidly than previously thought. World Resources Institute has estimated that tropical forests are being lost at a rate of 20.4 million hectares annually, an increase of 79% from FAO's estimate of total forest loss (WRI, 1990). Myers (1989), examined rates of loss of closed tropical forest and concluded that 14.2 million hectares were being lost per year in the 1980s--90% greater than FAO's estimate of closed forest loss (in part because the studies used different definitions of 'deforestation') (Sayer and Whitmore, 1990). Finally, FAO released updated statistics placing tropical closed and open forest loss at 16.8 million hectares per year (FAO, 1990). The most up-to-date estimate of deforestation rates for closed tropical forest is 10.5 million hectares per year (see the Appendix1 2).

Deforestation rates are imprecise to begin with, and are likely to change significantly in coming decades. In some regions, rates may increase because of population growth or increased access to forest resources. Satellite data for the southern portion of the Amazon basin of Brazil, for example, indicate that deforestation rates rose exponentially between 1975 and 1985 (Malingreau and Tucker, 1988). Eventually, however, rates of forest loss will slow when the most accessible land has been cleared. Accordingly, in order to reflect the considerable uncertainty in predicting future rates of forest loss, species extinction rate is modelled here based on three scenarios of closed tropical forest loss: 5 million, 10 million, and 15 million hectares per year. The low scenario is well below any current estimates, but allows examination of the decrease in extinction rates that could be achieved if a concerted global effort were made to slow tropical forest loss. The middle scenario is comparable to the best current estimate of rates of closed forest loss, and the high scenario exceeds the highest current estimate of closed forest loss.

Species-area curve

Species-area curves generally fit closely to equations in the form:

S = cA[z] (1)

where S = number of species, A = area, and c and z are constants. The exponent z determines the slope of the curve and is the critical parameter in estimating extinction rates.

The estimation of extinction rates is sensitive to the form of the species-area curve. Slopes of species-area curves may differ among various regions of tropical forests because of differences in the numbers of habitats or life-zones present in the region. The slopes are also likely to differ among taxa due to differences in the average size of species' ranges. Curves for groups that tend to have small ranges (high local endemism) should have relatively steep slopes. In addition, species-area curves measured for island flora and fauna differ from those measured for subsets of continental habitats, and the choice of which type of curve to use is not straightforward, as will be discussed further below. Because of the uncertainty associated with the choice of the proper slope for the curve, species extinction rates are modelled over a broad range of slopes that have been found in empirical studies (0.15 < z < 0.35, Connor and McCoy, 1979).


Extinction rates are estimated for each scenario of deforestation by first calculating regional deforestation rates for Africa and Madagascar, Asia and Pacific, and tropical America assuming that under any deforestation scenario the rate of deforestation for each region would be a constant fraction of the global rate. Initial forest cover in 1990 is then calculated from estimated forest cover in the 1980s (Table 3.3), by subtracting five years of forest loss at the appropriate regional deforestation rate. Then, for each region proportionate species loss is calculated using Equation (1) over time periods of 0-50 years and over species-area slopes in the chosen range of 0.15 < z < 0.35

The analysis is performed regionally rather than globally because species diversity differs considerably among the three regions. Africa and Madagascar are believed to contain about 23% of the world's tropical plant species, Asia contains 26% and the neotropics contain 51% (Raven, 1987; Reid and Miller, 1989). Thus, a 5% loss of species in Africa would contribute much less to global species extinction rates than a 5% loss in Latin America. For this same reason it could be argued that the analysis should be performed at an even more fine-grained resolution (i.e., at a country level), but the uncertainties in estimates of both forest loss and species richness at this level of resolution outweigh any theoretical gains in precision.

Global estimates of species extinction rates are then obtained from the regional analyses by weighting each regional estimate by the fraction of the tropical flora occurring in that region. This step assumes that regional patterns of diversity in all tropical forest species parallel the pattern found in plants.


