TREE vol. 13, no. 7 July 1998 275

The study of the distributionof species, which has longbeen a central focus ofecology and biogeogra-

phy, is taking on new urgency as evidence of the global biodiver-sity crisis mounts. The question of what geographical regions toprotect in order to maintain themost biological diversity is centralto the design of effective conser-vation programs. For a field ofstudy with a lineage that includesAndrewartha and Birch’s (1954)Distribution and Abundance of Ani-mals1 and Wallace’s (1876) TheGeographical Distribution of Ani-mals2, the answer to this questionwould seem to be readily at hand.However, the question is provingchallenging. Indeed, recent studiesare revealing surprising patternsthat challenge long-held biogeographical assumptions.

The term ‘biodiversity hotspot’ was coined by NormanMyers3,4 in the late 1980s in two papers that identified 18geographical regions as conservation priorities because theycontained large numbers of endemic species found in rela-tively small areas that were facing significant threats of habi-tat loss. It was reasoned that protecting hotspots, defined inthis manner, should prevent the extinction of a larger number

of species than would protectingareas of a similar size elsewhere.This definition of a hotspot con-tinues to be used in many studiestoday5,6. More generally, however,the term hotspot is now applied toa geographical area that ranks par-ticularly high on one or more axesof species richness, levels ofendemism, numbers of rare orthreatened species, and intensityof threat.

The term biodiversity hotspotis now most commonly used withreference to regions of high spe-cies richness. For example, theGAP analysis program being car-ried out in the United States useshotspots to identify gaps in the ex-isting network of protected areas7,8.This analysis begins by mappinghotspots of species richness, then

determines which species are already well-conserved inexisting protected areas, then maps the pattern of speciesrichness for the remaining species, and, using various selec-tion algorithms, chooses the minimum set of grid cells thatencompass the unprotected species.

Alternatively, hotspots of species rarity or endemism –regions rich in species with restricted distribution ranges –have been used to help set priorities for bird conservation9,10.

Copyright © 1998, Elsevier Science Ltd. All rights reserved. 0169-5347/98/$19.00 PII: S0169-5347(98)01363-9

REVIEWS

Biodiversity hotspotsWalter V. Reid

Hotspots of biodiversity – areasparticularly rich in species, rare species,threatened species, or some combination

of these attributes – are increasinglybeing delineated to help set priorities for

conservation. Only recently have webegun to test key assumptions that

determine how useful a hotspot approachcan be for conservation planning. The

evidence suggests that although at largegeographic scales hotspots do provide

useful information for conservationplanning, at smaller scales their value

may be more limited.

Walter Reid is at the World Resources Institute, 1709 New York Ave NW, Washington, DC 20006, USA

(walt@wri.org).

27 Sarich, V.M. (1993) Mammalian systematics: twenty-five yearsamong their albumins and transferrins, in Mammal Phylogeny:Placentals (Szalay, F.S., Novacek, M.J. and McKenna, M.C., eds), pp. 103–114, Springer-Verlag

28 Lavergne, A. et al. (1996) Interordinal mammalian relationships:evidence for paenungulate monophyly is provided by completemitochondrial 12S rRNA sequences, Mol. Phylog. Evol. 6, 245–258

29 de Jong, W.W., Leunissen, J.A.M. and Wistow, G.J. (1993) Eye lenscrystallins and the phylogeny of placental orders: evidence for amacroscelid–paenungulate clade? in Mammal Phylogeny: Placentals(Szalay, F.S., Novacek, M.J. and McKenna, M.C., eds), pp. 5–12, Springer-Verlag

30 Douzery, E. and Catzeflis, F.M. (1995) Molecular evolution of themitochondrial 12S rRNA in Ungulata (Mammalia), J. Mol. Evol. 41,622–636

31 Lecointre, G. et al. (1993) Species sampling has a major impact onphylogenetic inference, Mol. Phylog. Evol. 2, 205–224

32 Philippe, H. and Douzery, E. (1994) The pitfalls of molecularphylogeny based on four species, as illustrated by theCetacea/Artiodactyla relationships, J. Mammal. Evol. 2, 133–152

33 Naylor, G.J.P. and Brown, W.M. (1997) Structural biology andphylogenetic estimation, Nature 388, 527–528

34 Cao, Y. et al. (1994) Phylogenetic relationships among eutherianorders estimated from inferred sequences of mitochondrialproteins: instability of a tree based on a single gene, J. Mol. Evol.39, 519–527

