Elephants versus butterﬂies: the ecological role of large ...

Elephants versus butterflies: the ecological role oflarge herbivores in the evolutionary history oftwo tropical worldsCris Cristoffer1* and Carlos A. Peres2 1CES/CEVN, Luke Air Force Base, AZ, USA, 2Centre

for Ecology, Evolution & Conservation, University of East Anglia, Norwich, UK

Abstract

Aim Large herbivores have important effects upon Paleotropical ecosystems, but attainmuch lower biomass densities in the Neotropics. We assess how this difference in her-bivore activity has generated different ecological and evolutionary trajectories in theNew and Old World tropics. We also propose an explanation for how the greaterbiomass density in the Old World came about.

Location Data were compiled primarily from moist tropical forests, although more ofthe relevant information to address most of our hypotheses was available from themainland areas of Africa, Asia, and South America than elsewhere.

Methods We gleaned data from published information and personal communication.We compared body masses and a variety of other types of information for the New- andOld-World tropics. We proposed that interhemispheric differences exist in a variety ofprocesses, including herbivory, frugivory, and flower visitation. We erected hypothesesand evaluated them qualitatively, and, when information was available, tested themusing simple ratios of species in various taxonomic and trophic categories. To make thecomparisons more meaningful, we specified appropriate data selection criteria.

Results A general pattern of differences emerges from this review. Compared withNeotropical forests, the much greater biomass densities of large herbivores in Paleo-tropical forests are associated with a lesser diversity of small herbivores, differenthunting methods used by indigenous humans, larger arboreal vertebrates, larger fruits,different patterns of fruit and flower dispersion in space and time, a lesser abundance ofmost types of reproductive plant parts, and other features. The existence of a species-richfauna of large herbivores in the pre-Holocene Neotropical rain forest was not supported.

Main conclusions The potential for large herbivores to cause functional differencesbetween the New and Old World tropical forests has been virtually unexplored, despitethe well-known importance of large herbivores in the Old World tropics. The evalua-tions of our hypotheses suggest that the abundance of large herbivores in the Old Worldtropics has launched it onto a different evolutionary trajectory than that of theNewWorld tropics. The relevant evidence, although scanty, suggests that the inter-hemispheric ecological differences are not an artefact of recent megafaunal extinctions inthe New World. Recent human activities have, however, reduced population sizes oflarge wild herbivores in the Old World, and increased population sizes of livestock. Thishas likely created a rather homogeneous, anthropogenic selection pressure that tends toerase the evolutionary differences between the two tropical worlds.

Keywords

Macroevolution, Neotropics, Paleotropics, herbivores, rain forest, savanna, frugivory,flower visitors.

*Correspondence: Cris Cristoffer, 56 CES/CEVN, 13970 W. Lightning Street, Luke Air Force Base, AZ 85309-1149, USA. E-mail: [email protected]

Journal of Biogeography, 30, 1357–1380

� 2003 Blackwell Publishing Ltd

INTRODUCTION

Biogeographers have long been aware of taxonomic differ-ences between the biotas of the Neotropics and the Paleo-tropics. Taxa exclusive or nearly exclusive to onehemisphere include bromeliads, cacti, leafcutter ants, mor-pho butterflies, hummingbirds, toucans, hornbills, sunbirds,marmosets, elephants, rhinos, prosimians, prehensile-tailedmonkeys, true apes, gliding rodents (Anomaluridae or Sci-uridae), among many others. Indeed, pairwise comparisonsof some New and OldWorld taxa have often been used toexemplify convergent evolution (Bourliere, 1973; Eisenberg,1981; Terborgh & van Schaik, 1987).

Convergence may not be as close as it appears, but thesediscrepancies can be explained. We suggest that the Paleo-tropics are rich in large mammals whose closest counterpartsin the Neotropics are small mammals, reptiles and inverte-brates. Furthermore, there are fundamental interhemisphericdifferences in frugivory and flower visitation. We posit thatmany of the differences between the Neotropics and thePaleotropics have a common underlying cause: the abun-dance of large, tropical rain forest herbivores (hereafter,LRFH). Since the largest tropical herbivores are all butimmune to attack from predators (Owen-Smith, 1988), theirbiomass densities under non-hunted conditions may often begreater than for smaller, predated herbivores. It seemsplausible that the lesser abundance of LRFH and totalabsence of very large herbivores in the Neotropics, com-pared with the Paleotropics (Fa & Purvis, 1997), hasresulted in functionally different rain forest ecosystems.

Our primary focus is on lowland tropical rain forest; forour purposes, the maps in Whitmore (1998) depict the areaswhere these occur. Rain forest landscapes include a varietyof vegetation formations of smaller extent, including heathforests, isolated savannas, etc.; however, our analysis willnot address such relatively small areas. There are alsoundoubtedly quantitative abiotic differences between sitesthat might confound hypotheses of the factors that deter-mine species composition. Nevertheless, since large herbiv-ores have been widespread and occurred at high biomassdensities in the Paleotropics (see below) until recently, theconsequences of their activities should be observable.Indeed, at least one author has invoked the activities of largeherbivores as agents of climate change (Retallack, 2001).

There also seems to be a widely-held misconception thatlarge herbivores are present predominantly in savanna andgrassland habitats, but not in rain forest. There is a grain oftruth to this idea; for example, approximately twenty-ninespecies of ungulates from ten genera occur in African forests,compared with forty-three species from twenty-eight generain African savannas (Owen-Smith, 1982). Nevertheless, on alocal scale, forest ungulate communities can be very diverse,including up to sixteen co-occurring species in some Africanforests, compared with about nineteen species (excludingelephants) in the most diverse savanna areas (McNaughton& Georgiadis, 1986). Although biomass data for Asianforests are scarcer, we will present other informationindicating that Asian herbivores affect the forests they inhabit.

By contrast, large herbivores are not typical of tropicalrain forests outside of Asia and Africa. Indeed, part of per-ception that large herbivores may be unimportant in rainforests probably stems from the fact that much tropical rainforest research has taken place in the Neotropics, wherelarge herbivores are nearly absent. Another source of biasmay be because of a simple lack of information and severedifficulties in studying LRFH (e.g. Strickland, 1967; Tang-ley, 1997; Walsh & White, 1999).

Furthermore, several LRFH are in danger of extinctionand have been missing for decades from much of their for-mer geographical ranges, hence researchers have been unableto assess their ecological importance. For example, India hasbeen the site of a flurry of research activity pertaining tolarge mammals, yet several species of large herbivores thatdid occur in Indian tropical forests in the 1800s, includingSumatran rhinos (Dicerorhinus sumatrensis) and Javan rhi-nos (Rhinoceros sondaicus), are now missing (Groves, 1967;Nowak, 1999; Toon & Toon, 2002). Later we will presentinformation about these species as pertains to the hypothesesraised; for the moment, let us note that there are only a fewhundred individuals of both species in the wild. These spe-cies are primarily browsers or mixed feeders, not grazers,and rain forests, not grasslands, are their primary habitat. InIndochina a third large odd-toed herbivore, the Malayantapir (Tapirus indicus), also occurs (Nowak, 1999). There-fore, this review pertains to how LRFH affected their envi-ronment over evolutionary time, rather than during the mostrecent centuries of reduced range and diversity.

Note also that the great majority of recent studies on largepaleotropical herbivores have taken place in habitats wherethe animals are more readily observed, such as Africansavannas and the monsoon forests and alluvial grasslands ofIndia. Thus, our perceptions of rain forest herbivores havebeen coloured by non-forest species. But make no mistake:absence of evidence is not evidence of absence. Rain forestherbivores existed whether or not we happened to studythem. Fortunately, recent studies have begun to lift the veilof mystery surrounding these animals; for example, it is nowcommonly accepted that African elephants probably belongto two or more species, one of which, Loxodonta cyclotis, isa tropical forest specialist (Tangley, 1997). Fenton et al.(1998) reported that the woodland at half of the sites theystudied had been disturbed by high elephant densities to theextent that the tree canopy was greatly reduced. There weresignificant differences between intact and impacted sites withregards to diversity, numbers, activity, and diets of bats.

Other studies have revealed that elephants have played asignificant role in shaping West African rain forest veget-ation (Hawthorne & Parren, 2000). Furthermore, forestelephants make up 52% of the biomass of mammalianherbivores in the rain forest of Gabon and play a key role inAfrican rain forest ecosystems (Prins & Reitsma, 1989).Later in this paper, we will summarize other natural historyliterature documenting the importance of large herbivores inrain forests.

There are undoubtedly quantitative abiotic differencesbetween sites that might confound hypotheses of the factors

� 2003 Blackwell Publishing Ltd, Journal of Biogeography, 30, 1357–1380

1358 C. Cristoffer and C. A. Peres

that determine species composition. Nevertheless, since largeherbivores have been widespread in the Paleotropics untilrecently, the consequences of their activities should beobservable.

Herein we define �large herbivores� to include terrestrialmammals greater than 5 kg in weight, and that obtainmost of their diet from vegetative plant parts. Phytopha-gous insects, on the other hand, are referred to as �smallherbivores�. Rodents of the family Echimyidae that weightless than 1 kg in weight and feed upon vegetative plantparts will also be considered small herbivores, as will thehoatzin (Opisthocomus hoatzin), which is volant. We alsoexclude arboreal folivores larger than c. 2 kg andbelonging to several classes, because they face fundamentalecological problems different from those of large, terrest-rial ungulates (Eisenberg, 1978). Throughout this paper,we will emphasize a comparison of the mainland Neo-tropics with the mainland Paleotropics. Some of ourhypotheses are less applicable to tropical islands, includingMadagascar and New Guinea, according to the rationalewe describe below. We will, however, revive discussion ofsuch places when appropriate data are available to makecomparisons.

Densities of LRFH and their most apparent effects

The largest terrestrial herbivores in the Neotropics, genusTapirus, are less than 10% of the mass of the largestPaleotropical herbivores, but are also selective, if general-ized, herbivores. Tapirs do not occur in Africa, but the okapi(Okapia johnstoni) seems similar in some respects.

In the Paleotropics large herbivores have a greater range inbody size, attain greater biomass densities, and are morediverse. Using data from Lekagul & McNeely (1977);Kingdon (1997) and Fonseca et al. (1996), we compiled dataon the present-day non-volant mammal fauna in tropicalforest regions of Africa, southeast Asia, and South America.This comparison indicates that the mean body mass ofmammals in African forests (37.45 � 17.19 kg; n ¼ 284)and Thailand (98.45 � 43.26 kg; n ¼ 112) is significantlygreater than those in Amazonian forests (4.80 � 1.44 kg;n ¼ 192). This is largely because the number of large-bodiedspecies is considerably greater in Afrotropical and southeastAsian forests than in the Neotropics. In African forests, 60%of species are larger than 1 kg, and 22% are larger than10 kg. The corresponding figures for Thai forests are verysimilar at 62% and 20%, respectively. In contrast, only 38%and 7% of the non-volant mammal fauna found in theAmazon exceed a body mass of 1 and 10 kg, respectively.The weights of the sixty-six species of mammalian primaryconsumers of a forest in north-eastern Gabon are uniformlydistributed across five orders of magnitude (Emmons et al.,1983), whereas those of a typical terra firme forest of centralAmazonia are markedly skewed towards small- and mid-sized species (Peres, 1999). We excluded bats because theirupper size limit is probably far more constrained by flightrequirements (Brown & Maurer, 1987; Cristoffer, 1991);nevertheless, since the size spectrum of Paleotropical bats is

greater than that of Neotropical bats, their inclusion wouldhave further augmented this discrepancy.

The greater abundance of large animals in the Paleotropicshas also influenced indigenous hunting tactics. Snares andother prey capture devices effective on large prey species arecommon in the Old World, but rare in the Neotropics (Fa &Peres, 2001). Indeed, the relatively high abundance ofungulates such as small bovids in west African forests,sustains thriving bushmeat markets, a phenomenon that isvirtually absent from Neotropical forests (Peres, 2000; Fa &Peres, 2001).

The extent to which LRFH affect their environment couldvary with their species richness. Thus, we hypothesized thatthe density of trees would correlate inversely with the speciesrichness of large herbivores, which can injure or kill them.We used tree data from forest plots of 0.5–2.0 ha to partiallycontrol for any effects of sampling area, and we averagedvalues for eight sites in South America, five sites in Africa,two sites in mainland tropical Asia, and four sites in islandtropical Asia, and New Guinea/Australia (Richards, 1996).Within Asia and Australia, the species richness of largeherbivores declines in the direction from mainland Asia,through large islands in Malaysia and Indonesia, to NewGuinea–Australia (Flannery, 1995; Francis, 2001). Althoughthe data points were too few for rigorous statistical testing,there was a monotonic trend of decreasing tree abundancewith increasing number of large herbivore species. Africahad the fewest trees per hectare, and within Asia, tree den-sities increased from the mainland, through Indonesianislands, to New Guinea–Australia.

In theory, we could investigate the effects of large her-bivores by means of an experiment in which biotas at severalsites were randomly split into control and treatment groupsthat were identical except for large herbivores, which wouldbe abundant on control sites, and absent or uncommon ontreatment sites. Because this is unfeasible, we considerednatural experiments (sensu Diamond, 1996) by comparingvarious tropical locations. Natural experiments differ fromfield and laboratory experiments in that the experimenterdoes not establish the perturbation but instead selects siteswhere the perturbation is operating or has already run. Wesought sites as similar as possible, both ecologically andtaxonomically, in all variables except the abundance of largeherbivores. Thus the mainlands of both the Neotropics andPaleotropics have assemblages of anthropoids (monkeys orapes), megabeaked birds (toucans or hornbills), woodpeck-ers, trogons, parrots, specialized flower-feeding birds (hum-mingbirds or sunbirds), bees, butterflies, etc. Many islandswould violate this similarity condition, because islands havedisharmonic biotas with reference to the mainland (Carl-quist, 1965, 1974; Eisenberg, 1981). For example, prior tohuman colonization, the Hawaiian Islands had no ants,monkeys, toucans or hornbills, parrots, or trogons, and therewere few butterflies and bees – taxa that figure in our theoryand are common in the mainland. We also exclude theAustralian tropics, which have a similar disharmonic biotato that of New Guinea (Carlquist, 1965, 1974; Eisenberg,1981).