The extinction rates calculated with this model do not represent the actual loss of species over the time period indicated. Rather, they estimate the proportion of species that will eventually go extinct when the system reaches equilibrium following the loss of a given amount of forest. This distinction can best be seen by way of example. Barro Colorado Island in Panama was created around 1914 when the Panama Canal was built and surrounding valleys were dammed. Of the approximately 200 land bird species known to have bred on the island, 47 had disappeared by 1981 (Karr, 1982). The isolation of the island from the formerly continuous habitat committed these species to local extinction, but in many cases the species persisted on Barro Colorado for decades after the island was created. Similarly, estimates derived from the use of species-area curves indicate only the portion of species that will go extinct unless their habitat is restored as species numbers move toward a new equilibrium. Particularly vulnerable species may go extinct immediately and other species with short generation times might be lost within a few years, but some long-lived trees may persist for centuries or even millennia.

The extinction rates predicted by the model for the three regions differ considerably due to regional differences in both deforestation rates and the extent of current forest cover (Table 3.4 and Fig. 3.2). For example, at deforestation rates of 10 million hectares per year, the average extinction rate for Africa is 1-2% of species per decade, whereas the rate is 2-5% per decade in Asia. Globally, the model predicts that in the next 25 years, current rates of forest loss (approximated by the middle scenario) will commit between 4 and 8% of the world's closed tropical forest species to extinction. These estimates would be reduced to as low as 2% under the low scenario of deforestation, but could be as great as 6-14% of species under the high scenario. History has shown an accelerating rate of deforestation in tropical regions, thus the 'high' scenario may be the most probable in coming decades.

These estimates of species extinction in closed tropical forests can be placed in a global perspective by assuming that the bulk of species loss in coming decades will result from tropical deforestation. Since an estimated 50-90% of the world's species occur in closed tropical forest (Myers, 1980; Reid and Miller, 1989), then at current rates of deforestation the world stands to commit 2-7% of species to extinction in the next quarter century and if rates of deforestation accelerate this loss would increase to as much as 13%. With roughly 10 million species on earth, at current rates of forest loss this would amount to between 8000 and 28,000 species per year, or 20-75 species per day.

The magnitude of this estimate, although lower than some other predictions (Table 3.2), seems to defy common sense and experience. But consider that most of us are barely conscious of the vast majority of species on earth. Large, visible, species like birds, mammals and plants make up less than 5% of the world's species. Considering only these three groups, the results of this study imply that one bird, mammal, or plant will be committed to extinction every 0.5-1.7 days at current rates of forest loss.


Like any model, the species-area model is an approximation that is based upon several assumptions. First, the model assumes that deforestation eliminates all species originally present in the forest. In many instances, this would overestimate the actual impact of deforestation. For instance, land under shifting cultivation is considered deforested, yet many forest-dependent species could persist in the woody vegetation of fallow land. Birds and mammals in tropical rain forest have been found to survive in habitat that is slightly or moderately disturbed by low-density shifting agriculture and low levels of selective logging (Johns, 1986, 1988, Chapter 2), and deforestation may leave relict fragments of forest in which some species continue to survive (Brown and Brown, Chapter 6).

However, this bias is somewhat offset because the model also assumes that species richness in all forested land is at historic levels. In fact, much of the forested land included in the calculation of species loss has been logged, and logging in any significant scale changes habitat structure and species composition. Arguably such logged forest could be considered 'deforested', but more likely the effects on the number of the original species present will be intermediate. For example, one study in Malaysia found that a 25-year-old logged forest contained nearly 75% of the avifauna of a virgin forest (Wong, 1985). In that study the presence of adjacent undisturbed forest probably increased the number of species found in the disturbed site and, moreover, the pattern cannot be readily extrapolated to other groups such as insects. Nevertheless, the study does indicate that more work is needed to clarify the relationship between various types of habitat disturbance and species persistence.