35 Zardoya, R. and Meyer, A. (1996) Phylogenetic performance ofmitochondrial protein-coding genes in resolving relationshipsamong vertebrates, Mol. Biol. Evol. 13, 933–942

36 Russo, C.A.M., Takezaki, N. and Nei, M. (1996) Efficiencies ofdifferent genes and different tree-building methods in recovering a known vertebrate phylogeny, Mol. Biol. Evol. 13,525–536

37 Swofford, D.L. et al. (1996) Phylogenetic inference, in MolecularSystematics (2nd edn) (Hillis, D.M., Moritz, C. and Mable, B.K., eds),pp. 407–492, Sinauer

38 Shimamura, M. et al. (1997) Molecular evidence from retroposonsthat whales form a clade within even-toed ungulates, Nature 388,666–670

39 Hedges, S.B. and Maxson, L.R. (1996) Molecules and morphology inamniote phylogeny, Mol. Phylog. Evol. 6, 312–314

40 Luckett, W.P. and Hartenberger, J-L. (1993) Monophyly orpolyphyly of the order Rodentia: possible conflict betweenmorphological and molecular interpretations, J. Mammal. Evol. 1,127–147

41 Milinkovitch, M.C. and Thewissen, J.G.M. (1997) Even-toedfingerprints on whale ancestry, Nature 388, 622–624

42 Hartenberger, J.L. (1986) Hypothèse paléontologique sur l’originedes Macroscelidea (Mammalia), C. R. Acad. Sci. Ser. II 302, 247–249

43 Butler, P.M. (1995) Fossil Macroscelidea, Mamm. Rev. 25, 3–1444 Woodall, P.F. (1995) The penis of elephant shrews (Mammalia:

Macroscelidea), J. Zool. 237, 399–41045 Hedges, S.B. et al. (1996) Continental breakup and the ordinal

diversification of birds and mammals, Nature 381, 226–22946 McKenna, M.C. and Bell, S.K. (1997) Classification of Mammals Above

the Species Level, Columbia University Press47 Simpson, G.G. (1945) The principles of classification and a

classification of mammals, Bull. Am. Mus. Nat. Hist. 85, 1–350

TREE vol. 13, no. 7 July 1998

For example, the International Council for Bird Preservation(ICBP), now known as BirdLife International, mapped thedistribution of land bird species with range sizes less than50 000 km2 and subsequently identified 221 areas that con-tained at least two restricted-range species. These areasaccounted for 95% of all restricted-range birds9.

Other studies have defined hotspots based on both spe-cies richness and endemism. Megadiversity countries, forexample, are nations that either have extremely high spe-cies richness of plants and vertebrates (e.g. Brazil, Colombiaand Indonesia) or are relatively less species rich but haveextremely high levels of endemism (e.g. Madagascar andAustralia)11. Finally, hotspots have also been defined as re-gions with the largest number of threatened species, inde-pendent of the overall species richness or endemism of theregion12,13.

Intensive research is currently focused on two aspectsof the analysis of biodiversity hotspots that are key to itsvalue for setting conservation priorities. First, because spe-cies distribution data for most taxa are limited or unavailable,

the use of a hotspot approach in setting priorities rests onthe assumption that patterns of diversity among relativelywell-studied ‘indicator’ groups, such as birds, mammals andplants, are good predictors of patterns of diversity in less-studied groups. Second, the research aims to determine theoptimal method of analysis for using hotspot information insetting conservation priorities.

Overcoming data constraintsDo geographic regions that rank high on a scale of species

richness, endemism or threat in well-known taxa also rankhigh for other taxa? At coarse-grained geographic scales,diversity patterns do correspond across taxa, as shown bythe pattern of increased species richness at lower latitudesfound in many taxonomic groups. Species richness of tigerbeetles (Cicindelidae), for example, is positively correlatedwith measures of bird and butterfly diversity across NorthAmerica, Australia and the Indian subcontinent14 (Fig. 1),although when the effect of latitude is removed the cor-relations are much weaker15. Similarly, correspondence ofpatterns of endemism across taxonomic groups would beexpected at global or continental scales because of the com-mon trend of species range size to decrease as one movesfrom higher to lower latitudes16.