Macroevolutionary role of large herbivores 1359

The importance of large herbivores in the mainlandPaleotropics has been long recognized. Large herbivores,especially elephants, have been implicated in blatant eco-system alterations (Mueller-Dombois, 1972; Owen-Smith,1988; Dinerstein, 1992; Power & Tilman, 1996; Cumming& Brock, 1997; Keesing, 2000). Prior to the invasion byEuropeans and their weapons, large herbivores were likely tohave been a dominant influence on ecosystem structure anddynamics (Owen-Smith, 1988). Owen-Smith (1988) sum-marized the feeding behaviour of very large Paleotropicalherbivores. Most of the information available is for ele-phants, particularly those in African savannas. Althoughelephants in general are mixed feeders on grass, browse andfruit, grass is insignificant in the diets of forest elephants inGhana and Ivory Coast, where woody browse and fruits arethe main components. Browsing elephants strip leaves andbreak small branches, and the woody material ingested fromsome trees outweighs the foliage, and small woody plants areoften eaten whole. Nearly all browsing occurs below 2 mabove ground, but trees taller than 6 m may be pushed overto bring high branches within reach. Trees greater thanabout 25 cm in diameter commonly withstand attempts topush them over. Asian elephants appear to feed on grass to agreater extent than the African forest species, and to pushover trees less often. Nevertheless, the damage that elephantscause to structural features of vegetation during episodes oflow rainfall may leave a lasting impact on vegetationstructure, thereby altering habitat for many other species.

Only recently has extensive information about Africanforest elephants become more readily available. Forest ele-phants alter the structure of their habitats by creating anetwork of paths and �boulevards� that are regularly used forlong distance migration or foraging (Turkalo & Fay, 1995;Vanleeuwe & Gautier-Hion, 1998). We will not describe indetail the role of elephants and other large herbivores as seeddispersal agents (but see Barlow, 2000).

The following natural history information pertaining totwo species of Asiatic rhinos could be taken as typical forvery large herbivores living in tropical forest; we present thisinformation to substantiate some hypotheses to be presentedlater regarding long distance movements, effects on veget-ation physiognomy, seed dispersal, need for mineral licks,etc. Sumatran rhinos (Dicerorhinus sumatrensis) inhabittropical rain forest (Nowak, 1999), and in Gunung LeuserPark in Sumatra feed mainly on small trees or saplings. Theyconsume twigs, small branches and leaves, and fruits. Toreach higher shoots of woody saplings, animals bend orbreak the stems by walking over the plant and pressing downon the trunk with the body. Sumatran rhinos snapped plantswith stem diameters of up to 5 cm (Borner, 1979). Hubback(1939) reported that feeding Sumatran rhinos will �get asapling behind his front horn and twist it round and rounduntil it is thoroughly decorticated and covered with mudfrom his head�. A favorite fruit seems to be Mangifera, a typeof wild sour mango, and young seedlings have been seengrowing from old dung deposits. Strickland (1967) notedthat young trees had been bitten off, regenerated and beenbitten off again repeatedly. The home range size for two

adult animals was about 10 km2, but this was probablyunderestimated. The animals wandered widely, and spentvery little time in the extensive swamps, which merelyfunctioned as paths from one patch of high ground toanother. Although the animals used trails repeatedly, thesmallest animal often wandered off the main trails. Therhinos were attracted to wallows baited with salt, and thetracks of several animals were found to converge at a saltlick. Although the rhinos were capable of covering manymiles in a day, in certain areas (mud wallows, salt licks, andfeeding and resting areas), they sometimes moved less thanhalf a mile in a 24-hour period. Asian elephants (Elephasmaximus) co-existed with the rhinos and were even found inthe same wallow. Rhinos created wallows used by otherspecies. Young saplings provided the majority of the diet,but these were invariably damaged. Some plants were socompletely damaged that they could not be identified in thefield. Trees up to 7 cm were sometimes snapped off com-pletely up to 2 m above the ground. Uprooted trees werestepped on and broken into smaller pieces. The damage doneby D. sumatrensis could be easily distinguished from that ofother mammals. A minimum of 40 species of plants wereeaten, but many plants were too devastated to permit iden-tification. Although D. sumatrensis fed extensively on plantsof secondary forest and forest edge, untouched primaryforest would provide adequate habitat for this species. In-deed, under modern conditions of human disturbance (suchas hunting) in Sumatra, rhinos and elephants avoid forestedges (Kinnaird et al., 2003).

Javan rhinos (Rhinoceros sondaicus) are also primarilybrowsers. The diet consists of shoots, twigs, leaves, youngfoliage, and fallen fruit; almost 150 species have been iden-tified as food plants (Schenkel & Schenkel-Hulliger, 1969;Hoogerwerf, 1970). In the course of feeding, branches up to2 cm thick are torn off and trees up to 15 cm in diameter areuprooted (Talbot, 1960). Damaged trees and shrubs oftensurvive and after some time put out new shoots and continueto grow in a horizontal direction if they are uprooted(Hoogerwerf, 1970), so certain localities have a �rhinocharacter�. This appears to hamper regeneration of the for-ests so that they continue to form suitable forage ranges forthe rhino for long periods. According to Talbot (1960), theanimals make and use game trails, ruts in the mud up to3 feet deep with roots and logs worn smooth by elephantand rhino. Strickland (1967) also reported that many youngtrees that had been eaten had parts of the bark scraped offabout a meter from the ground; in a few cases, he found treesthat had been scraped, but not eaten. Lekagul & McNeely(1977) reported that R. sondaicus inhabits dense rain forests,and in the recent past they built up large populations(Groves, 1967).

Native pigs play an important role in plant dynamics atthe understorey level in Malaysian rain forest (Ickes et al.,2001), affectcing stem density, species richness, growth andpossibly mortality. They also noted that although they wereunaware of any studies documenting the effects on theunderstorey of large mammals such as deer, elephants,rhinos, wild cattle and tapirs on tropical vegetation in



Southeast Asia, their impacts may be considerable. Similarly,Hirsch & Marler (2002) observed the fate of that plantstoppled by a typhoon. All the plants ultimately died, andherbivory by introduced Sambar deer (Cervus unicolor) andpigs (Sus scrofa var. vittatus or Sus celebensis) was the pri-mary cause of mortality.

Rhinos are not the only large animals whose feedingbehaviour can enhance their own habitat. Although we lackinformation specific to rain forest, elephants in semi-aridsavanna preferentially browse Colophospermum mopanethat has already been browsed, even when other foods arereadily available (Smallie & O’Conner, 2000).

Large herbivores can have a number of indirect ecosys-tem effects. For example feeding by feral horses in saltmarsh apparently increased the diversity of foraging birdsand crabs, but decreased the density and species richness offishes (Levin et al., 2002). Furthermore, Cumming & Brock(1997) document a loss of diversity in fruit bats, birds,ants, and mantises as a consequence of high elephantdensities. Natural levels of herbivory have been shown todiminish reproductive output, such as pollen performance(Delph et al., 1997). Conversely, lack of herbivory couldresult in greater reproductive productivity than wouldoccur under high herbivory pressure. Not surprisingly,loblolly pines (Pinus taeda) grown in a herbivore-free,CO2-rich environment are twice as likely to be reproduc-tively mature and produce three times as many cones andseeds as trees in ambient CO2 concentrations (LaDeau &Clark, 2001).

Herbivores thus have many indirect effects on plants byrendering them more susceptible to drought, other herbi-vores, fungi, and competitors (see also Hendrix, 1988).Although we have emphasized elephants and other verylarge herbivores, ungulates such as deer can also have pro-found effects and may extirpate plant species locally(e.g. Warren, 1991; McShea et al., 1997), and deer-sizedherbivores, although present in the Neotropics, are far morecommon in the Paleotropics.

Long-distance movements by LRFH

Unlike insects, large herbivores might not have the option ofspecializing on particular plant species if these occur in lowbiomass densities. They might be forced to feed on manyspecies simply to meet metabolic needs and dilute the effectsof the various toxins they ingest (Janis & du Toit, 2001).Also, cost of locomotion per unit of body mass is lower forlarge animals (Peters, 1983), thus they probably incur alower cost/benefit ratio in moving about to feed on a varietyof plant species.

We, therefore, hypothesized that LRFH move long dis-tances to meet their various needs. For example, a femaleforest elephant traveled with a maximum straight-line dis-placement of �60 km, and undertook return journeys ofmore than 35 km over a home range of at least 880 km2

(Blake et al., 2001). Forest elephants move long distances inresponse to fruit availability, thereby maximizing the avail-ability of favoured food trees (White, 1994).

Rhinoceros sondaicus may travel 15–20 km within 24 hGroves (1967) and Dicerorhinus sumatrensis have been seenswimming in the sea, but have also been found in remoteareas of steep montane forest (Hoogerwerf, 1970).

Talbot (1960) opined that it was difficult to believe that ananimal the size of a D. sumatrensis could get through suchrough and steep country. Undisturbed rhinos had wanderedthrough swift rivers five feet deep. They also used gametrails, ruts in the mud up to three feet deep with roots andlogs worn smooth by elephant and rhino. They also justwandered cross-country. Muddy, vine-covered slopes toosteep for men to climb straight up, were ascended with easeby wandering rhinos. The D. sumatrensis studied byStrickland (1967) had home ranges of at least 10 km2.

Generalized herbivores and extirpation of plants

Another consequence of the generalized herbivory of largevertebrates is that it permits a less variable population sizethan that of small herbivores, hence they might preferentiallyfeed on certain plant species, then switch to others when thepreferred species are depleted. This is similar to the phe-nomenon referred to as �apparent competition� by Holt(1977), and reviewed by Morin (1999). Lindroth (1989)posits essentially the same phenomenon with regards toplants and notes that plant species can become locallyextinct because of selective feeding by generalist mammalianherbivores because herbivore populations may not declineappreciably when preferred plants become scarce.

By contrast, specialist herbivore populations are likely toperiodically fluctuate. Stochastic fluctuations resulting fromthe combination of large and small herbivores might be morelikely to drive plants to extinction than those exposed onlyto small herbivores because plants exposed to both small andlarge herbivores are never free of herbivore pressure. Despitethe lack of direct evidence in support of this contention, wehypothesize that larger herbivores can bring about localextirpation of plant species. The literature provides supportfor this hypothesis (e.g. Buechner & Dawkins, 1961;Harper, 1969; Laws, 1970), and there is ample evidence thatanimals can extinguish plant populations and species(Niklas, 1997). The greater species richness of plants inNeotropical than Afrotropical forests (Turner, 2001) couldthus be attributed in part to the abundance of large herbiv-ores.

Large species writ small: the proliferation of small

Neotropical herbivores

In a review of the role that niche opportunities play in theprocess of ecological adaptive radiation, Schluter (2000)concluded that some studies of the diversity of herbivoresand the resin defenses of their prey (plants) are the first clearindication that freedom from enemies (in that instance,herbivores) either promotes speciation or slows extinctionrates. We hypothesized that the paucity of large Neotropicalherbivores relative to their Paleotropical counterpartshas fostered a proliferation of smaller herbivores, both



vertebrate and invertebrate. Consistent with our hypothesis,various small Neotropical species are unmatched by Old-World analogues.

The hoatzin, arguably the most folivorous bird on earth(Grajal et al., 1989), is strictly Neotropical; although otherherbivorous birds exist, the hoatzin is the only bird in theworld with foregut fermentation to cope with potent allel-ochemics.

Leafcutter ants (Atta cephalotes and relatives) are alsofound only in the New World and could be thought of asindirectly folivorous ants. Much like large mammalian her-bivores, their fungal mutualists enable them to make use ofat least eighty plant species (Steven, 1983). Furthermore, atone Neotropical site the biomass of plants eaten by leafcutterants was equivalent to that of folivorous vertebrates andonly slightly less than that of all the other herbivorousinsects combined (Leigh & Windsor, 1982). It seems plaus-ible that the metabolic needs of large leafcutter colonies canonly be supplied by a substantial quantity of vegetation froma large variety of plants, which precludes specialization onone or a few plant species.

Although the role of leafcutter ants in the New Worldcould be taken by termites (Isoptera) in the Old World, wehypothesize to the contrary. Bignell & Eggleton (2000)provide a world-wide summary of termite feeding groupsincluding soil-feeders, soil/wood interface-feeders, wood-feeders, litter-foragers, grass-feeders, and minor feedinggroups. Of these, none seem to be folivores on trees andshrubs, although the litter-foragers and grass-feeders prob-ably come closest. Although litter-foragers forage on leavesand small woody items, these seem to be taken on or near theground, and primarily dead. Grass-feeders forage for(usually dry) standing grass and other low vegetation stems,usually cutting and removing it to the nest–again, not thesame as arboreal folivory.

A slightly different classification by feeding substrate waspresented by Bignell & Tayasu (2001). It included wood,detritus, grass, wet wood, dry wood, inquiline, fungus, soil,soil–wood interface, litter, microepiphytes, or some combi-nation of these. Green tree and shrub leaves were notincluded. Hence, tropical rain forest termites feed primarilyas detritivores, not herbivores.

Caviomorph rodents are diverse and abundant in theNeotropics (Nowak, 1999). Neotropical forest caviomorphsrange in size from small rats of several genera to porcupines(Erethizontidae) and pacas (Agouti paca). Although cavio-morphs are trophically diverse, they are typically more her-bivorous and less faunivorous than the dominantPaleotropical rodents of the family Muridae. The Neo-tropical family Echimyidae includes several species ofarboreal folivores, such as bamboo rats (Kannabateomysamblyonyx) in the Atlantic forest and Dactylomys dactylinusin the Amazon, which can become locally abundant (Olmoset al., 1993; Patton et al., 2000). By contrast, the continentalmurid species in Paleotropical forests cannot cope withplants that are chemically well protected; not surprisingly,folivorous murids in Paleotropical forests only occur inislands such as the Philippines that have few large herbivores

(Eisenberg, 1978). Herbivorous snails may also be moreabundant in the Neotropics, such as the Pantanal wetlandsof Brazil (C. Peres, pers. obs.), but we are unaware ofintercontinental comparisons of population densities.

Iguanids in the subfamily Iguaninae comprise the mostspecies-rich lineage of herbivorous lizards (Pough et al.,2001), and occur in the Neotropics and Madagascar but noton mainland Africa and Asia. By contrast, the Paleotropicallizards in the same size range as iguanids – especially mon-itors of the family Varanidae – are primarily carnivorous.Finally, we also note that butterflies are most speciose in theNeotropics (Robbins & Opler, 1997), and most caterpillarsare herbivorous.

One could argue that the differences we propose for but-terflies, ants, and bees, are simply because of random vica-riant events. If this were true, then there should be manyhigher insect taxa with peculiar distributions that have norelationship to the presence or absence of large herbivores.We hypothesized to the contrary. In fact, tropical forestinsect faunas on the various continents share a variety offeatures, including low dominance and frequency of species.Additionally, for beetles (and by inference, other orders ofinsects), the most species-rich and abundant families areremarkably constant in all comparable studies (Wagner,2000). Thus our hypothesis was supported.