Second, the model assumes that extinction rates are unaffected by forest fragmentation. Obviously, this is unrealistic. In practice, deforestation converts relatively continuous tracts of forest into a fragmented array of smaller patches. Many of the species present in the patches may be lost if their populations are reduced below their minimum viable population size (Shaffer, 1981). But exactly how fragmentation affects species richness within a large region depends upon which areas remain under forest cover. For example, more species may persist if several patches of forest with high species richness or high endemism are selected for protection than if one large tract of similar area but lower richness or endemism is protected (Simberloff and Abele, 1982; Quinn and Harrison, 1988). In general, however, there is every reason to think that by not including the effect of fragmentation, species-area models substantially underestimate the extinctions that will actually occur (Simberloff, Chapter 4). FAO has now begun to include a 'Fragmentation Index' in their reports of forest area which may help ecologists to refine estimates of species loss (FAO, 1990).

Related to this point, the model also assumes that habitat loss occurs randomly among regions with various levels of species richness. If lowland forest sites with high species richness are preferentially deforested, the analysis would underestimate extinction rates. Alternatively, if species-rich sites are preferentially protected, the model would overestimate extinction rates. To circumvent the uncertainties inherent in this assumption, other studies have focused on more detailed assessments of the number of species in regions experiencing high rates of deforestation (Raven, 1988a,b; Myers, 1988).

Finally, the model differs from some applications of the species-area analysis by not assuming either a 'continental' or 'island' slope for the species-area curve. Species-area curves for habitat islands have steeper slopes than curves for subsamples of continuous habitats due to the 'relaxation effect' (Connor and McCoy, 1979). The relaxation effect refers to the gradual loss of species through time after a habitat is fragmented. Many populations are reduced below their minimum viable population size in small habitat fragments. Although these species may persist for varying periods of time after the fragment is created, with time the number of species present decreases until only species with sufficiently large populations remain. Accordingly, although many exceptions exist, the value of the exponent z in the species-area curve for islands generally falls between 0.2 and 0.4 whereas for subsamples of continuous habitat it is less steep, lying between 0.12 and 0.19 (MacArthur and Wilson, 1967; Connor and McCoy, 1979).

Clearly, the goal of this analysis is to examine species loss associated with deforestation, thus fitting the conditions of habitat 'islands' rather than continuous habitats. However, because the model assumes that remaining tropical forest is a continuous, unfragmented tract in each region, the size of each of the three 'islands' greatly exceeds the size of islands where species-area curves have been measured. Even species-area curves that have been measured on continents have involved smaller and generally much more uniform tracts of habitat than would be found in the large areas involved in this study like the Amazon basin.

The choice of the appropriate species-area curve thus becomes quite problematic. Strictly speaking, since the model assumes that the forest is unfragmented and since the area involved is so large, there is little justification for assuming that species number will 'relax' to lower levels with time. Thus a continental species-area curve seems most suited for the analysis. However, because the area involved is so large and heterogeneous, it is not entirely clear if a standard continental curve (measuring primarily beta or 'within-habitat' diversity) is appropriate since considerable gamma or 'between-habitat' diversity would exist. For example, I calculated a species-area curve based on data for plant species richness in 57 tropical countries (data from Davis et al., 1986, supplemented by Gentry, 1986). The exponent z for this curve is 0.30 (+/- 0.066 SE)--falling within the range of 'island' exponents and outside the range of 'continental' exponents.

Added to these issues are the already-mentioned difficulties related to the fact that species-area slopes may differ among regions because of differences in the numbers of habitats present in the region, and slopes are also likely to differ among various taxa due to differences in the level of local endemism. Thus, the best option appears to be to model extinction rates using a broad range of estimates of the slope of the species-area curve.

Because the model assumes that the forest area in each region is continuous, a case could be made that the estimates of species loss apply only to immediate extinctions (due to the complete loss of certain habitats) and do not incorporate relaxation effects. However, in reality, tropical forests are being extensively fragmented and thus it is more realistic to treat the estimates derived as indicating the number of species that will remain when the system ultimately reaches equilibrium.