At finer scales of resolution, however, the correspond-ence of patterns of species richness and endemism acrosstaxonomic groups breaks down17. Little concordance isfound among the most species-rich regions in the USA formammals, birds, reptiles, amphibians, tiger beetles andtrees15. Similarly, if the regions highest in species-richness(the top 5% of 10 km grid squares) in the UK are identifiedfor five different taxonomic groups (butterflies, dragonflies,liverworts, aquatic plants and breeding birds), the maxi-mum overlap of diversity hotspots between groups is only34% (between butterflies and dragonflies)18. In South Africa,a similar approach examining overlap in richness and en-demism hotspots among fish, frogs, tortoises, snakes, birdsand mammals found substantial overlap for species rich-ness hotspots between some groups (e.g. frogs and birds,72%; frogs and mammals, 62%) but low overlap for manyother combinations (e.g. tortoise diversity hotspots showedless than 10% overlap with other groups)19. For endemismhotspots, the maximum overlap was only 44% (betweenfrogs and mammals), and for 12 out of 15 combinations ofspecies the overlap was 25% or less (Table 1).

If patterns of species richness and endemism do not cor-respond across taxa at finer scales of resolution, we wouldnot expect hotspots of threatened or endangered species indifferent taxonomic groups to correspond either, which isindeed the case. For example, there is relatively little over-lap among areas (counties) in the USA containing the high-est numbers of federally listed endangered species in differ-ent taxa (molluscs, birds, fish, mammals, arthropods,herptiles)13 (Fig. 2).

There is a good explanation for the weak correspondencebetween patterns of richness and endemism at fine scales ofgeographic resolution, and this explanation suggests thatweaker correspondence would also be expected for morefine-grained taxonomic scales (e.g. genera as opposed toclasses). Fine-scale geographic samples and fine-scale taxo-nomic groups are likely to contain species that share nar-rower habitat requirements; thus, areas of high richness orendemism for those samples are less likely to correspondacross taxonomic groups. Amphibian diversity patterns areless likely to correspond to diversity in gymnosperms, forexample, than vertebrate diversity patterns would be ex-pected to correspond to vascular plant diversity. It is not

REVIEWS

276

Fig. 1. Correlation between the number of tiger beetle and bird species per squareon grids across North America, Australia and the Indian subcontinent. From Ref. 14,with permission.

North America160

120

80

40

0 5 10 15 20 25

r = 0.375n = 208

Australia

200

150

100

0 5 10 15 20 25

r = 0.531n = 67

India

Number of tiger beetle species per square

450

300

150

0 12 24 36 48 60

r = 0.726n = 61

Num

ber

of b

ird s

peci

es p

er s

quar

e

TREE vol. 13, no. 7 July 1998

surprising, therefore, thatstudies are showing that atfine scales of resolution hot-spots of certain taxa do notreflect hotspots of other taxa. For indicator taxa to beuseful in setting priorities, a careful balance needs to be struck between the geo-graphic and taxonomic scaleof resolution, and the type ofconservation question beingaddressed.

Higher-taxon diversity as asurrogate for speciesdiversity

Instead of relying on in-dicator taxa to reflect dis-tribution patterns in lesser-known taxa, other studieshave examined the potentialvalue of other surrogate meas-ures of species richness or endemism, including vegetationclasses, land classes, environmental variables (e.g. pre-cipitation and net primary productivity) and richness athigher taxonomic levels. A number of these surrogates havebeen shown to have useful predictive power, but highertaxon-diversity appears to be most closely correlated topatterns of species diversity20. Most of this research hasfocused on patterns of species richness.

At both coarse and fine geographic scales, richness ofgenera and families have proved to be relatively good pre-dictors of species richness21,22. For example, at a continentalscale (using 611 000 km2 grid cells), 99% of the variation inbird species richness in North America can be explained bygenus richness, and 91% of variation by family richness20.Similarly, at a fine scale in 35 forest reserves ranging in sizefrom 18–30 000 ha in Sri Lanka, 96% of woody-plant speciesrichness can be explained by genus richness, and 86% ofvariation by family richness22.

Can higher-taxon richness patterns be combined to cre-ate even better surrogates for overall species diversity? Thechallenge here is to determine how best to weight numbersof genera or families in different taxa, because higher taxo-nomic groups may be defined in different ways, particularlywhere the taxa are distantly related. In one coarse-grained(global) study examining different methods for combiningfamily richness in plants, reptiles, amphibians and mammals(absolute number of families; proportional family richness;and proportional family richness weighted for total speciesrichness in each group), all three methods gave fairly simi-lar results23. At other scales of analysis or using other taxo-nomic groups, however, the different methods would be ex-pected to produce different patterns. Which method to usedepends on the objective of the analysis and the availabilityof data (e.g. good estimates of total species richness mightnot be available for some taxa).