Diversity of predators of herbivorous snails

We hypothesized that the paucity or large herbivores in theNeotropics has increased the abundance of snails. We haveno direct evidence of a greater abundance of snails in theNeotropics, but exclusion experiments from Fennoscandiaindicate that cervids can decrease the abundance of terrest-rial gastropods (Suominen, 1999). Furthermore, the fol-lowing evidence is consistent with our hypothesis. Ataxonomically diverse array of unrelated Neotropical ver-tebrate species have become snail-eaters, and it seemsplausible that a pre-requisite for the evolution of so manyspecialized predators is a corresponding abundance of her-bivorous snails. Among others, specialized (and oftensympatric) snail-eaters include the limpkin (Aramus guar-auma), snail kite (Rostrhamus sociabilis), slender-billed kite(Rostrhamus hamatus), hook-billed kite (Chondrohieraxunicinctus), the caiman lizard (Dracaena guianensis), andDipsadinae snakes (although the Asiatic Pareatinae appearto have similar adaptations) (Greene, 1997). Although manyvertebrates eat snails, the aforementioned taxa are foundonly in the New World and are arguably the most specializedmembers of their respective taxa for a diet of snails.

Herbivore motility and plant phenology

We hypothesized that plants could offset the mobility oflarge herbivores by synchronizing phenology such that allplants would be in the same stage of palatability at the sametime, thus precluding sequential selection of individuals inthe most palatable stages. Although this strategy is notapparent among the non-reproductive portions of tropical



forest plants, an analogous situation occurs with Asiandipterocarps, which are believed to limit the population sizeof vertebrate seed-predators by mast-fruiting at supra-annual intervals (Curran & Leighton, 2000).

We hypothesized that inducible defenses (sensu Karban &Baldwin, 1997) might be more effective over short timescales on small herbivores than on large herbivores. Someevidence is consistent with this hypothesis. For example, acaptive tapir rejected at least 300 species of native broad-leafed plants and accepted another 150, eating small quan-tities of each species during a feeding bout (Janzen, 1983). Itseems plausible that many food species used by tapirs are toouncommon to serve as dietary staples. Some of their pre-ferred foods are rare where tapirs occur in apparently typicalnumbers and are common where tapirs have been selectivelyoverhunted. Janzen (1983) suggested that food scarcity maybe in part generated by tapir browsing. This is consistentwith dietary and radio-tracking studies that show that tapirsfeed primarily on fruits wherever conditions are favourable(Bodmer, 1990) or otherwise typically trapline betweendifferent regenerating canopy gaps where they selectivelybrowse on herbaceous plants and tree saplings (Salas &Fuller, 1996; Salas, 1996; C. Peres & E. Dias, unpubl. data).Similarly, okapis selectively feed on leaves, buds and smallbranches (Kingdon, 1997) that are relatively scarce (Hart &Hart, 1988, 1989). The rhinoceros feeding informationpresented above also supports this. Thus LRFH probablyselect forage plants when they are most palatable.

Fragile vegetation and the evolution of small arboreal

vertebrates

Emmons & Gentry (1983) speculated that larger herbivoreshave selected for sturdier vegetation in the Paleotropics,which in turn favoured the evolution of certain forms oflocomotion over others. There is evidence consistent withthe hypothesis that the activities of large herbivores in thePaleotropics have selected for trees that are faster growingand more responsive to light gaps. Whitmore (1998) notedthat forests in Brazil tend to be dominated by tree speciesthat are shade tolerant and slow growing. These hardwoodspecies tend to respond too slowly to canopy opening tobecome attractive for silviculture. When canopy gaps areformed, few seedlings are able to grow vigorously. Whit-more suggests that this degree of canopy disturbance leads tothe forest �tumbling down� to a mass of woody climbers andcommercially useless pioneer species. Whitmore & Silva(1990) also noted that Amazonian trees tend to be denserthan those from other regions. Wood density is, however,negatively correlated with growth rate (Turner, 2001), hencePaleotropical species could have been selected for a fastergrowth so that they can take advantage of disturbance (suchas light gap formation) caused by large herbivores.

There is also some evidence to support our hypothesis thatlarge Paleotropical herbivores could increase the geo-graphical range size of the trees whose fruits they disperse.Many African forest plant species have very wide distribu-tions, whereas many Neotropical species are much more

restricted (Richards, 1996). Dry climates have been proposedas the mechanism that capped African gamma diversity(Axelrod, 1952, 1972; Raven & Axelrod, 1974), but themovements of large herbivores among forest blocks, and theenormous distances that they disperse fruits (Blake et al.,2001), might also have played a role in homogenizing biotas.

The removal of fragile vegetation in the Paleotropics mayhave enabled climbing vertebrates to obtain sturdier sup-ports than their Neotropical counterparts, thus providingfewer constraints to body size increments over time. Thishypothesis is supported by the wider body size spectrum ofPaleotropical arboreal vertebrates (Cristoffer, 1987).

Trophical shifts were conjectured to be a consequence of asmaller body size, if not phyletic dwarfism, in Neotropicalarboreal endotherms, which often eat more animal matter(e.g. arthropods) and less vegetation. This would seem to bea partial contradiction of the previous paragraphs, but this ismore apparent than real. There are anatomical and physio-logical constraints on the evolution of folivory in endother-mic vertebrates (Eisenberg, 1978), and the folivorouslifestyle is all but precluded for very small arboreal endo-therms. Thus, as Neotropical arboreal endotherms are con-strained to remain small to enable them to move about onfragile vegetation, they have little choice but to feed on foodsthat are easily and quickly digested.

By contrast, ectotherms and terrestrial herbivores can af-ford to become smaller and retain or evolve a plant diet.Most of the small herbivores we conjectured to have filled infor large herbivores are ectotherms or terrestrial species, andthe few folivorous arboreal endotherms listed, such as thehoatzin, are profoundly modified behaviourally and physi-ologically.

There is some slight support for the hypothesis that thefeeding activities of large herbivores would extinguish somespecies of epiphytes and lianas in the Paleotropics. Therichest herb and epiphyte floras in tropical forests areprobably in the wetter parts of the Neotropics (Richards,1996). These plants would likely qualify as �fragile veget-ation� (sensu Emmons & Gentry, 1983) and would beespecially susceptible to the effects of large herbivores.

The data on woody lianas is slightly more complicated.Although lianas have similar species richness in the Neo-tropics and Africa, their biomass tends to be greater in thelatter (Hegarty & Caballe, 1991). This is consistent with theidea that large herbivores remove more fragile vegetation inthe Paleotropics (Emmons & Gentry, 1983), leading tosturdier or more massive lianas.

We hypothesized that the activities of large herbivores hasreduced the abundance of palms in the Paleotropics. Her-bivorous vertebrates have been shown to cause mortality inpalms (Pacheco, 2001), presumably by feeding directly onthem. We also hypothesized that the less-effective feedingupon lianas by Neotropical LRFH than their Paleotropicalcounterparts enhances palm survival in the Neotropics byencumbering the palms� competitors. Woody lianas mayreduce the survival, growth, and fecundity of arborescentpalms (Svenning, 2001); nevertheless, palms are wellequipped to avoid liana infestation, mainly because of the



continual shedding of their long leaves (Putz, 1984; Richet al., 1987). Note also that the abundance canopy palms inupland rain forests is a phenomenon largely confined to theNeotropics, Madagascar, New Caledonia, and a few otherislands (Gentry, 1988). These also happen to be places withfew large herbivores. Thus, the available information isconsistent with our hypothesis.

Smaller fruits and frugivore motility

We hypothesized that the removal of fragile vegetation bylarge herbivores could thwart reproduction of small-fruitedplants, especially in the understorey. Since both terrestrialand arboreal herbivores are larger in the Paleotropics, wemight expect them to feed on larger fruit, and they do:Paleotropical fruits average larger than those in the Neo-tropics (Mack, 1993).

Since larger animals within trophical guilds tend to havelarger home ranges, we hypothesized that Paleotropicalfrugivores would forage, on average, over larger areas thantheir Neotropical counterparts (cf. Fleming et al., 1987).The available information, although limited, supports thishypothesis.

For example, there seem to be ecologically relevant dif-ferences between the Paleotropical hornbills (Bucerotidae)and their purported Neotropical counterparts, the toucans(Ramphastidae, in part). Hornbills are strong-flying, wide-ranging birds that could only have reached islands ofMalesia or the Malay Archipelago by flying over largestretches of open ocean (Kemp, 1995). Hornbills such as thewreathed hornbill (Aceros undulatus) move long distances insearch of patches of figs, thereby causing local populationsto fluctuate (Kinnaird et al., 1996). Similarly, red-knobbedhornbills cover daily distances of up to 13 km (Poonsward& Tsuji, 1994), and large flocks of wreathed hornbills mighttravel more than 10 km between feeding sites and their roost(Leighton, 1986; Tsuji et al., 1987). Hornbills in generalfavour larger fruits taken from relatively few large trees, andsome of their foods are quite scarce (Kemp, 1995). Fruitsfavoured by hornbills are both produced at infrequentintervals (Lambert & Marshall, 1981) and ephemeral(Leighton, 1982).

By contrast, toucans are not known to travel long dis-tances, and massive die-offs of large monospecific flocks ofRamphastos toucans attempting to cross large rivers havebeen occasionally observed in central Amazonia (A. Whit-taker, pers. comm.), presumably following a generalizedfailure in fruit crops. Toucans are rather clumsy and weakflyers that may fall into the water if they fail to achievesufficient trajectory when crossing wide rivers (Sick, 1993).The course of their flight is undulatory and connects tree-tops that are rarely far apart. Indeed, the species ranges ofAmazonian toucans are effectively separated by river bar-riers (Haffer, 1974). Although quantitative data on toucanhome range sizes are few, one study reported that keel-billedtoucans (Ramphastos sulfuratus) have home ranges that varyfrom 18 to 111 ha in size (Graham, 2001); this is far smallerthan those of three hornbill species that vary from 370 to

1000 ha in the breeding season to as much as 2800 ha in thenon-breeding season (Poonsward & Tsuji, 1994).

Hornbills may be better adapted than toucans to exploitwidely dispersed but large clumps of food. This in turnsuggests that the type and dispersion of fruits in Neotropicalforests – at least those consumed by toucans – is significantlydifferent from those in Paleotropical forests. Note thatalthough Paleotropical fleshy fruits are somewhat larger thanthose in the Neotropics, fruit productivity in Neotropicalforests is equivalent to or greater than that in the Paleo-tropics (Hladik, 1978; Chapman et al., 1999; Ganesh &Davidar, 1999).

Although data are scant, Paleotropical fruit bats (familyPteropidae) also appear to fly farther when foraging thando Neotropical fruit bats (family Phyllostomidae) (Nowak,1999). The roosts of the African Eidolon helvum suggest aforaging range of at least 30 km. Eidolon makes extensiveseasonal migrations; one colony left its roost in a forest inthe Ivory Coast in February, moved northward into thesavanna zone, and migrated at least as far as the NigerRiver Basin by the middle of the wet season (Thomas,1983). Some colonies in East Africa may make a roundtrip of over 2500 km. Hypsignathus forage up to 10 kmfrom the roost at night (Bradbury, 1977) and Epomo-phorus wahlberg switches day roosts every few days andflies up to 4 km from these to nocturnal feeding areas(Fenton et al., 1985). Rousettus leschenaulti shifts inresponse to the supply of fruit and can make foraging tripsof 40–50 km in a night (Lekagul & McNeely, 1977).Cynopterus may travel 97–113 km in one night. ManyPteropus mariannus have been found to move betweenislands in the southern Marianas on an irregular basis(Wiles & Glass, 1990). Most Pteropus species roost inemergent trees that rise above the forest canopy and for-age far from their roosts (Pierson & Rainey, 1992), andpopulations on large land masses may travel 40–60 km tofeed. Nowak characterized Pteropus as strong fliers, andSterndale (1884) noted that a P. giganteus landed on aboat 320 km from land.

By contrast, most New World fruit bats do not appear tofly so far as some Paleotropical species. Carollia perspicillatadisperse nightly, with each individual going to two to sixfeeding areas and flying an average of 4.7 km per night(Heithaus & Fleming, 1978). Vampyrodes caraccioli batswere found to emerge from their roosts and fly 850 m to afruiting tree, and visit several other feeding areas within arestricted radius (Morrison, 1980). Female Artibeus jamai-censis fly an average of 8 km to forage (Morrison, 1978),and the same species in Barro Colorado Island, Panama, flies1–4 km between day roosts and feeding sites (Handley et al.,1991).

With regards to use of the understorey or canopy, Bernard(2001) reported that short-tailed fruit-eating bats, Carolliaperspicillata and C. brevicauda, feed principally on under-storey plants such as Piper, Solanum, and Vismia (Fleming,1988). In fact, C. brevicauda was exclusively captured inground nets, and just 8% of the captures of C. perspicillatawere in canopy nets.



Although pteropids are on average larger than phyllosto-mids (Cristoffer, 1987), considerable overlap exists in bodysize. In fact, one feature of the overlap suggests that thedifference between the families is explained better by envi-ronment than by phylogeny. We refer to the genus Vampy-rum and, to a lesser extent, several other related genera.Vampyrum is a large bat with a wingspan of �1 m thatdemonstrates that phyllostomids can reach a fairly large size.Vampyrum, however, is a carnivore, not a frugivore. We areaware of no physiological or morphological constraints onphyllostomids that would render them incapable of beingboth large and frugivorous, hence look to extrinsic, ecolog-ical causes. We suggest that the frugivorous/omnivorousPhyllostomidae and Pteropidae differ in several ways rele-vant to this discussion.

Phyllostomids are omnivorous in general, but some generaare more specialized as frugivores. Some of the frugivoresmay even obtain protein via seed predation (Nogueira &Peracchi, 2003), which is so far unrecorded for pteropids.Omnivory enables them to feed on both animal and plantfood, thus they are able to subsist on small patches of forestunderstorey or scrub that lack large-fruited trees. Smallunderstorey trees typically have small, easily-digested fruitsthat are relatively evenly dispersed compared with large-fruited trees, and produce small fruit crops for an extendedperiods. Even if these fruit crops fail, the omnivorous speciesare not necessarily forced to undergo the uncertainties ofmigration or nomadism to find new fruit sources, becausethey can subsist on locally available animal food. Echolo-cation enables them to navigate even in the darkest forestunderstorey at night. The average densities of these speciestend to be high and probably less variable locally than thoseof pteropids.