The estimates of extinction rates derived here are comparable to estimates derived in several recent studies (Table 3.2). All these studies have yielded rates in the range of 1-10% extinction (or commitment to extinction at equilibrium) per decade. In some cases these estimates have applied only to plants, but it is reasonable to assume that rates of extinction would be at least as great for all species (since numbers of invertebrate species are even more concentrated in the tropics than plants).

More significantly, although all the estimates in Table 3.2 are based on species-area analysis, two different methods have been used. This study has examined global patterns of forest loss, whereas others such as Raven (1988a, b) and Myers (1988) have focused on specific regions facing serious threats. Accordingly, the general concurrence of results suggests that the conclusions are robust. The fact that the results of this study are somewhat lower than those of Raven (1988a, b) could well be due to the violation of the assumption noted above regarding the pattern of forest loss. If species-rich habitats are being preferentially lost, then the approach used in this study would underestimate global extinction rates. Historical trends provide ample reason to believe that many species-rich regions are under serious threat. For example, over 95% of the species-rich Atlantic coastal forest of Brazil has been lost and less than 9% of the original primary forest of the Philippines remains (Forest Management Bureau, 1988). Myers (1988) has calculated that 10 tropical forest regions, covering only 3.5% of remaining tropical forest contain at least 27% of higher plant species and some 14% of these species are endemic to these regions. Thus, the results of this study may well be conservative estimates of extinction rates. The differences among estimates are due more to methodological than biological factors since key data relating to patterns of fragmentation and endemism are missing.

But especially since they are conservative, the results are cause for alarm. If forest loss continues to accelerate, as many as 35% of tropical forest species may be committed to eventual extinction by the year 2040. Climatic changes associated with the loss of such large areas of forest could exacerbate this potential loss. If instead, a commitment is made to slow deforestation rates to roughly one-half of current levels, the threat to species could drop to well under 10% by the year 2040.

Species loss can be slowed not only by blanket efforts to slow tropical deforestation, but also by protection of key forest habitats with high species-richness and endemism. Tropical forests are being fragmented and lost, but by controlling the pattern and location of fragments and by ensuring that relatively large areas remain in natural and semi-natural habitats, the loss of species could be significantly reduced. Unfortunately, the detailed studies of tropical forest species distributions required to choose such sites are still lacking despite eloquent calls for their implementation made a decade ago (NRC, 1980).

The fact that tropical forests are becoming increasingly fragmented also points to the need for greatly expanded research on the management of small populations. As habitat area is reduced, extinction can be slowed to the extent that managers are able to maintain forest fragments in a supersaturated state; that is, to the extent that species numbers are prevented from 'relaxing' to their natural equilibria.

Species loss can also be slowed by enhancing the conservation of biodiversity in disturbed habitats. As with the need for rapid biological inventories, there is a pressing need for research to identify resource management techniques that will better meet conservation needs under the constraints of timber harvest or rural agriculture. Clearly, regional land-use planning must increasingly incorporate biodiversity conservation as a major planning goal.

A substantial loss of species is already underway, and we are certain to experience species extinctions for decades because of the habitat loss and forest fragmentation that has already occurred. But while the stage is already set for future losses, the magnitude of the extinction can still be influenced by national and international policies and programmes. Many species already 'condemned' to extinction could in fact be saved through restoration of habitat or through new management techniques. But more importantly, a global effort to slow tropical forest loss, rationalize forest use, and maintain key forest habitats could dramatically slow the rate of extinction. Based on the model described above, the 40-90% increase in rates of closed forest loss in the 1980s, if maintained until 2040, will increase the number of tropical species condemned to extinction by between 60 and 160%. These trends must be reversed.