Patterns of higher taxon richness can therefore serve assurrogates for species richness. It remains to be seen whetherthis same correspondence holds for patterns of endemism.The utility of this surrogate measure for conservation plan-ning hinges on the seemingly reasonable but untested as-sumption that if only a partial survey of a region has beenundertaken, the cumulative list of higher taxa encounteredwill converge on the total number of taxa more rapidly thanwill the cumulative list of species.

Setting conservation prioritiesSetting global priorities

Biodiversity hotspot analysis was originally used toidentify large regions, typically the size of an entire nation,that deserved conservation attention, such as Madagascar,Northern Borneo, or the Philippines. The 18 global hot-spots defined by Norman Myers have since been used by theMacArthur Foundation to target its grantmaking. Organ-izations including the World Wide Fund for Nature-India andConservation International also set priorities among coun-tries or regions using Myers’ rankings.

The studies discussed here provide reasonable supportfor the utility of hotspots that are defined by using speciesrichness and endemism at a global or continental scale ofanalysis. At this coarse scale, patterns of richness and en-demism tend to correspond reasonably well across taxa. Byits nature, a continental or global hotspot analysis tends toidentify regions with relatively small overlap in species com-position – Madagascar is unlikely to share many species withBorneo, for example.

Using hotspots to guide conservation decisionsAt more fine-grained geographic scales, which tend to be

the scales at which conservation and development decisionsare made, the application of hotspot analysis is more chal-lenging19,24,25. To begin with, the poor correspondence be-tween species hotspots in different taxa implies that prior-ities determined on the basis of either richness or endemismin a few taxonomic groups cannot be relied upon to capturesimilar patterns in other groups.

Just as important, the choice of method for defining a hot-spot – whether it is based on richness, endemism, threat or acombination of these factors – significantly influences whichregions or sites are identified as conservation priorities.

Site selection algorithms based on identifying hotspotsof species richness tend to be the least efficient in maximiz-ing the protection of species diversity. This is becausehotspots of species richness do not often include relativelyrare species – hotspots that are ranked highest for richnessoften contain overlapping sets of common species, whilefailing to capture rarer species. In the UK, 43% of rare birdspecies do not occur in the top 5% of species richness hot-spots18. In South Africa, the top 5% of richness hotspots forfish only include 66% of the total diversity of fish species19.

REVIEWS

277

Table 1. The proportional overlap of (a) hotspots (above diagonal), cold spots (below diagonal), and (b) endemic species hotspots,

among six vertebrate taxa in greater South Africaa,19

Fish Frogs Tortoises Snakes Birds Mammals

(a) Hotspots and coldspotsFish – 0.35 (31) 0.00 (5) 0.35 (31) 0.48 (31) 0.36 (28)Frogs 0.24 (110) – 0.00 (9) 0.49 (47) 0.72 (47) 0.62 (47)Tortoises 0.54 (98) 0.61 (128) – 0.10 (10) 0.00 (12) 0.08 (12)Snakes 0.19 (118) 0.25 (153) 0.62 (133) – 0.59 (61) 0.42 (59)Birds 0.67 (3) 1.00 (2) 0.88 (32) 0.47 (19) – 0.47 (68)Mammals 0.19 (144) 0.22 (172) 0.65 (155) 0.34 (177) 0.41 (39) –

(b) Endemic hotspotsFish – 0.00 (21) 0.00 (7) 0.14 (14) 0.18 (22) 0.19 (16)Frogs – 0.20 (10) 0.25 (24) 0.42 (33) 0.44 (18)Tortoises – 0.00 (10) 0.20 (15) 0.22 (9)Snakes – 0.24 (25) 0.16 (19)Birds – 0.40 (20)Mammals – –

aNumbers in parentheses are total possible overlaps.