Since Old World frugivorous pteropids are on averagelarger than their Neotropical counterparts, they probablyfeed on, and therefore disperse the seeds of, larger fruit.However, even good night vision requires a minimumambient light intensity, which may not occur in the forestunderstorey at night. Therefore, their roost trees tend to berelatively exposed and well illuminated. They are unable tosubsist on local fruits alone, or on animal matter (but seeCourts, 1998), so must forage more widely. Presumablyflying long distances in the canopy would be taxing ofenergy. Fortunately, their specialized retinas permit them tonavigate well under an open sky, above the canopy.Although they require scarce, high-quality fruiting trees,their large body size and rather narrow wings are efficientfor flying long distances. They can occur at high biomassdensities locally but are patchy both in space and time.

These syndromes are not absolute and there are species ineach family that possess at least some of the characterstypical of the other. Nevertheless, we hypothesize that there isa general New World/Old World dichotomy among tropicalforest frugivores. When New and Old World ecologicalsurrogates exist, we predict that OldWorld frugivores willreach the largest size, will be the most mobile, and will feedon the largest fruits. Furthermore, these fruits will be morepatchy in both space and time, and their seeds will be

dispersed a greater distance. The bat fauna of New Guineaprovides tests of this.

New Guinea is inhabited by pteropid bats, not phyllos-tomids (Bonaccorso, 1998). On that ground, we wouldexpect it to have large, mobile, species that feed and roost inthe canopy. In fact, it does, especially in coastal areas and onislands adjacent to New Guinea. However, New Guinealacks large herbivores, which should have resulted in selec-tion pressures more similar to those in the mainland tropicsof the New World than in the Old. Therefore, we hypo-thesized that some of the New Guinea pteropids wouldresemble their Neotropical counterparts. This is confirmed:several of the pteropid bats described in Bonaccorso (1998)have characteristics similar to those of phyllostomids. Forexample Dobsonia anderseni feeds on subcanopy as well ascanopy fruits, and sometimes roosts in tree hollows andcaves, as does D. inermis; D. minor is highly maneuverablein the understorey, feeds on introduced Piper fruits (whichare often consumed by phyllostomids), and is often caught intraps less than 5 m above ground. D. moluccensis is able toaccess understorey fruits not available to Pteropus, whichare sympatric; it even alights on the ground to obtain them.Finally, the subfamily Nyctimeninae, which is endemic to thevicinity of New Guinea (one species reaches mainland Aus-tralia), is considered convergent morpho-ecologically withphyllostomids of the subfamily Stenoderminae.

We also hypothesized that anthropoid primates wouldexhibit a size and foraging dichotomy. In the Neotropics, thetiny, diurnal monkeys (families Callitrichidae and someCebidae) tend to feed on fruits that produce small crops overlong periods of time (Terborgh, 1983). Superficially, itwould appear that the Paleotropics have counterparts tosmall monkeys among the prosimians. For reasons toodetailed to go into here, small Neotropical primates prob-ably do not have obvious counterparts among small Paleo-tropical mammals; we refer to Charles-Dominique (1983)and Sussman & Kinzey (1984), respectively, for discussionof how small Neotropical primates differ ecologically fromprosimians and squirrels. The peculiarly small size of manyNeotropical anthropoids has been discussed elsewhere(Leutenegger, 1979; Sussman & Kinzey, 1984; Cristoffer,1987). We noted that the overlap in size between phyllos-tomid and pteropid bats suggested that there are no phylo-genetic constraints on large size in phyllostomids. Theprimates exhibit a similar overlap, suggesting that the sizediscrepancy is imposed by natural selection rather than byphylogeny. In general, the predominance of small frugivoresthat feed on evenly-dispersed resources in several unrelatedfamilies of Neotropical vertebrates supports the hypothesisthat a common constraint has shaped the evolution of theirforaging behaviour.

Are consumers of reproductive plant products more

important in the Neotropics?

Herbivores cause plants to divert resources from reproduc-tive plant products (RPP) into defenses against herbivores, orinto repairing damage caused by herbivores (Marquis, 1984;



Bazzaz et al., 1987; Simms & Rausher, 1987; Hendrix &Trapp, 1989; Doak, 1992; Mauricio et al., 1993; Sagers &Coley, 1995; Strauss et al., 1996; Juenger & Bergelson,1997; Strauss, 1997; Van Der Wal et al., 2000; but seeGronemeyer et al., 1997). We suggest that there is a trade-off between allocation to reproduction and defense againstlarge herbivores, resulting in a decrease in production ofreproductive tissue. This could then stymie the diversifica-tion of animals that consume reproductive plant products,including nectar, pollen and fragrant oils.

ASSESSING REPRODUCTIVE ALLOCATION

AT A CONTINENTAL SCALE

It is difficult to compare reproductive effort in the Neo-tropics with that of the Paleotropics because there is nosingle currency to measure reproductive effort. Furthermore,when we speak of reproductive enhancement of Neotropicalplants, we are making a generalization. A more realistictheory would accommodate the likelihood that large her-bivores suppress many types of reproduction but enhance afew others. For example, although large herbivore activitiesprobably diminish plant reproduction in many circum-stances, they might actually enhance it for those plant specieswhose fruits are especially adapted for dispersal by largevertebrates. We would then need to somehow control for theeffect of the so-called �megafaunal seed dispersal syndrome�(Barlow, 2000), as well as for grains, as elephants at leasttend to promote grassland over forest (Owen-Smith, 1988).

Measuring the relative importance of various plant def-enses against both small and large herbivores is also difficult.We hypothesized, however, that effects on vegetationstructure would be noticeable. At least one trait more likelyto be affected by large herbivores, namely physical fragility,has been addressed (Emmons & Gentry, 1983).

We could also speculate on what characteristics contributeto the fragility of Neotropical plants. Whitmore (1998)suggests that mechanical toughness rather than chemicalcomposition is the major deterrent to insect herbivory, andnotes that softer, young leaves are preferred to older leavesdespite their greater chemical protection. Perhaps largeherbivores are more likely to damage tough leaves than areinsects, because they can generate bite forces over a largerarea. We have no direct evidence to assess this, but note thatamong folivorous insects, larger individuals are able to bitethicker leaves (c. Bernays, 1998).

We suggest that many fragile understorey plants in thePaleotropics simply cannot invest as much in reproductionand, therefore, produce less RPP. This lower allocationcould cut across many taxonomic lines, and therefore the netresult would be a reduced overall availability of RPP forconsumers. We will emphasize flower nectar and pollen,nuts, fruit pulp, and seeds within fruits (other than grains)more than others because they have been easier to study.

One of our first hypotheses to address this question is thatthe aggregate biomass density of RPP consumers (exceptthose that specialize on grains and megafaunal fruits) shouldbe greater in the Neotropics than in the Paleotropics.

However, site-to-site variation due to factors such as soilfertility and rainfall regime could obscure differences due tolarge herbivore activities (or lack thereof) and thus the dif-ferences in consumer biomass density. Ratios of trophicalcategories are frequently used in insect ecology (e.g. Warren& Gaston, 1992; Basset, 1995). Therefore, a more relevantmeasure of the effects of large herbivores might, in somecases, be ratios of the aggregate biomass of consumers ofRPP to the aggregate biomass of consumers of some otherfoodstuff. Thus, we might predict that a Neotropical site willhave a higher RPP: non-RPP ratio than a Paleotropical site,regardless of their overall productivity.

Another hypothesis based on ratios is based upon thegeneralization that primary productivity is a strong nonlin-ear predictor of primate species richness (Kay et al., 1997;Mittelbach et al., 2001). Furthermore, fruit abundance is afar better predictor of primate biomass than is vegetationbiomass, and the abundance of certain RPP resources, suchas large-seeded trees in the Lecythidaceae, may explain thegeographical distribution of pitheciines such as uakaries(Cacajao spp.) and bearded saki monkeys (Chiropotes spp.)(Peres & Janson, 1999; Stevenson, 2001). This suggests thata greater availability of certain types of RPP make possible aproliferation of RPP consumers. The examples we havealready given of small frugivores are consistent with the ideathat the small fruit component of RPP resources is especiallyimportant in the Neotropics.

It does not necessarily follow, however, that a paucity ofmegafauna should result in a decrease in species richness ofall taxa. For instance, large Paleotropical herbivores mayhave obligate parasites that would not occur in the Neo-tropics because their hosts are absent. Another example ofmegaherbivore facilitation pertaining to birds, concernsvultures. The Pleistocene of North America had a morpho-logically more species-rich vulture fauna than does theextant Nearctic, presumably because the Pleistocene eco-system, which was much richer in megafauna, providedmore large-herbivore carcasses for scavengers (Hertel,1994).

Nevertheless, the importance of the activities of largeherbivores in determining differences between the Neotrop-ics and Paleotropics has received little attention, despite thefact that herbivory has been shown to have a greater effecton plant growth habits at sites with high biomass, such astropical moist forests, than at sites with low biomass (Bonser& Reader, 1995).

RPP/non-RPP ratios

One rough measure of the potential suppression of theproliferation of RPP consumers is the ratio of RPP species tonon-RPP species. For the latter, we primarily use predators,which are unquestionably not directly dependent upon RPP.Table 1 shows the species richness ratios of RPP consumersto non-RPP consumers for Neotropical and Paleotropicalregions. We were unable to find exactly comparable sites foreach table entry, hence have indicated the extent covered.Although species–area relationships, as originally pointed



out by MacArthur & Wilson (1963), would predict thatlarger geographical regions should have a larger speciespool, the trophic ratios should be unaffected by area sizeexcept in the extreme cases (e.g. the loss of large predatorsfrom small habitat fragments). The taxa used were those forwhich data were available, so it is possible that they weresomehow not representative.

Many taxa which varied in food habits were excluded forclarity. For example, the Tyrannidae as arrayed in Sibley &Monroe (1990) encompasses several trophically diverse groupsof birds, including cotingas, manakins, tyrant flycatchers, andothers. Furthermore, splitting tyrannids into smaller cladesdoes not clarify the situation, because there is substantialdietary variation even within some subfamilies. Similarly,phyllostomid bats display virtually the same dietary breadth asthe entire order Chiroptera, although in this case we feltcomfortable enough with the data to exclude those species thatdiffer dramatically in diet from the Paleotropical Pteropidae.

Kingfishers sensu latu are another example. Kingfishersinclude the specialized piscivores in the Neotropics (Remsen,1990), as well as many Paleotropical taxa that are partly orentirely insectivorous, therefore, we compared only the fish-catchers. In most cases, we took pains to assure that ouradmittedly subjective demarcations were made in such a waythat the selection would favour the null hypothesis, discussed

below, rather than our own hypothesis. Likewise, guans(Cracidae) overlap in use of fruits with toucans (Guix et al.,2001); again, their inclusion would have strengthened ourhypothesis at the expense of the null hypothesis. Bees wereexcluded from statistical tests because we lacked numericaldata, although the differences between continents are con-sistent with the alternative theory. As will be seen, theseconcessions made little difference.

Our null model is that the ratio of the Neotropics toPaleotropics for consumers of RPP should not be significantlydifferent than the ratio for consumers of other major dietaryguilds. The alternative model is that the ratio of the Neotropicsto Paleotropics for consumers of RPP should be greater thanthe ratio for consumers of other foods. A Wilcoxon signedranks test (Agresti & Agresti, 1979) resulted in a P value of0.0066 for a one-sided test. Thus, we can reject the null hy-pothesis that RPP-dependent taxa make up the same propor-tion of the fauna in the Paleotropics and the Neotropics.

The greater species richness of certain trophic guilds in theNeotropics could be because of factors other than largeherbivore activities. The Pleistocene Refugia Hypothesis(Haffer, 1969, 1974) has been used to explain the greaterspecies richness of various taxa in the Neotropics comparedwith the Paleotropics. An alternative hypothesis to explainAmazonian diversity patterns is the Riverine Barrier

Table 1 Species richness of various animal taxa, expressed as ratios between the number of species occurring in different regions of the New-

and Old World tropics

Taxa* N/O ratio� N/A ratio� Geographic regions covered

Taxa dependent upon

reproductive plant parts

Toucans and Hornbills§ 2.41 1.64 All of Neotropics & Paleotropics

Parrots 6.27 6.27 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinentHumming-birds and sunbirds 3.65 2.21 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinent

Trogons 3.0 4.5 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinent

Bees >1.0 >1.0 All of Neotropics & PaleotropicsButterflies 1.50 2.50 Costa Rica, Malaysia, Liberia

Fruit and flower-feeding bats 3.48 4.21 Mainland Neotropics, mainland Paleotropics and large islands (except Madagascar)

Taxa not dependent upon

reproductive plant partsHerons 1.0 0.95 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinent

Raptors– 0.9 0.92 South America, Asia & Africa (including Madagascar)

Ibises and storks** 0.88 0.88 Neotropics & PaleotropicsJacamars and bee-eaters 2.33 1.0 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinent

Owls (Strigidae) 0.69 1.11 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinent

Nightjars and nighthawks 2.56 1.15 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinent

Ducks and geese 0.57 0.96 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinentSwifts 1.13 1.07 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinent

Kingfishers�� 0.71 0.86 All of Neotropics, tropical Africa, mainland tropics of Asia plus large islands

Swallows 1.0 0.65 Brazil, Senegal, the Gambia, Kenya, northern Tanzania, Indian subcontinent

*Data are from Gardner (1977); Handley et al. (1991); Sick (1993); Barlow & Wacher (1997); Robbins & Opler (1997); Fry & Fry (1999);

Grimmett et al. (1999); Nowak (1999); Zimmerman et al. (1999); Clements (2000); Michener (2000); Weidensaul (2000).

�Number of species in the Neotropics: number of species in the oriental tropics.

�Number of species in the Neotropics: number of species in the Afrotropics.§Bucerotidae and Ramphastidae excluding New World barbets.

–Cathartidae, Accipitridae and Falconidae.

**Threskiornithidae and Ciconiidae.��Includes piscivorous species only.



Hypothesis (Wallace, 1852; Hershkovitz, 1972; Gasconet al., 2000). Unfortunately, the Pleistocene Refugia andRiverine Barrier Hypotheses predict a greater overall speciesrichness because of allopatric speciation, but they do notspecify which trophic categories will be most affected. Bycontrast, we predict that consumers of RPP will be relativelymore important in the Neotropics than in the Paleotropics.In other words, alternative hypotheses may explain whythere are so many Neotropical birds, but they do not explainwhy so many of them feed on RPP.

Note also that allopatric mechanisms of speciation do notincorporate constraints on species richness. More geograph-ical isolation events simply correlate with more species, andfrom this one can infer that if the Paleotropics had more iso-lating events, it would have more species. This seems simplisticto us. The results of natural experiments, such as described byMoulton & Pimm (1986), confirm that one cannot simplypack similar species indefinitely. If unconstrained by energeticor other limits, then combining habitat fragments producingallopatric speciation by forest refugia and fluvial barriers,would result in the sum total of all the species present in eachfragment. Most of the species produced by allopatric specia-tion within tropical forest isolates would have ecologicallysimilar sister species in other isolates, and it seems unrealisticto us that so many similar species would co-exist indefinitelywhen brought back into sympatry. In reality, energy, nitrogen,nest cavities, or some other currency would be in short supply,and not all species would survive. Our theory incorporates thislikelihood, at least at a coarse level.