Recent evidence that tropical deforestation has accelerated in the 1980s has profound implications for the persistence of tropical species. Although estimates of global extinction rates are fraught with imprecision, species-area techniques do provide a means for shedding light on the probable effects of deforestation on species extinction. Based on species-area techniques and current rates of forest loss, over the next 25 years an estimated 4-8% of closed tropical forest species are likely to be committed to extinction. This loss will take place over a number of years as a new equilibrium number of species is achieved. If forest loss continues to accelerate, by the year 2040 some 17-35% of tropical forest species could be committed to eventual extinction when equilibrium numbers are reached. The 40-90% rise in the rate of deforestation of closed tropical forests that has occurred in the last decade has increased by 60-160% the number of tropical species likely to be condemned to extinction in the next half-century.


I thank Dan Janzen, Daniel Simberloff, and Mike Soulé for their helpful comments on a preliminary version of this manuscript, and Mohamed El-Ashry, Kenton Miller, Norman Myers, Jeff McNeely, and Peter Raven for their comments on more recent drafts.

[a] Based on total species number of 10 million. Estimates in bold face indicate the actual loss of species over that time period (or shortly thereafter). Estimates in standard type refer to the number of species that will be committed to extinction during that time period as a new equilibrium is attained.

[b] See text for definition of z.

[c] Extinction estimates apply to the number of species committed to extinction by the year 2000 at current rates of forest loss. How long it will take for the new equilibrium to be achieved is not known.

[d] Estimate refers to number of species committed to eventual extinction when species numbers reach equilibrium following forest loss.

[e] This estimate applies only to hot spot regions, thus the global extrapolation is conservative.


Connor, E.F. and McCoy, E.D, (1979) The statistics and biology of the species-area relationship. American Naturalist 13, 791-833.

Davis, S.D., Droop, S.J.M., Gregerson, P., Henson, L., Leon, C.J., Villa-Lobos, J.L., Synge, H. and Zantovska. J. (1986) Plants in Danger: What Do We Know? International Union for Conservation of Nature and Natural Resources, Gland, Switzerland.

FAO (Food and Agriculture Organization of the UN) (1981) Tropical Forest Resources Assessment Project. 4 vols. FAO, Rome.

FAO (1988) An Interim Report on the State of Forest Resources in the Developing Countries. FO:MISC/88/7, FAO, Rome.

FAO (1990) Interim Report on Forest Resources Assessment 1990 Project. Committee on Forestry, Tenth Session 24-28 September, COFO-90/8(a) Rome.

Fearnside, P.M. (1990) Deforestation in Brazilian Amazonia: the rates and causes of forest destruction (unpublished manuscript). National Secretariat of Science and Technology, National Institute for Research in Amazonia (INPA), Manaus, Brazil.

Forest Management Bureau (1988) Natural Forest Resources of the Philippines. Philippine German Forest Resources Inventory Project (pp. 62 mimeo).

Gentry, A.H. (1986) Endemism in tropical versus temperate plant communities, in Conservation Biology: The Science of Scarcity and Diversity (ed. M.E. Soulé) Sinauer Associates, Sunderland, MA, pp. 153-81.

Grainger, A. (1984) Quantifying changes in forest cover in the humid tropics: overcoming current limitations. Journal of World Forest Resource Management 1, 3-63.

Johns, A.D. (1986) Effects of Habitat Disturbance on Rainforest Wildlife in Brazilian Amazon. Final report, World Wildlife Fund US (WWF) project US-302. WWF, Washington, DC.

Johns, A.D. (1988) Economic development and wildlife conservation in Brazilian Amazonia. Ambio, 17, 302-6.

Karr, J.R. (1982) Avian extinction on Barro Colorado Island, Panama: a reassessment. American Naturalist, 119, 220--39.

Lanly, J.P. (1982) Tropical Forest Resources. Forestry Paper No. 30, Food and Agriculture Organization of the United Nations, Rome.

Lovejoy, T.E. (1980) A projection of species extinctions, in Council on Environmental Quality (CEQ), The Global 2000 Report to the President, Vol. 2. CEQ, Washington, DC, pp. 328-31.

Lugo, A.E. (1988) Estimating reductions in the diversity of tropical forest species, in Biodiversity (eds E.O. Wilson and F.M. Peter), National Academy Press, Washington, DC, pp. 58-70.