TREE vol. 13, no. 7 July 1998

The failure of species richness hotspots to capture rarespecies is scale dependent. As the size of the sample unitincreases, an increasing number of rare species will be foundin richness hotspots because the sample unit will ‘need’ rarespecies to be unusually species rich. This has been suggestedas one explanation for why nearly twice as many hotspots ofbird species richness in Australia (defined using 100-km2

grid cells) hold at least one rare bird species compared withhotspots in the UK (defined using 10-km2 grid cells)26. Simi-larly, although endemism or rarity and species richness areonly weakly correlated at national scales27, at a continentalscale, patterns of richness and endemism coincide28.

A more efficient method for site selection relies on choos-ing hotspots of rarity or endemism – sites that are richest inspecies with the most restricted ranges24. Because sites thatcontain species with narrow distributions must be includedin a protected area network in order to cover all species, thehigh priority that rarity-based algorithms give these sitestends to reduce the number of total sites needed.

A still more efficient mechanism for maximizing thenumber of species protected in a given land area is to usemethods of complementarity, where the species content ofexisting reserves is identified and then further sites are se-lected in a stepwise fashion to add areas that contribute thegreatest number of new species24,25,29. These selection algo-rithms are most efficient (i.e. capture all species in the small-est number of sites) when they begin with sites containingspecies found nowhere else (and, therefore, which often

have relatively low species richness) rather than with themost species-rich locations.

Finally, the most efficient site selection is achieved usinga maximal-covering-location model which uses integer linearprogramming methods to choose simultaneously the optimalset of sites24,30. Such methods, however, are prohibitive forlarge datasets and also fail to provide information on the bestsequence for adding new sites to an existing reserve networkso as to provide the greatest marginal gain with each newsite31.

Comparisons among site selection algorithms tend to bebased on the algorithms’ efficiency in capturing all species ina given region. Using this criterion of complete species cov-erage, hotspots of richness or endemism prove to be a lessefficient means of setting conservation priorities thanapproaches using complementarity. However, in the realworld of conservation decision-making, a more relevantquestion to ask often is: given the ability to protect only asmall number of sites, what proportion of overall diversityis captured in priority areas that are identified on the basisof the various methodologies? Where the number of sites islimited, richness-based approaches perform relatively well24.

The utility of richness- or endemism-based hotspots forconservation decision-making can also be enhanced by com-bining hotspot analysis with an attempt to protect repre-sentative samples of ecosystems. The World ConservationUnion, for example, has recommended a minimum conser-vation goal of protecting 10% of every biogeographic region.

REVIEWS

278

Fig. 2. The geographic distribution of four groups of endangered species in the United States. (a) Plants, (b) birds, (c) fish and (d) molluscs. The maps illustrate thenumber of Federally listed endangered species in each country. Alaska and Hawaii are shown in the bottom left-hand corner of the maps (not to scale). From Ref. 13,with permission.

Number of plants

Zero/no dataOneTwo3 to 45 to 78 to 1819 to 77

Number of birds

Zero/no dataOneTwoThreeFour5 to 13

Number of fish

Zero/no dataOneTwoThreeFour5 to 7

Number of molluscs

Zero/no dataOneTwoThree4 to 56 to 14

(b)(a)

(c) (d)

TREE vol. 13, no. 7 July 1998

This objective helps to reduce the potential overlap in speciescomposition if only a limited number of protected areas areto be established. For example, if we were to identify the toptwo hotspots of species richness in a region encompassingtwo biogeographic regions, both hotspots could fall within asingle region and the overlap in species composition couldbe quite high. If, instead, the top hotspot in each region wasselected – that is, representative samples of the ecosystemswere chosen – then overlap in species lists would decrease,and the total number of species protected could increase.When only richness or endemism hotspots are used in theanalysis, protection of representative samples using thismethod helps to increase the complementarity of site selec-tion. World Wildlife Fund-USA, for example, recently identi-fied 232 conservation priorities by first stratifying data intomajor habitat types and biogeographic realms and then ex-amining species richness and endemism (and other factors)to identify relatively rich ecoregions within each group32.

No matter which approach to priority setting is used,however, if the areas conserved capture the diversity of theindicator taxa only and not that of the more poorly-knowntaxa, then the use of the method for conserving biodiversityis questionable. Recent studies, demonstrating the failure ofhotspots of richness and endemism to correspond acrossdifferent taxa, seem to undermine the utility of hotspots inpriority setting. However, on closer examination the messageof these studies is not so bleak: these same studies lend sup-port to the hypothesis that a set of areas in which one ma-jor taxon is well represented can also represent diversity inunrelated taxa.