Our intention here is not necessarily to replace one ormore hypotheses with another; the truth may be betterserved by a combination. Thus, the forest refuges and riverbarriers hypotheses provide a mechanism for speciation,whereas we suggest what kinds of proliferations will be mostsuccessful in independent evolutionary experiments mouldedby large herbivores.

We hypothesized that New Guinea, which lacks largeherbivores, would be inhabited by a disproportionate num-ber of frugivorous and flower-feeding birds. This hypothesiswas supported. Beehler et al. (1986) noted that New Guineais unusual in supporting large numbers of fruit- and nectar-feeders. The roster includes obligate frugivores, which arerelatively uncommon elsewhere. For example, New Guineahas twice as many fruit-eaters and nearly twice as manynectar-eaters, as measured by proportion of the fauna, ascomparable lowland forest in Peru. Thus in some respects,New Guinea, which lacks herbivores even as large as tapirs,may contrast even more vividly with Paleotropical foreststhan do the Neotropics.

Neotropical vs. Paleotropical flowers and their visitors

We hypothesized that since there are fewer LRFH in theNeotropics to damage flowers and to force plants to allocateresources to defense against large herbivores, that the Neo-tropics would have a greater diversity of pollination typesand would produce more RPP than those in the Paleotropics.In fact, several pollination �subsyndromes� attributed to non-

flying mammals are exclusive to the New World. One ofthese is referred to as the �upright-flowered lowland tropicalsyndrome� and the other is the �mossy forest syndrome�(Proctor et al., 1996). The absence of these syndromes in theOld World is explicable with reference to large herbivores.Let us assume for the moment that large herbivores wouldconsume many flowers given the opportunity. If, say, anelephant attempted to eat a downward-facing flower, itmight still inadvertently be dusted by pollen as it thrust itstrunk among them, simply because gravity would causepollen to fall onto the trunk as the flower shook. By contrast,an upright flower would not normally scatter pollen onto thetrunk. Natural selection might tend to eliminate uprightflowers from the population since they would not reproduceas successfully. Upright flowers might also be more likely tobe detected and eaten by large herbivores.

Similarly, the mossy forest pollination subsyndrome mightnot be viable in forests of low stature that are frequentlybrowsed by large herbivores. Since mossy forests exist inboth the Old World and the New World, the absence of thesyndrome in the former might be because of feeding by largeherbivores.

Proctor et al. (1996) list several other features of Neo-tropical flowers that are rare or absent in the Paleotropics.Oil-producing flowers occur primarily in the Neotropics.Note that oil is more expensive for a plant to produce thansugar – another indication that Neotropical plants canallocate more to reproduction. Bees are more diverse in theNeotropics (Michener, 2000), which is reinforced by themore regular flowering of Neotropical canopy trees, incontrast to the irregular condition in southeast Asia.Sucrose-rich nectars occur in flowers pollinated by hum-mingbirds, but not by passerines such as sunbirds (Baker &Baker, 1983), and hummingbirds prefer sucroses overhexoses (Hainsworth & Wolf, 1976). In fact, hummingbirdswill not even utilize flowers that lack sucrose altogether,whereas sunbirds in the Old World take nectar high inmonosaccharides and will not utilize flowers that lack them.Production of disaccharides should require more energy forthe plant than monosaccarides, thus hummingbird flowersmight produce more expensive nectar than do sunbirdflowers. This again suggests a greater allocation toreproduction in the Neotropics. Curiously, most of thesunbirds, flower peckers and megachiropteran bats thatpollinate Asian tropical trees are not even attracted tounderstorey flowers, although they may often visit or evenconfine themselves to forest gaps (Bawa et al., 1990). Bycontrast, in the New World some species of both hum-mingbird and phyllostomid bat pollinators are understoreyspecialists relying heavily on a few plant taxa exhibitingprolonged or staggered flowering periods.

The plants most strongly associated with hummingbirdsare the bromeliads (Bromeliaceae), the most numerous of allNeotropical epiphytes; the African-violet family (Gesneria-ceae), a family of herbs and epiphytes; the large herbHeliconia spp., related to the bananas; and the climbingpassion-flowers (Passifloraceae) (Proctor et al., 1996). Thegrowth habits of these plants might render them especially



prone to destruction by large herbivores. In Paleotropicalrain forests pollination by birds appears to be less commonthan in the Neotropics and almost confined to the canopy, intotal contrast to the hummingbird-pollinated plants (Pettet,1977; Appanah, 1981), although Cheke & Mann (2001)note that some sunbirds occasionally visit understoreyflowers.

Hummingbirds are more specialized for exploiting flowersthan any Old-World birds (Stiles, 1981) and are the onlyfamily in which most foraging is accomplished while hov-ering. This difference in foraging styles is reflected in floralmorphology (Westerkamp, 1990). By contrast, sunbirds aremore omnivorous, and will consume flowerheads, seeds andfruit (Cheke & Mann, 2001). Paleotropical flower visitorsnormally perch while feeding, and often forage in flocks(Stiles, 1981). These traits appear to make them less suitablefor pollinating understorey plants than for large forest trees.Furthermore, ornithophilous plants appear to be lessnumerous in West Africa than in other parts of Africa(Pettet, 1977), which suggests that adaptation to tropicalforest is not as great as in the Neotropics. Perhaps plantssimilar to hummingbird–pollinated plants would be eaten inthe Paleotropics.

We hypothesized that RPP might be more evenly dispersedin time in the Neotropics. Mass flowering is uncommon inthe Neotropics, but has been observed in southeast Asianrain forests (Yap & Chan, 1990). We have little informationabout this for the Afrotropics.

We hypothesized that understorey plants in the Neotropicswill be able to produce flowers over a longer period of time toattract potential pollinators. According to Bawa et al.(1990), understorey flowers produce just a few flowers at atime, often spread throughout the year. This perfectly suitsthe non-migratory habits of hermit hummingbirds (Phae-thornis spp.), but not most Paleotropical flower visitors.

Similarly, DeVries (1987) discussed the peculiar life his-tory and longevity of Heliconius and related Neotropicalbutterflies. Unlike most butterflies, Heliconius is able toutilize the nutrients in pollen in addition to nectar (Gilbert,1975). The pollen-feeding behaviour centres around a groupof Psiguria and Gurania (Cucurbitaceae) vines, which flowercontinuously over many years and produce mostly maleflowers. Perhaps these long-lived butterflies evolved only inthe Neotropics because only there is RPP consistentlyavailable.

We hypothesized that Neotropical flowers will be able toallocate more energy to enhance reproduction by thermogenicrespiration. Dynastine beetles (Scarabeidae: Dynastinae) alsoexemplify the reproductive extravagance of the Neotropics(Schatz, 1990). Approximately 900 species of Neotropicalplants may rely on dynastines for pollination. These plantsexpend considerable energy attracting their dynastine mutu-alists by elevating flower and inflorescence temperatures.

Extinct Neotropical megafauna

One might wonder why the Pleistocene megafauna in SouthAmerica did not transform Neotropical forests as do large

herbivores in the Paleotropics. If megafauna disappearedonly a few thousand years ago in the Neotropics, how couldthe presence or absence of large herbivores explain theinterhemispheric discrepancies we have described?

We suggest that there is little evidence that large herbiv-ores were common in the rain forests of the Neotropics sincethe Miocene. South America has a rich fossil record of largevertebrates, but there is no perfect method to determinewhether they lived in tropical rain forest.

We removed from consideration fossil faunas older thanthe Miocene, on the grounds that the effects produced byextremely ancient herbivores would become progressivelyless important, and less detectable, with the passage of time.We are also reluctant to include species that are believed tohave occurred outside the tropics. As many South Americanfossil localities occur outside the lowland tropics, this greatlyreduced the number of fossils for consideration.

Indeed, perhaps instead of asking why the rain forests ofthe Neotropics have so few fossils of large herbivores, maybe we should turn the question on its head and ask why thePaleotropics have so many. What percentage of large her-bivore fossils in other parts of the world actually come fromrain forests, as opposed to some other biome, such assavanna? We believe that most large herbivores, outside ofAfrica and Eurasia, have lived outside of rain forests. Again,for the sake of argument, let’s assume this is true.The question then becomes, why are Africa and Eurasia(hereafter, Afrasia) different?

The answer may have to do with the sheer size of non-forested biomes that have occurred in Afrasia during theCenozoic. A world map of savannas, such as in Bourliere(1983a), reveals a huge savanna area in Africa, a somewhatsmaller one in Asia, a medium-sized area in the Cerrado (anda smaller area in the Llanos) of South America, a large butvery isolated one in Australia, and the rest all small pieces.Furthermore, if you go back in time you will find extensivetropical savannas in Asia and subtropical savannas in Eur-ope (Potts & Behrensmeyer, 1980; Agustı & Anton, 2002).So Afrasia as a whole has been a huge staging area forsavanna vertebrates. Populations could thrive in some tem-porarily isolated pocket, speciate, then later intermingle. Areview of later Cenozoic vertebrate evolution (Potts &Behrensmeyer, 1992) suggests that alternating contractionand expansion of different environmental conditions hadmajor effects on speciation, extinction, and overall diversityof organisms linked to specific climatic conditions andvegetation.

Furthermore, savanna formations have occurred in somepart or another of Afrasia since at least the late Miocene. Forexample, although tropical savanna occupies only a smallportion of Asia now, the diversity of the savanna fauna ofthe late Miocene Siwaliks of the Indian subcontinent isunmatched today (Potts & Behrensmeyer, 1980). The largeextent of savanna biome would mean that large populationsof large animals could occur, and large populations spreadover a wide geographical area are theoretically less likely togo extinct than small and localized populations. There isgood evidence that some Old-World savanna taxa have had



enormous geographical ranges in the past: there werebaboons (Papio and Procynocephalus), ostriches (Struthio),gazelles (Gazella), cheetahs (Acinonyx), lions (Panthera),giraffes (Giraffidae), aardvarks (Orycteropodidae), and hip-pos (Hippopotamidae) in Asia as well as Africa (Potts &Behrensmeyer, 1980). If we think of biomes as islands, thenthe tropical savanna island in the Old World is old, huge,and rich in higher taxa. Paleotropical savanna could serve asa source of emigrants to other biomes, such as rain forest.Indeed, savanna mammals may have repeatedly colonizedthe forests of Africa (Kingdon, 1974). More recently, Smith& Wayne (1997) implicate the savanna/rain forest ecotone,often greater than 1000 km wide, in generating rain forestbiodiversity. With regard to the great biotic interchangebetween North and South America in the Pliocene andPleistocene, Webb (1985) and Marshall (1988) suggestedthat the greater success of the North American fauna couldbe explained in part by the much greater �staging area� andfaunal diversity in the North.

By contrast, South America was much more isolated thanAfrica, and there was no land connection between SouthAmerica and any other continent from the late Cretaceousuntil the end of the Pliocene (Carroll, 1988). We suggest thata large biome island (such as Afrasian savannas) is morelikely to produce emigrants than is a small biome island,such as the South American savanna. This occurs for severalreasons. First, there is the sheer disparity in population sizesof species that could serve as potential colonists. Secondly,large areas typically have more species than small areas(MacArthur & Wilson, 1963), hence potential colonistsfrom a small biome fragment attempting to establish abeachhead in a new biome would be up against a species-richcommunity of competitors, predators, diseases, parasites,and plants protected with potent allelochemics.

To assess this possibility we perused literature pertainingto the birds of the Neotropics and Paleotropics. We hypo-thesized that the ratio of species richness inhabiting savannaand forest would differ depending on the relative size of eachbiome on each continent. This was confirmed, assuming thatwe rely on subjective information about niche equivalenciesfrom the literature. One comparison would be of Neo-tropical jacamars (Galbulidae) with their Paleotropicalcounterparts, the bee-eaters (Meropidae). The former areoverwhelmingly forest birds, but there are approximatelythree times as many species of bee-eaters restricted to openhabitats as those found mainly in forests (Fry & Fry, 1999).A similar comparison could be made of African and SouthAmerican hornbills and toucans, using information fromKemp (1995) and Short & Horne (2001). There is only onespecies of South American toucan that prefers open habitats(Ramphastos toco). Since there are c. 25 species of toucan-like Ramphastids in lowland South America, c. 4% of themcould be said to prefer savanna. By contrast, of those Africanhornbills that exhibit a pronounced preference for eithertropical forests or savannas (defined broadly), 69% preferforests and 31% prefer savanna. Even red-legged seriemas(Cariama cristata), a quintessential savanna species fromBrazil, has a relative, the black-legged seriema (Chunga

burmeisteri), that is more often found in chaco, forest, andforest edge (Schmitt & Cole, 1981). Thus in the Cerradoscrublands, there are few bird taxa at the level of family orabove, in which most of the species are savanna specialists.

We also hypothesized that the Brazilian Cerrado, whichoccupies a much smaller fraction of the South Americancontinent than does forest, would show evidence of sub-stantial colonization by forest fauna. This is confirmed.Marinho-Filho et al. (2002) note that fewer than 17% of theCerrado mammal species are restricted to open areas, 29%are exclusive to forests, and 54% occur in both. Only onespecies of Cerrado cervid, the pampas deer Ozotocerosbezoarticus, rarely occurs in forest, and even this specieslacks some features found in extreme savanna dwellers (e.g.zebras and wild asses, family Equidae) in Afrasia. Forexample, pampas deer are not highly specialized grazers(Rodrigues & Moneiro-Filho, 1999), and they apparentlylack endurance and produce neonates that hide rather thanimmediately follow the mother (Geist, 1998).

Movement into more open habitats, such as woodlandsavanna, savanna, and scrub, favoured a suite of adaptationsamong ungulates (Vaughan, 1978; Geist, 1998). Theseinclude, but are not limited to: large body size, springingligaments in the foot, reduction of toe number, elongation ofdistal foot elements, elongation of the digestive tract,rumination, development of a rete mirabile to cool the brainduring running, and high-crowned or ever-growing teeth.Not all ungulates of open habitats possess all of theseadaptations, but it is conceivable that successful colonists ofopen habitats possessed most of them. Note that with thepossible exception of pacas, all LRFH in the Neotropicsbelong to families that immigrated from North America,which had an extensive fauna of savanna or savannawoodland herbivores in the Miocene (MacFadden, 2000).Furthermore, some of these may have had their ultimateorigins in Afrasia. For example, deer, family Cervidae,occurred first in Afrasia, then North America, then SouthAmerica. We suggest that some of the adaptations acquiredby savanna species have proved useful in colonizing rainforest.