MacArthur, R.H. and Wilson, E.O. (1967) The Theory of Island Biogeography. Princeton University Press, Princeton, RI.

Malingreau, J.P. and Tucker, C.J. (1988) Large-scale deforestation in the southeastern Amazon basin of Brazil. Ambio, 17, 49-55.

Mares, M.A. (1986) Conservation in South America: Problems, consequences, and solutions. Science, 233, 734-9.

Myers, N. (1979) The Sinking Ark: A New Look at the Problem of Disappearing Species. Pergamon Press, Oxford.

Myers, N. (1980) Conversion of Tropical Moist Forests. The National Research Council, National Academy of Sciences, Washington, DC.

Myers, N. (1988) Threatened biotas: 'hotspots' in tropical forests. Environmentalist, 8(3), 1-20.

Myers, N. (1989) Deforestation Rates in Tropical Forests and their Climatic implications. Friends of the Earth, London

NRC. (1980) Research Priorities in Tropical Biology. National Academy of Sciences, Washington, DC.

Ono, R.D., Williams, J.D. and Wagner, A. (1983) Vanishing Fishes of North America. Stone Wall Press, Inc., Washington, DC.

Quinn, J.F. and Harrison, S.P. (1988) Effects of habitat fragmentation and isolation on species richness: evidence from biogeographic patterns. Oecologia, 75, 132-40.

Raup, D. M. (1978) Cohort analysis of generic survivorship. Paleobiology, 4, 1-15.

Raven, P. H. (1987) The scope of the plant conservation problem world-wide, in Botanic Gardens and the World Conservation Strategy (eds D. Bramwell, O. Hamann, V. Heywood, and H. Synge), Academic Press, London, pp. 19-29.

Raven, P. H. (1988a) Biological resources and global stability, in Evolution and Coadaptation in Biotic Communities (eds S. Kawano, J.H. Connell, and T. Hidaka), University of Tokyo Press, Tokyo, pp. 3-27.

Raven, P.H. (1988b) Our diminishing tropical forests, in Biodiversity (eds E.O. Wilson and F.M. Peter), National Academy Press, Washington, DC, pp. 1l9-22.

Reid, W.V. and Miller, K.R. (1989) Keeping Options Alive: The Scientific Basis for Conserving Biodiversity. World Resources Institute, Washington, DC.

Sayer, J.A. and Whitmore, T.C. (1990) Tropical moist forests: destruction and species extinction. Biological Conservation, 55, 199-201.

Shaffer, M.L. (1981) Minimum population sizes for species conservation. BioScience, 31, 131 -4.

Simberloff, D. (1986) Are we on the verge of a mass extinction in tropical rain forests? in Dynamics of Extinction, (ed. D.K. Elliott), Wiley, New York, NY, pp. 165-80.

Simberloff, D. and Abele, L.G. (1982) Refuge design and island biogeographic theory: effects of fragmentation. American Naturalist, 120, 41-50.

Simon, J.L. (1986) Disappearing species, deforestation and data. New Scientist, 15 May, 60-3.

Sommer, A. (1976) Attempt at an assessment of the world's tropical moist forest. Unasylva, 28(112 + 113), 5-24.

Wilson, E.O. (1988) The current state of biological diversity, in Biodiversity (eds E.O. Wilson and F.M. Peter), National Academy Press, Washington, DC, pp. 3-18.

Wilson, E.O. (1989) Threats to biodiversity. Scientific American, September 1990, 108-16.

Wong, M. (1985) Understorey birds as indicators of regeneration in a patch of selectively logged west Malaysian rainforest, in Conservation of Tropical Forest Birds (eds A.W. Diamond and T.E. Lovejoy), Technical Publication No. 4. International Council for Bird Preservation, Cambridge, UK, pp. 249-63.

WRI (World Resources Institute) (1990) World Resources 1990-1991. Oxford University Press, NY.