For example, in the UK, a recent study found that thegreatest overlap in hotspots among five taxonomic groupswas only 34%. However, if it were possible to protect everyhotspot designated for birds (that is, the top 5% of grid cellswith the most species of birds), then 87% of birds, 100% ofbutterfly species and over 90% of dragonflies, liverworts andaquatic plants would be encompassed18. Similarly, it appearsthat a set of areas in Oregon, USA, which completely repre-sents one major taxon does a good job of representing di-versity in others24. This result is not surprising. Any site se-lection approach, whether it is based on richness, rarity orcomplementarity, that captures most of the diversity in onetaxon is likely to include a diversity of habitats and, there-fore, capture a substantial amount of the diversity of othertaxa as well. This is the case even if the areas of high rich-ness or endemism do not correspond.

How hot are hotspots?One of the most important lessons that we are learning

from the study of diversity hotspots is that the extent of thebiodiversity crisis is often highly localized. Although hot-spots do not always correspond across taxa, a substantialfraction of species diversity can be found in very small re-gions and most threatened species can be found in smallerregions still. Rarity hotspots covering just 5% of the UK rep-resent 98% of British species of breeding birds25. Richnesshotspots in the UK covering a similar area encompass 91%of butterflies, 92% of dragonflies, 95% of liverworts, 96% ofaquatic plants and 87% of breeding birds18. Within the coter-minous US, more than 50% of endangered species of plants,birds, fish and molluscs are found in less than 2% of the landarea13. This is not to say that these hotspots are sufficientlylarge to maintain viable populations of all the species foundin them, but it does re-affirm the emphasis that conser-vation strategies place on protected areas: significant speciesdiversity can be encompassed in relatively small fractionsof the landscape.

Hotspot analysis is one of a set of different tools nowavailable to help set priorities for conservation planning. Dif-ferent methods of analysis of hotspot information have dif-ferent strengths depending on the conservation goal beingpursued33 and the availability of data. For example, the ana-lytic method chosen to maximize the number of survivingspecies would not be the best approach for maximizing thegenetic or taxonomic diversity of surviving species34 ormaintaining critical ecosystem functions. Although the util-ity of a hotspot approach to setting conservation prioritiesis probably greatest at relatively coarse spatial scales, a hot-spot analysis can be useful at finer scales, particularly if usedin combination with other analytical methods.

References1 Andrewartha, H.B. and Birch, L.C. (1954) The Distribution and

Abundance of Animals, University of Chicago Press2 Wallace, A.R. (1876) The Geographic Distribution of Animals,

Macmillan3 Myers, N. (1989) Threatened biotas: “Hotspots” in tropical forests,

Environmentalist 8, 1–204 Myers, N. (1990) The biodiversity challenge: expanded hotspots

analysis, Environmentalist 10, 243–2565 Sisk, T.D. et al. (1994) Identifying extinction threats: global

analyses of the distribution of biodiversity and the expansion ofhuman enterprise, BioScience 44, 592–604

6 Médail, F. and Quézel, P. (1997) Hot-spots analysis for conservationof plant biodiversity in the Mediterranean basin, Ann. MO Bot.Gard. 84, 112–127

7 Scott, J.M. et al. (1993) Gap analysis: a geographic approach toprotection of biological diversity, Wildl. Monogr. 123, 1–41

8 Kiester, A.R. et al. (1996) Conservation prioritization using GAPdata, Conserv. Biol. 10, 1332–1242

9 Bibby, C.J. et al. (1992) Putting Biodiversity on the Map: Priority Areasfor Global Conservation, International Council for Bird Preservation

10 Balmford, A. and Long, A. (1994) Avian endemism and forest loss,Nature 372, 623–624

11 McNeely, J.A. et al. (1990) Conserving the World’s BiologicalDiversity, World Resources Institute

12 Flather, C.H., Knowles, M.S. and Kendall, I.A. Threatened andendangered species geography: characteristics of hot spots in theconterminous United States, BioScience (in press)

13 Dobson, A.P. et al. (1997) Geographic distribution of endangeredspecies in the United States, Science 275, 550–553

14 Pearson, D.L. and Cassola, F. (1992) World-wide species richnesspatterns of tiger beetles (Coleoptera: Cicindelidae): indicatortaxon for biodiversity and conservation studies, Conserv. Biol. 6,376–391

15 Flather, C.H. et al. (1997) Identifying gaps in conservationnetworks: of indicators and uncertainty in geographic-basedanalyses, Ecol. Appl. 7, 531–542