Nevertheless, although the large mammals of rain forestsmight have originated elsewhere, it also seems plausible thatfew extra-forest large herbivores might have the character-istics necessary to colonize rain forests. We suggest that onlyvery large extents of savanna or similar biome would pro-duce a diverse enough array of species of which at least a fewhave the characteristics needed to successfully colonize rainforest. The history of the planet is full of examples of specieswith special adaptations colonizing new areas. Some of thecharacteristics acquired by herbivores during the extra-forestphase – such as high mobility – might enable large herbivoresto colonize rain forest.

One could argue that the paleofaunas existing prior tohuman contact demonstrate that large herbivores inhabitedtropical rain forests. The evidence for this is, however,equivocal. First, many South American paleofaunas aredominated by grazers (Anaya & MacFadden, 1995) or haveother characteristics indicative of open habitats. To consider



just one example, the pollen flora of the Pleistocene Tarijasite in Bolivia is composed almost entirely of grass species,and therefore the locality is reconstructed as a predomin-antly dry open-country grassland with scattered trees andshrubs that were mostly concentrated along the margins ofrivers and lakes (Yoshida & Yamazaki, 1982). Based uponthis and morphological and isotopical data from large her-bivore fossils, MacFadden & Shockley (1997) consideredmany of the large herbivores from the Tarija site to be gra-zers. Although savannas are known to be inhabited by richfaunas of browsers and grazers, a diverse fauna of largegrazers is inconsistent with pristine tropical rain forest ha-bitat. Since most post-Oligocene, South American faunalassemblages contain several species of grazers (MacFadden,2000), we find it difficult to envision rain forest as a suitablecontext for them. Furthermore, many fossil sites in SouthAmerica are at too high a latitude or too high an altitude tosustain tropical rain forest. Thus, many of the large herbi-vores that lived during the later Cenozoic in South Americamay not have inhabited tropical rain forest.

The caveats of Kay & Madden (1997), which pertain toone of the best-studied Neotropical �forest� paleofaunascontaining large herbivores, are particularly instructive.They note that we are ignorant of nearly all taphonomicprocesses in tropical lowland forest environments. Hence,taphonomic and sampling bias may impair our knowledge ofwhat species were present, such that taxa at the same localitycould have been derived from several different habitats.They further note that the percentage of arboreal species isnot as high as that commonly found in modern rain forests.They concluded that the site was more likely a forest mosaic(e.g. disturbed riparian succession), rather than a continu-ous, uninterrupted, multistratal, evergreen forest. We areunable to locate any paleofaunas from South America forwhich we are reasonably certain that large herbivores werecommon in uninterrupted, multistratal, evergreen forest.Large herbivores may have been common in the vicinity ofrain forests in secondary forests (e.g., rhino referencesabove). Kay & Madden (1997) specifically mentionedriparian mosaic, aquatic, river-margin, and tree-fall gaphabitats as having the potential to sustain an abundant faunaof large herbivores. They also noted that today, openclearings are maintained within forests by the activities ofmegaherbivores (Laws, 1970; Owen-Smith, 1988). Thenatural history information, we presented earlier pertainingto African forest elephants and Asian rhinos suggests muchthe same thing. We, therefore, envision ancient Neotropicalforests virtually empty of large herbivores in their interiors,but sustaining them at the edges. Pleistocene extinctions ofNeotropical megaherbivores would not have adverselyaffected the rich fauna of small herbivores, frugivores, andflower-feeders, and might even have been of benefit to them.

The co-existence of primary forest and savanna faunas inclose proximity still occurs in South America. For example,the Cerrado landscape of Brazil and Bolivia, which is amosaic of grasslands, shrublands, gallery forests, and morecontinuous woodlands, is home to a mix of mammal speciesfrom both open habitats and closed forest. In close juxta-

position can be found tree sloths (Bradypus variegatus) andfour genera of primates (Callithrix, Aotus, Cebus, andAlouatta), as well as pampas deer and hoary foxes (Pseu-doalopex vetulus) (Marinho-Filho et al., 2002). A similarinterdigitation of forest-dwelling and open-habitat species isalso apparent in the avifauna (Macedo, 2000). Note, how-ever, that this does not mean that all the species are evenlydistributed within the habitat mosaic: a tree sloth and ahoary fox do not occur in the same habitat. We suggest thatvirtually all the trophic impacts of large extinct Neotropicalherbivores could have been confined to areas outside closed-canopy tropical evergreen forest, although the animals mightpenetrate the forest temporarily to escape direct exposure tohot sunlight, or seek cover from predators.

We hypothesized that Africa has more families ofsavanna-adapted birds than the Neotropics. This was con-firmed by Fry (1983). Similarly, until recent deforestation byhumans, the lowlands of Madagascar were largely forested,and the savanna avifauna of Madagascar is depauperatecompared to that of mainland Africa. Madagascar lacks overtwenty African families, including many with importantsavanna representatives (Dorst, 1972).

By contrast, although savanna dominates a large portionof the Old-World tropics, there are places elsewhere thattropical forest, by virtue of its large extent, dominates openbiomes. Howell (1971) notes that small savannas in Nicar-agua have only twenty-six resident bird species, only a tinyfraction of the number found in the Llanos or Cerrado.

The distribution and abundance of extant large herbivoressuggests that large body size may be a handicap to rain forestspecies. We, therefore, hypothesized that selection againstlarge body size would be discernible in forest races or siblingspecies of those species that occupy both savanna and rainforest. This is confirmed for several species: in Africa forestelephants, and buffalo are smaller than their savanna (bush)counterparts (Kingdon, 1974). The same principle also oftenseems to be the case when we compare species within thesame family: the forest-dwelling hyraxes (Procaviidae), hip-pos (Hippopotamidae), giraffes and okapis (Giraffidae) andpantherine cats (genus Panthera) are smaller than their open-country counterparts. In a review of Cenozoic vertebratesand their habitats, Potts & Behrensmeyer (1992) noted thatwith increasing aridity and decreasing tree cover in the mid-Miocene, larger-bodied rhinos replaced the medium-sizedforms. In South America, the largest cavy (family Caviidae),Dolichotis patagonum, occurs outside of rain forest. Thesame is true for deer (Blastoceros dichotomus), toucans(Ramphastos toco), and canids (Chrysocyon brachyurus).Furthermore, the jaguars (Panthera onca) are considerablylarger in the primarily open Pantanal wetlands than in therain forests of Amazonia and Belize (Sunquist & Sunquist,2002).

Note also that all the Paleotropical rain forest herbivoreswith a greater biomass than that of the largest Neotropicalherbivore belong to just three orders and three to six familiesin any given forest: Proboscidea (Elephantidae), Perissodac-tyla (Rhinocerotidae), and Artiodactyla (Bovidae and per-haps one species each from Suidae, Hippopotamidae and



Cervidae). Several smaller species are similar in size toNeotropical tapirs, including the Malayan tapir (Tapirusindicus), gorilla (Gorilla gorilla), and okapi (Okapia john-stoni). Note, however, that the same families are not equallyrepresented in both Africa and Asia: Africa lacks rain forestrhinos, deer, and tapirs, whereas Asia lacks rain forestgiraffids and hippos. Furthermore, several species seem tolack the full suite of key adaptations that could occur in rainforest herbivores. Kingdon (1974) suggested that the smallergeographical range of gorillas than of chimpanzees (Pantroglodytes) reflects an incomplete adaptation to a terrestrialherbivore niche. In particular, if gorillas are compared withsimilar-size bovids, it can be seen that they have relativelyunspecialized digestive tracts and are less cursorial.

The anthropoid primates also support a New-World/Old-World dichotomy. The earliest Old-World anthro-poids were adaptively more similar to the New-Worldplatyrrhines than to later anthropoids in their relativelysmall size, paucity of folivorous taxa, and absence ofterrestrial forms (Kay & Simons, 1980; Fleagle & Kay,1985). This suggests that Paleotropical forests were latersubjected to a drying of the climate and reduction of treedensity (McCrossin et al., 1998), that selected for terrest-rial adaptations in primates. Ancestral Cercopithecus, themost speciose African primate genus, passed through astage in which their habits were more terrestrial and theirhabitat less densely forested (Kingdon, 1974). Their sub-sequent success upon recolonizing forest could have beenmade possible by adaptations acquired in savanna orwoodland savanna. Note that the Old-World cercopit-hecids have either cheek pouches or sacculated stomachs,adaptations that could have arisen as a result of foragingconstraints imposed by environmental change. Kingdon(1997) also pointed out that the subfamily Cercopithecinaeis much more speciose in Africa than is the subfamilyColobinae, and that the former recolonized forest morerecently. It thus seems plausible that the adaptations thatthe Cercopithecinae acquired in non-forest environmentswere particularly beneficial later. Note also that trueruminants, which may have evolved their food processingspecialization as habitats opened up (Geist, 1998), did notevolve in South America, but colonized it.

Furthermore, climatic changes that increased savanna atthe expense of forest might have eliminated some forest-dependent species, thereby providing unexploited forestresources for immigrant species. The general picture thatemerges is that in Africa – and perhaps in Miocene Asia – thesavanna biome has been so extensive, and the large mammalfauna so rich, that understudies have always been availablein the savanna to enter the rain forest evolutionary theatreand fill in for missing forest fauna.

The propensity of large herbivores to disappear frombiome fragments could be explored by examining the faunaof islands that once were part of continuous habitat but arenow just fragments. We, therefore, hypothesized that thespecies richness of large mammals from the islands ofSundaland would be less than on the Asiatic mainland. Thiswas confirmed: although various islands in the area have

several species of large mammals, none has as many asmainland Thailand (Francis, 2001).

In the Neotropics, and in most tropical areas of the worldoutside of Afrasia, forest predominates over savanna. We,therefore, hypothesized that the number of species in thesesmaller savannas would be less than in Afrasia. This wasconfirmed for ungulates, using data from Bourliere (1983b).Thus, the Llanos of South America has only two ungulatespecies, and both are derived from forest. The Cerrado isconsiderably larger, so there was more opportunity forevolution and persistence of savanna fauna, and there are sixor seven species of ungulates present. Overall, however,many of the Cerrado vertebrate species are either generalizedforms that can live in both grassland or forest, or they areclosely related to, and presumably derived from, forestforms.

Therefore, we assume that small biome islands (such asNeotropical savannas enmeshed in a matrix of tropical for-est, Caatinga, and Chaco) are more likely to import immi-grants than to export emigrants. We suggest that the mainreason that there have been so many large herbivores in theAfrasian rain forest during the latter half of the Cenozoic, isthat they have colonized from outside the forest. We,therefore, theorized that rain forest is not a suitable biomefor most large herbivores, as we can see by looking attropical forests in the rest of the world.

We have conjectured that there may be an adaptation oradaptations that animals acquire when outside the rainforest, that they or their descendants possess when theyrecolonize. The conquest of land by tetrapods might be aninstructive analogy. Evolutionary biologists argue that fishevolved legs and toes not to walk on land but in order tomove underwater (Zimmer, 1998). The morphology of thefirst tetrapods could then be considered a pre-adaptation orexaptation to enhanced vagility on land. Perhaps large her-bivores in rain forests inherited enhanced vagility from theirextra-forest ancestors. Enhanced mobility would enable thecolonists to acquire the widely distributed resources theyneed, including a variety of browse and fruits, mineral licks,drinking water, and mud wallows. This is consistent with thefossil record, which indicates that ungulates became highlycursorial prior to their alleged predators and a need for largeherbivores to move long distances to track resources (Janis& Wilhelm, 1993).

We hypothesized that most large herbivores living in rainforest in the Holocene are more mobile, and move morewidely over the landscape to meet their needs than did mostextinct, large Neotropical herbivores. We have no directmeans to compare the mobility of fossil Neotropical herbi-vores with herbivores from the Holocene. We do, however,know that Holocene rain forest herbivores often move largedistances, as we noted above.

By contrast, some large Neotropical herbivores wouldhave been hard pressed to move long distances; the multi-tonsloths of several families come to mind. For example, two ofthe three genera of fossil sloths at one sight were highlymodified for digging (White, 1997; Bargo et al., 2000),compromising the speed of above-ground, terrestrial



locomotion. Perhaps unsurprisingly, McDonald (1997) sug-gested that the earliest ground sloths inhabited open wood-land, and a preference for open or semi-open habitatspersisted throughout the latter part of the Cenozoic. Simi-larly, only very slow locomotion was likely to have beenpossible for the tank-like glyptodonts, family Glyptodonti-dae (Carroll, 1988). If one considers the six genera of cin-gulates (armored mammals such as armadillos andglyptodonts) at one of the better-studied sites, LaVenta inColombia, no single habitat type emerges (Carlini et al.,1997). Only one of the genera present, the 2-kg armadilloNanoastegothermium, was considered to be restricted tohumid forest, and only one other genus, Anadasypus, eveninhabited humid forest. All of the genera larger than 15 kgwere considered to be primarily inhabitants of savannas.

Food habits can also be suggestive of habitat. Althoughbrowsers occur in both forest and savanna, a diverse array ofgrazers is inconsistent with unbroken primary rain forest.Furthermore, grazers tend to be large, so as to accommodatelong digestive tracts that can process low-nutritive-valueforage (MacFadden, 2000). There is usually a strong pre-dominance of grazers in the mammalian paleofaunas ofSouth America, which suggests a period during the middleTertiary of extensive grasslands that were inhabited byendemic Neotropical herbivores, such as notoungulates(MacFadden, 1995). We, therefore, hypothesized that manyNeotropical fossil herbivores would possess dental charac-teristics, such as hypsodonty, associated with open-countryhabitat. This was confirmed. A review of fossil mammalsfrom South America indicates that many species of largeherbivores were grazers; in fact, some achieved hypsodontymillions of years before their supposed counterparts in NorthAmerica (MacFadden, 2000). Unfortunately, few fossilassemblages can be unequivocally assigned to a particularbiome; furthermore, the fossil record for rain forests is poor.

Large herbivores in small biomes may be prone toextinction, as follows. One could also make predictionsabout how many endemics should occur in a taxon, basedupon population size. That is, species that persist in largepopulation sizes should be less prone to extinction, hencethere should be more surviving endemics in taxa with largepopulation sizes. It follows that species with large geo-graphical ranges should be less prone to extinction, since(other things being equal), they will have larger populationsizes (MacArthur & Wilson, 1963; Brown & Maurer, 1987).As plants are more abundant than vertebrates, there shouldbe more endemic plants than vertebrates. Likewise, asectotherms are more abundant than endotherms, we wouldpredict that there should be more endemic herptiles thanendemic birds and mammals. Thus, we hypothesized that thepercentage of endemics of a typically-sized biome decreasefrom plants, to ectothermic vertebrates, to endotherms.Furthermore, since large endotherms exist at lower densitiesthan small endotherms, we expect that the larger endemicendotherms will have a higher extinction rate, hence therewill be more endemics among smaller species. Thesehypotheses were confirmed. The flora of the Cerrado ofBrazil and Bolivia is perhaps the richest of any savanna,

c. 10,000 species, and about half are endemic (Heringeret al., 1977; Oliveira-Filho & Ratter, 2002). The endemismof the Cerrado herpetofauna is c. 21% (Colli et al., 2002),and the avifauna, which at more than 800 species (somespecialists say over 900 species) is very rich, but has only c.4% endemics (Silva, 1995). Finally, within the Cerradotheriofauna it is the small species that are endemic (Marinho-Filho et al., 2002); presumably the larger population sizes ofsmall endemic mammals tend to buffer them from extinction.

New Guinea exhibits a similar pattern. Using figures fromBonaccorso (1998), the endemism for vascular plants rangesfrom 70% to 76%; for amphibians, 67%; for mammals,24%; and for birds, 11%. Hence, our hypothesis was sup-ported; over long spans of geological time, the guild ofspecies of large herbivores accumulating in a typical tropicalforest region, such as in the Neotropics, will not be partic-ularly large. Only in Afrasia are conditions optimal for ahigh species richness of LRFH to develop.

We earlier noted that the activities of large herbivoresmight tend to retain habitat in a state compatible with theirpersistence. It is also plausible that the activities of somelarge herbivores could increase the likelihood that otherswould colonize. We have noted that large herbivores intropical forests disperse large fruits and create wallows,paths and gaps into which forage plants can grow. PerhapsLRFH are less likely to colonize rain forest if other LRFHhave not already become successfully established.

We have no reason to assume that the effects of largeherbivores follow an all-or-nothing pattern. We noted abovethat tree density increased gradually with a decrease inspecies richness of large herbivores on Asian islands. Thus, itis quite plausible that large herbivores were present in thePleistocene forests of the Neotropical rain forests at lowspecies richness and low biomass densities, thereby affectingthe forests in only a minor way.

A diverse array of large herbivores on New Guinea priorto human occupancy of the area might have been expected toalter tree diversity and abundance, so that the aforemen-tioned trend would be incongruous. We, therefore, hypo-thesized that New Guinea would not have a diverse record ofextinct LRFH. This was confirmed: the paucity of largeherbivores in New Guinea is not new. Flannery (1995)reported that none of the species of extinct theriofauna wasvery large and only ten species (<5% of the total) are knownor suspected to have become extinct during the late Qua-ternary. Furthermore, the largest New Guinea mammalknown, which weighed 200–400 kg, may not have survivedpast the Pliocene.

Milewski & Diamond (2000) argue that three micro-nutrients (iodine, cobalt and selenium) are necessary for themetabolism of large herbivores, and that the scarcity ofsources of these micronutrients limits large herbivore abun-dance and distribution. They considered South America tobe a nutrient-poor continent that did not have the potentialto attain the biomass densities of large mammals that prevailin some of the more pristine parts of Africa and Asia. Even ifa few species of LRFH were present in Neotropical forests,their relatively small biomass densities may have hardly



affected the evolution of primary Neotropical rain forests.Indeed, a survey of Pleistocene mammals from westernAmazonia did not reveal a single extinct taxa confined toforest habitat; at best, some could be considered to be forestedge species (Rancy, 1999).

As a final comment on paleofaunas, we would be remiss ifwe did not explain why Central America has not been dis-cussed. The reason is simply that there are confoundingvariables that frustrate easy analysis. These include thepresence of savanna vertebrates in North America, the rug-ged topography (which might allow herbivores to climbfrom one biome to another), the relative recency of emer-gence of the Panamanian Isthmus, and the fact that many ofthe potential rain forest colonists originate in temperateclimates. Thus, we emphasized Neotropical rain forests inSouth, not Central America.

CONCLUSION

Large herbivores affect rain forests in many ways, only someof which have been addressed herein. Indeed, earlier draftsof this paper contained many untested hypotheses. Given theknown importance of large herbivores in Paleotropical for-ests, it is rather surprising that few have speculated on theconsequences of a lack of large herbivores in Neotropicalforests. This question is of more than academic interest,because the threats faced by large herbivores in the Paleo-tropics, and the introduction of exotic livestock, havemodified selection pressures on these forests. We suggest thatanthropogenic disturbances introduced by humansthroughout the low-latitude humid tropics can bring aboutthe same selective pressures that can functionally homo-genize biotas on a pantropical scale, thus replacing the verydifferent, natural selection pressures that had previouslyoperated in the Paleo- and Neotropics.

ACKNOWLEDGMENTS

Many of the ideas presented in this paper evolved from ourdays in graduate school at the University of Florida (1984–88). We thank our fellow graduate students and professorsat that time, including John Eisenberg, John Robinson, BrianMcNab, Kent Redford, Jay Malcolm and Peter Feinsinger,for the initial impetus, stimulating discussions and encour-agement. For additional discussions over the years of trop-ical communities and species diversity we thank JohnTerborgh, Carel van Schaik, Al Gentry, Louise Emmons,Colin Chapman, Andrew Jones, Richard Bodmer and JohnFa. Carlos Peres� fieldwork in Neotropical forests (1984–2002) has been funded by the Wildlife Conservation Societyand Conservation International. We thank Anina Carkeekfor constructive comments on the manuscript.

REFERENCES

Agresti, A. & Agresti, B.F. (1979) Statistical methods for thesocial sciences. Dellen Publishing Company, San Francisco,CA, p. 554.

Agustı, J. & Anton, M. (2002) Mammoths, sabertooths, andhominids: 65 million years of mammalian evolution inEurope. Columbia University Press, New York.

Ananya, F & MacFadden, B.J. (1995) Pliocene mammals fromInchasi, Bolivia: The endemic fauna just before the GreatAmerican Interchange. Bulletin of the Florida Museum ofNatural History, 39, 87–140.

Appanah, S. (1981) Pollination in Malaysian primary forests.Malaysian Forester, 44, 37–42.

Axelrod, D.I. (1952) A theory of angiosperm evolution.Evolution, 6, 29–59.

Axelrod, D.I. (1972) Ocean-floor spreading in relation to ecosys-tematic problems. University of Arkansas Museum occasionalpapers, Vol. 4. (ed. by R.T. Allen and F.C. James), pp. 15–76.University of Arkansas Press, Fayetteviller Arkansas.

Baker, H.G. & Baker, I. (1983) Floral nectar sugar constituentsin relation to pollinator type. Handbook of experimentalpollination biology (ed. by C.E. Jones and R.J. Little),pp. 117–141. Van Nostrand Reinhold, New York.

Bargo, M.S., Vizcaıno, S.F., Archuby, F.M. & Blanco, R.E.(2000) Limb bone proportions, strength and digging in someLujanian (Late Pleistocene-Early Holocene) mylodontidground sloths (Mammalia, Xenarthra). Journal of VertebratePaleontology, 20, 601–610.

Barlow, C. (2000) The ghosts of evolution: nonsensical fruit,missing partners, and other ecological anachronisms, p. 291.Basic Books, New York.

Barlow, C. & Wacher, T. (1997) A field guide to the birds of theGambia and Senegal, p. 400. Yale University Press, NewHaven.

Basset, Y. (1995) Arthropod predator-prey ratios on vegetation atWau, Papua New Guinea. Science in New Guinea, 21, 103–12.

Bawa, K.S., Ashton, P.S., Nor, S.M. (1990) Reproductiveecology of tropical forest plants: management issues. Repro-ductive ecology of tropical forest plants (ed. by K.S. Bawa andM. Hadley), UNESCO and Parthenon Publishing, Paris,pp. 3–13.

Bazzaz, F.A., Chairello, N.R., Coley, P.D., Pitelka, L.F. (1987)Allocating resources to reproduction and defense. Bioscience,37, 58–67.

Beehler, B.M., Pratt, T.K. & Zimmerman, D.A. (1986) Birds ofNew Guinea. Princeton University Press, Princeton, NJ.

Bernard, E. (2001) Vertical stratification of bat communities inprimary forests of Central Amazon, Brazil. Journal ofTropical Ecology, 17, 115–126.

Bernays, E.A. (1998) Evolution of feeding behavior in insectherbivores. Bioscience, 48, 35–44.

Bignell, D.E. & Eggleton, P. (2000) Termites in ecosystems.Termites: evolution, sociality, symbioses, ecology (ed. byT. Abe, M. Higashi and D.E. Bignell), pp. 363–387. KluwerAcademic, Dordrecht, The Netherlands.

Bignell, D.E. & Tayasu, I. (2001) Feeding groups, lifetypes andthe global ecology of termites. Ecological Research 16, 941–960.

Blake, S., Douglas-Hamilton, I. & Karesh, W.B. (2001) GPStelemetry of forest elephants in Central Africa: results of apreliminary study. African Journal of Ecology, 39, 178–186.

Bodmer, R.E. (1990) Fruit patch size and frugivory in thelowland tapir (Tapirus terrestris). Journal of Zoology, 222,121–128.



Bonaccorso, F.J. (1998) Bats of Papua New Guinea. Conserva-tion International, Washington, DC.

Bonser, S.P. & Reader, R.J. (1995) Plant competition andherbivory in relation to vegetation biomass. Ecology, 76,2176–2183.

Borner, M. (1979) A field study of the Sumatran rhinoceros.Unpublished Doctoral thesis, University of Basel, Basel.

Bourliere, F. (1973) The comparative ecology of rainforestmammals in Africa and Tropical America: some introductoryremarks. Tropical forest ecosystems in Africa and SouthAmerica: a comparative review (ed. by B.J. Meggers, E.S.Ayensu, W.D. Duckworth) pp. 272–292. Smithsonian Insti-tution Press, Washington, DC.

Bourliere, F. (1983b) The savanna mammals: introduction.Ecosystems of the world 13: tropical savannas (ed. byF. Bourliere), pp. 359–361. Elseviere Scientific, Amsterdam.

Bourliere, F. (Ed.) Ecosystems of the world 13: tropicalsavannas. Elseviere Scientific, Amsterdam.

Bradbury, J.W. (1977) Lek mating behavior in the hammer-headed bat. Zeitschrift Fur Tierpsychologie, 45, 225–55.

Brown, J.H. & Maurer, B.A. (1987) Evolution of speciesassemblages: effect of energetic constraints and speciesdynamics on the diversification of the North Americanavifauna. American Naturalist, 130, 1–17.

Buechner, H.K. & Dawkins, H.C. (1961) Vegetation changeinduced by elephants and fire in Murchison Falls NationalPark, Uganda. Ecology, 42, 752–66.

Carlini, A.A., Vizcaıno, S.F. & Scillato-Yane, G.J. (1997)Armored xenarthrans: a unique taxonomic and ecologicassemblage. The Miocene fauna of La Venta, Colombia (ed.by R.F. Kay, R.H. Madden, R.L. Cifelli and J.J. Flynn) pp.213–226. Smithsonian Institution Press, Washington, DC.

Carlquist, S.H. (1965) Island life: a natural history of the islandsof the world, p. 451. Natural History Press, Garden City, NY.

Carlquist, S.H. (1974) Island biology, p. 660. Natural HistoryPress, Garden City, NY.

Carroll, R.L. (1988) Vertebrate paleontology and evolution.W.H. Freeman and Company, New York.

Chapman, C.A., Gautier-Hion, A., Oates, J.F., Onderdonk,D.A. (1999) African primate communities: determinants ofstructure and threats to survival. Primate communities (ed. byJ.G. Fleagle, C.H. Janson, and K.E. Reed), pp. 1–37.Cambridge University Press, Cambridge.

Charles-Dominique, P. (1983) Ecology and social adaptations indidelphid marsupials: comparison with eutherians of similarecology. Advances in the study of mammalian behavior. (ed.By J.F. Eisenberg & D.V. Kleiman), pp. 395–422. SpecialPublication No. 7, American Society of Mammalogists,Shippensburg, Pennsylvania.

Cheke, R.A. & Mann, C.F. (2001) Sunbirds: a guide to thesunbirds, flowerpeckers, spiderhunters and sugarbirds of theworld, p. 384. Cambridge University Press, New Haven andLondon.

Clements, J.F. (2000) Birds of the world: A checklist, 6th edn,p. 867. Ibis Publishing, Vista, CA.

Colli, G.R., Bastos, R.P. & Aruajo, A.F.B. (2002) The characterand dynamics of the Cerrado herpetofauna. The Cerrados ofBrazil: ecology and natural history of a Neotropical savanna(ed. by P.S. Oliveira & R.J. Marquis), pp. 223–241. Colum-bia University Press, New York.

Courts, S.E. (1998) Dietary strategies of Old World fruit bats(Megachiroptera, Pteropodidae): how do they obtain suffi-cient protein? Mammal Review, 28, 185–194.

Cristoffer, C. (1991) Scaling of body weight in bats and rodents.Journal of Mammalogy, 72, 731–33.

Cristoffer, C. (1987) Body size differences between New Worldand Old World, arboreal, tropical vertebrates: cause andconsequences. Journal of Biogeography, 14, 165–172.

Cumming, D.H.M. & Brock, F.M. (1997) Elephants, wood-lands and biodiversity in southern Africa. South AfricanJournal of Science, 93, 231–236.

Curran, L.M. & Leighton, M. (2000) Vertebrate responses tospatiotemporal variation in seed production of mast-fruitingDipterocarpaceae. Ecological Monographs, 70, 101–128.

DeVries, P.J. (1987) The Butterflies of Costa Rica and theirnatural history. Volume I: Papilionidae, Peridae, Nymphali-dae, p. 327. Princeton University Press, Princeton, NJ.

Delph, L.F., Johannsson, M.H., Stephenson, A.G. (1997) Howenvironmental factors affect pollen performance: ecologicaland evolutionary perspectives. Ecology, 78(6), 1632–9.

Diamond, J. (1996) Overview: laboratory experiments, fieldexperiments, and natural experiments. Community ecology(ed. by J. Diamond and T.J. Case), pp. 3–22. Harper andRow, New York.

Dinerstein, E. (1992) Effects of Rhinoceros unicornis on riverineforest structure in lowland Nepal. Ecology, 73, 701–704.

Doak, D.F. (1992) Lifetime impacts of herbivory for a perennialplant. Ecology, 73, 2086–99.

Dorst, J. (1972) Evolution and affinities of the birds ofMadagascar. Biogeography and ecology of Madagascar (ed.by G. Richard-Vindard and R. Batistini). Monographs inBiology 30, 615–627.