16 Lawton, J.H. (1994) Population dynamic principles, Philos. Trans. R.Soc. London Ser. B 334, 61–68

17 Prendergast, J.R. (1997) Species richness covariance in highertaxa: empirical tests of the biodiversity indicator concept,Ecography 20, 210–216

18 Prendergast, J.R. et al. (1993) Rare species, the coincidence ofdiversity hotspots and conservation strategies, Nature 365,335–337

19 Lombard, A.T. (1995) The problems with multi-speciesconservation: do hotspots, ideal reserves and existing reservescoincide? S. Afr. J. Zool. 30, 145–163

20 Gaston, K.J. and Blackburn, T.M. (1995) Mapping biodiversity usingsurrogates for species richness: macro-scales and New Worldbirds, Proc. R. Soc. London Ser. B 262, 335–341

21 Balmford, A., Green, M.J.B. and Murray, M.G. (1996) Using higher-taxon richness as a surrogate for species richness: I. Regionaltests, Proc. R. Soc. London Ser. B 263, 1267–1274

22 Balmford, A., Jayasuriya, A.H.M. and Green, M.J.B. (1996) Usinghigher-taxon richness as a surrogate for species richness: II. Localapplications, Proc. R. Soc. London Ser. B 263, 1571–1575

REVIEWS

279

280 TREE vol. 13, no. 7 July 1998

Extra-pair paternity inbirds: ‘good-genes’ andsomething else

In a recent TREE article, Petri and Kempenaers1

reviewed determinants and variation of extra-pairpaternity (EPP) in birds, emphasizing the ‘good-genes’ hypothesis: females paired torelatively ‘poor-quality’ males will seek extra-paircopulations (EPCs) from males of better quality toimprove the fitness of some of their offspring1.This is a widely accepted hypothesis for EPP inbirds and assumes a genetic basis for rankingquality traits and related fitness (good genes). Ifthe task of females is to select the highest rankingmale possible, they should not seek EPP whentheir social mate is a ‘good-quality’ mate.

However, increasing evidence indicates thatfemales often do so2,3, which suggests that(according to this hypothesis) they can stilldiscriminate some differences. This is unlikelybecause the amount of phenotypic variabilityresulting from gene–environment covariance andinteraction is not negligible4,5 and should masksmall genetic differences. In addition, male qualityis often not correlated with their ability to obtainextra-pair fertilizations2 and males bearingrelatively large ornaments can be cuckolded3. Theauthors also attributed low proportions of EPP inbottleneck populations to the ‘good genes’hypothesis because the proportion of EPP and theamount of variation in male quality are correlated.However, there are alternative explanations for

this association if reduced heterozygosity isdirectly responsible for lower reproductiveperformance6.

In the ‘heterozygosity’ theory7, the explanationfor sexual selection and EPP is not so much in‘better’ genes as in greater heterozygosity. Thistheory has generated new insights into the studyof EPP and explains a wider range of situations inwhich female birds obtain EPP. By masking lethaland sublethal genes, heterozygosity enhances theexpression of genes related to fitness (whichdetermine the extreme expression of sexuallyselected male traits7). Therefore, females shouldseek EPCs with other ‘good-quality’ (heterozygous)males to improve the genetic diversification(quality) of some of their offspring7. Also, unlikethe ‘good-genes’ hypothesis, the ‘best’ male forone female might not be the ‘best’ for another ifthere are fertility costs to mating, such as highgenetic similarity1, reduced male fertility1,6 orgenetic incompatibility8,9. By seeking EPCs withother high-quality (heterozygous) males, therefore,females obtain an extra benefit assuring, orenhancing, the fertilization of the clutch. Petri andKempenaers1 did not pay much attention to thefertility costs associated with mating in spite ofthe increased interest in this topic in multiplemating strategies9. They are right to argue thatthere is a lack of evidence for birds; however, thismay be because little effort has been made in this direction.

Future research should include not onlyobservational and comparative studies1, but alsoanalyses of genetic correlates of quality and costsof mating with proportions of EPP. This can betested using the major histocompatibility complex

(MHC)8,10. If the presence of MHC haplotypespredict genetic costs or assortative mating8,10, wecould test hypotheses for mating decisions and tryto understand the variation of EPP in birdpopulations from an integrated perspective.

Pedro J. Cordero

Museo Nacional de Ciencias Naturales,Consejo Superior de Investigaciones Científicas, José Gutierrez Abascal 2, 28006 Madrid, Spain(mcnctr@fresno.csic.es)

References1 Petrie, M. and Kempenaers, B. (1998) Trends

Ecol. Evol. 13, 52–582 Bridgot, J.M. et al. (1997) Behav. Ecol. Sociobiol.

40, 119–1263 Cordero, P.J., Wetton, J.H. and Parkin, D.T.

J. Avian Biol. (in press)4 Bailey, R.C. (1997) Genetica 99, 125–1335 Veiga, J.P. and Puerta, M. (1996) Proc. R. Soc.

London Ser. B 263, 229–2346 Charlesworth, D. and Charlesworth, B. (1987)

Annu. Rev. Ecol. Syst. 18, 237–2687 Brown, J.L. (1997) Behav. Ecol. 8, 60–658 Svensson, E. and Skarstein, F. (1997) Trends

Ecol. Evol. 12, 92–939 Jennions, M.D. (1997) Trends Ecol. Evol. 12,

251–25310 Potts, W.K. and Wakeland, E.K. (1990) Trends

Ecol. Evol. 5, 181–187

Reply from M. Petrie andB. Kempenaers

In his comment on our review, Cordero claims thatwe put too much emphasis on the ‘good genes’hypothesis to explain variation in extra-pairpaternity in birds. He suggests that alternativehypotheses, based on heterozygosity andinfertility, should also be considered and thatsome of the data may be better explained by thesealternatives.

Cordero states that ‘there is increasingevidence that females mated to high quality males

CORRESPONDENCE

Copyright © 1998, Elsevier Science Ltd. All rights reserved. 0169-5347/98/$19.00

23 Williams, P.H., Gaston, K.J. and Humphries, C.J. (1997) Mappingbiodiversity value worldwide: combining higher-taxon richnessfrom different groups, Proc. R. Soc. London Ser. B 264, 141–148

24 Csuti, B. et al. (1997) A comparison of reserve selection algorithmsusing data on terrestrial vertebrates in Oregon, Biol. Conserv. 80,83–97

25 Williams, P. et al. (1996) A comparison of richness hotspots, rarityhotspots, and complementary areas for conserving diversity ofBritish birds, Conserv. Biol. 10, 155–174

26 Curnutt, J. et al. (1994) Hotspots and species diversity, Nature 367,326–327

27 Ceballos, G. and Brown, J.H. (1995) Global patterns of mammaliandiversity, endemism, and endangerment, Conserv. Biol. 9, 559–568

28 Blackburn, T.M. and Gaston, K.J. (1996) Spatial patterns in thespecies richness of birds in the New World, Ecography 19, 369–376

29 Pressey, R.L. et al. (1993) Beyond opportunism: key principles forsystematic reserve selection, Trends Ecol. Evol. 8, 124–128

30 Church, R.L., Stoms, D.M. and Davis, F.W. (1996) Reserve selectionas a maximal covering location problem, Biol. Conserv. 76, 105–112

31 Pressey, R.L., Possingham, H.P. and Margules, C.R. (1996)Optimality in reserve selection algorithms: when does it matterand how much? Biol. Conserv. 76, 259–267

32 Olson, D.M. and Dinerstein, E. (1997) The Global 200: A Representation Approach to Conserving the Earth’s DistinctiveEcoregions, World Wildlife Fund

33 Reid, W.V. (1994) Setting objectives for conservation evaluation, inSystematics and Conservation Evaluation (Forey, P.I., Humphries, C.J.and Vane-Wright, R.I., eds), pp. 1–13, Oxford University Press

34 Vane-Wright, R.I., Humphries, C.J. and Williams, P.H. (1991) What toprotect? – systematics and the agony of choice, Biol. Conserv. 55,235–254

Letters to TREECorrespondence in TREE may address topics raised in very recent issues of TREE, or (occasionally) other matters of general current interest to ecologists andevolutionary biologists. Letters should be no more than 500 words long with amaximum of 12 references and one small figure; original results, new data or newmodels are not allowed. Letters should be sent by e-mail to TREE@elsevier.co.uk.The decision to publish rests with the Editor, and the author(s) of any TREE articlecriticized in a Letter will normally be invited to reply. Full-length manuscripts inresponse to previous TREE articles will not be considered.