Eisenberg, J.F. (1978) The evolution of arboreal herbivores inthe class Mammalia. The ecology of arboreal folivores (ed. byG.G. Montgomery), pp. 135–52. Smithsonian InstitutionPress, Washington, DC.

Eisenberg, J.F. (1981) The Mammalian radiations, p. 610.University of Chicago Press, Chicago.

Emmons, L. & Gentry, A. (1983) Tropical forest structure andthe distribution of gliding and prehensile-tailed vertebrates.American Naturalist, 121, 513–24.

Emmons, L.H., Gautier-Hion, A., Dubost, G. (1983) Commu-nity structure of the frugivorous-folivorous forest mammalsof Gabon. Journal of Zoology, London, 199, 209–222.

Fa, J.E. & Peres, C.A. (2001) Game vertebrate extraction inAfrican and Neotropical forests: an intercontinental com-parison. Conservation of exploited species (ed. by J.D.Reynolds, G.M. Mace, K.H. Redford, J.G. Robinson),pp. 203–41. Cambridge: Cambridge University Press, Cam-bridge.

Fa, J.E. & Purvis, A. (1997) Body size, diet and populationdensity in Afrotropical forest mammals: a comparison withNeotropical sites. Journal of Animal Ecology, 66, 98–112.

Fenton, M.B., Brigham, R.M., Mills, A.M., Rautenbach, I.L.(1985) The roosting and foraging areas of Epomophoruswahlbergi (Pteropodidae) and Scotophilus viridis (Vespertili-onidae) in Kruger National Park, South Africa. Journal ofMammalogy, 66, 461–8.

Fenton, M.B., Cumming, D.H.M., Rautenbach, I.L., Cum-ming, G.S., Cumming, M.S., Ford, G., Taylor, R.D.,



Dunlop, J., Hovorka, M.D., Johnston, D.S., Portfors, C.V.,Kalcounis, M.C. & Mahlanga, Z. (1998) Bats and the lossof tree canopy in African woodlands. Conservation Biology,12, 399–407.

Flannery, T. (1995). Mammals of New Guinea, CornellUniversity Press, Ithaca, NY.

Fleagle, J.G. & Kay, R.F. (1985) The paleobiology ofcatarrhines. Ancestors: the hard evidence (ed. by E. Delson),pp. 23–36. Alan R. Liss, New York.

Fleming, T.H. (1988) The short-tailed fruit bat. A study in plant-animal interactions. University of Chicago Press, Chicago.

Fleming, T.H., Breitwisch, R., Whitesides, G.H. (1987) Patternsof tropical vertebrate frugivore diversity. Annual Review ofEcology and Systematics, 18, 91–109.

Fonseca, G.A.B., Herrmann, G., Leite, Y.L.R., Mittermeier,R.A., Rylands, A.B., Patton, J.L. (1996) Lista anotada dosmamıferos do Brasil. Occasional Papers in ConservationBiology No. 4, Conservation International, Washington, DC.

Francis, C.M. (2001) A photographic guide to mammals ofSouth-East Asia, including Thailand, Malaysia, Singapore,Myanmar, Laos, Cambodia, Vietnam, Java, Sumatra, Baliand Borneo. New Holland Publishers (UK) Limited, London.

Fry, C.H. (1983) Birds in Savanna Ecosystems. Ecosystems ofthe world 13: tropical savannas (ed. by F. Bourliere), pp. 337–357. Elseviere Scientific, Amsterdam.

Fry, C.H. & Fry, K. (1999) Kingfishers, bee-eaters, & rollers,p. 324. Princeton University Press, Princeton, NJ.

Ganesh, T. & Davidar, P. (1999) Fruit biomass and relativeabundance of frugivores in a rain forest of southern WesternGhats, India. Journal of Tropical Ecology, 15, 399–413.

Gardner, A.L. (1977) Feeding habits. Biology of bats of the NewWorld family Phyllostomatidae. Part 1 (ed. by R.J. Baker, J.K.Jones and D.C. Carter) Special Publications of the MuseumTexas Tech University, 10, 293–350.

Gascon, C., Malcolm, J.R., Patton, J.L., da Silva, M.N.F.,Bogart, J.P., Lougheed, S.C., Peres, C.A., Neckel, S., Boag,P.T. (2000) Riverine barriers and the geographic distributionof Amazonian species. Proceedings of the National Academyof Sciences, 97, 13672–13677.

Geist, V. (1998) Deer of the world: their evolution, behavior,and ecology. Stackpole Books, Mechanicsburg, Pennsylvania.

Gentry, A.H. (1988) Changes in plant community diversity andfloristic composition on environmental and geographicalgradients. Annals of the Missouri Botanical Garden, 75, 1–34.

Gilbert, L.E. (1975) Ecological consequences of a coevolvedmutualism between butterflies and plants. Plant and animalcoevolution (ed. by L.E. Gilver & P.H. Raven), pp. 210–40.University of Texas Press, Austin.

Graham, C. (2001) Habitat selection and activity budgets ofkeel-billed toucans at the landscape level. Condor, 103, 776–84.

Grajal, A., Strahl, S.D., Parra, R., Dominguez, M.G., Neher A.(1989) Foregut fermentation in the hoatzin, a Neotropicalleaf-eating bird. Science, 245, 1236–1238.

Greene, H.W. (1997) Snakes: The evolution of mystery innature, p. 351. University of California Press, Berkeley.

Grimmett, R., Inskipp, C., Inskipp, T. (1999) A guide to thebirds of India, Pakistan, Nepal, Bangladesh, Bhutan, SriLanka and the Maldives, p. 888. Princeton University Press,Princeton, NJ.

Gronemeyer, P.A., Dilger, B.J., Bouzat, J.L., Paige, K.N. (1997)The effects of herbivory on paternal fitness in scarlet gilia:better moms also make better pops. American Naturalist,150, 592–602.

Groves, C.P. (1967) On the rhinoceroses of South-East Asia.Saugetierkundliche Mitteilungen, 15, 221–237.

Guix, J-C., Ruiz, X. & Jover, L. (2001) Resource partitioningand interspecific competition among coexisting species ofguans and toucans in SE Brazil. Netherlands Journal ofZoology, 51, 285–297.

Haffer, J. (1969) Speciation in Amazonian forest birds. Science,165, 131–137.

Haffer, J. (1974) Avian speciation in South America. Publica-tions of the Nuttall Ornithological Club, 14, 1–390.

Hainsworth, F.R. & Wolf, L.L. (1976) Nectar characteristicsand food selection by hummingbirds. Oecologia (Berlin), 25,101–113.

Handley, C.O. Jr, Wilson, D.E., Gardner, A.L. (1991) Demo-graphy and natural history of the common fruit bat, Artibeusjamaicensis, on Barro Colorado Island, Panama. SmithsonianContributions to Zoology, 511, 89–130.

Harper, J.L. (1969) The role of predation in vegetationaldiversity. Diversity and stability in ecological systems (ed. byG.M. Woodwell & H.H. Smith), pp. 48–62. SymposiumBrookhaven National Laboratory, Upton, NY..

Hart, J.A. & Hart, T.B. (1988) A summary report on thebehaviour, ecology, and conservation of the okapi (Okapiajohnstoni) in Zaire. Acta Zoologica et Pathologica Antver-piensia, 80, 19–28.

Hart, J.A. & Hart, T.B. (1989) Ranging and feeding behaviourof okapi (Okapia johnstoni) in the Ituri Forest of Zaire: foodlimitation in a rainforest herbivore? Symposium ZoologicalSociety of London, 61, 31–50.

Hawthorne, W.D. & Parren, M.P.E. (2000) How important areforest elephants to the survival of woody plant species inUpper Guinean forests? Journal of Tropical Ecology, 16,

133–150.Hegarty, E.E. & Caballe, C. (1991) Distribution and abundance

of vines in forest communities. The biology of vines (ed. byF.E. Putz & H.A. Mooney), pp. 313–35. Cambridge Univer-sity Press, Cambridge.

Heithaus, E.R. & Fleming, T.H. (1978) Foraging movements ofa frugivorous bat, Carollia perspicillata (Phyllostomidae).Ecological Monographs, 48, 127–143.

Hendrix, S.D. (1988) Herbivory and its impact on plantreproduction. Plant reproductive ecology: patterns and strat-egies (ed. By J. Lovett-Doust J & L. Lovett-Doust), pp. 246–63. Oxford University Press, Oxford.

Hendrix, S.D. & Trapp, E.J. (1989) Floral biology in Pastinacasativa: do compensatory responses offset reduction in fitness?Evolution, 43, 891–895.

Heringer, E.P., Barroso, G.M., Rizzo, J.A. & Rizzini, C.T.(1977) A flora do Cerrado. IV Simposio sobre o Cerrado (ed.by M.G. Ferri), pp. 211–232. Livraria Itataia Editora andEditora da Universidade de Sao Paulo, Belo Horizonte andSao Paulo.

Hershkovitz, P. (1972) The recent mammals of the Neotropicalregion: a zoo-geographic and ecological review. Evolution,mammals, and southern continents (ed. by A. Keast & B.Glass), pp. 311–41. State University of New York, Albany.



Hertel, F. (1994) Diversity in body size and feeding morphologywithin past and present vulture assemblages. Ecology, 75,1074–84.

Hirsch, H. & Marler, T. (2002) Damage and recovery ofCycas micronesica after Typhoon Paka. Biotropica, 34,598–602.

Hladik, A. (1978) Phenology of leaf production in rain forest ofGabon: distribution and composition of food for folivores.The ecology of tropical folivores (ed. By G. Montgomery),pp. 51–71. Smithsonian Institution Press, Washington, DC.

Holt, R.D. (1977) Predation, apparent competition, and thestructure of prey communities. Theoretical Population Bio-logy, 12, 197–229.

Hoogerwerf, A. (1970) Udjung Kulon. E.J. Brill, Leiden.Howell, T.R. (1971) An ecological study of the birds of the

lowland pine savanna and adjacent rain forest in northeasternNicaragua. Living Bird, 10:185–242.

Hubback, T.R. (1939). The Asiatic two-horned rhinoceros.Journal of the Bombay Natural History Society, 40, 594–617.

Ickes, K., Dewalt, S.J & Appanah, S. (2001) Effects of nativepigs (Sus scrofa) on woody understorey vegetation in aMalaysian lowland rain forest. Journal of Tropical Ecology,17, 191–206.

Janis, C.M. & du Toit, J. (2001) Hoofed mammals. Encyclo-pedia of mammals (ed. by D. Macdonald), pp. 454–65.Barnes & Noble, New York.

Janis, C.M. & Wilhelm, P.B. (1993) Were there mammalianpursuit predators in the Tertiary? Dances with wolvesavatars. Journal of Mammalian Evolution, 1, 103–125.

Janzen, D.H. (1983) Tapirus bairdii (Danto, Danta, Baird’sTapir) Costa Rican natural history (ed. by D.H. Janzen),pp. 496–7. University of Chicago Press, Chicago.

Juenger, T. & Bergelson, J. (1997) Pollen and resourcelimitation of compensation to herbivory in scarlet gilia,Ipomopsis aggregata. Ecology, 78, 1684–95.

Karban, R. & Baldwin, I.T. (1997) Induced responses toherbivory, p. 319. University of Chicago Press, Chicago, IL.

Kay, R.F. & Madden, R.H. (1997) Paleogeography andpaleoecology. Vertebrate paleontology in the Neotropics.The Miocene fauna of La Venta, Colombia (ed. by R.F. Kay,R.H. Madden, R.L. Cifelli and J.J. Flynn) pp. 520–550.Smithsonian Institution Press, Washington, DC.

Kay, R.F. & Simons, E.L. (1980) The ecology of OligoceneAfrican Anthropoidea. International Journal of Primatology,1, 21–37.

Kay, R.F., Madden, R.H., van Schaik, C. & Higdon, D. (1997)Primate species richness is determined by plant productivity:implications for conservation. Proceedings of the NationalAcademy of Sciences, 94, 13023–27.

Keesing, F. (2000) Cryptic consumers and the ecology of anAfrican savanna. Bioscience, 50, 205–15.

Kemp, A. (1995) The Hornbills, Bucerotiformes, p. 302. OxfordUniversity Press, Oxford.

Kingdon, J. (1974) East African mammals: an atlas of evolutionin Africa. Vol. 1. University of Chicago Press, Chicago, IL.

Kingdon, J. (1997) The Kingdon field guide of African mam-mals, p. 464. Academic Press, London.

Kinnaird, M.F., O’Brien, T.G., Suryadi, S. (1996) Populationfluctuations in Sulawesi red-knobbed hornbills: tracking figsin space and time. Auk, 113, 431–40.

Kinnaird, M.F., Sanderson, E.W., O’Brien, T.G., Wibisono,H.T. & Woolmer, G. (2003) Deforestation trends in atropical landscape and implications for endangered largemammals. Conservation Biology, 17, 245–257.

LaDeau, S. & Clark, J.S. (2001) Rising CO2 levels and thefecundity of forest trees. Science, 292, 95–98.

Lambert, F. R. & Marshall, A. G. (1981) Keystone characters ofa bird-dispersed Ficus in a Malaysian lowland rain forest.Journal of Ecology, 79, 793–809.

Laws, R.M. (1970) Elephants as agents of habitat and landscapechange in East Africa. Oikos, 21, 1–15.

Leigh, E.G. & Windsor, D.M. (1982) Forest production andregulation of primary Consumers in Barro Colorado Island.The ecology of a tropical forest: seasonal rhythms and long-term changes (ed. by E.G. Leigh, A.S. Rand, and D.M.Windsor), pp. 111–22. Smithsonian Institution Press, Wash-ington, DC.

Leighton, M. (1982) Fruit resources and patterns of feeding,spacing and grouping among sympatric Bornean hornbills(Bucerotidae). Unpublished PhD Dissertation, University ofCalifornia, Davis, CA.

Leighton, M. (1986) Hornbill social dispersion: variations on amonogamous theme. Ecological aspects of social evolution:birds and mammals (ed. by D.I. Rubinstein & R.W. Wrang-ham), pp. 108–30. Princeton University Press, Princeton, NJ.

Lekagul, B. & McNeely, J.A. (1977) Mammals of Thailand,p. 758. Sahakarnbhat, Bankok.

Leutenegger, W. (1979) Evolution of litter size in primates.American Naturalist, 114, 525–31.

Levin, P.S., Ellis, J., Petrik, R. & Hay, M.E. (2002) Indirecteffects of horses on estuarine communities. ConservationBiology, 16, 1364–1371.

Lindroth, R.L. (1989) Mammalian herbivore-plant interactions.Plant-animal interactions (ed. by W.G. Abrahamson),pp. 163–206. McGraw-Hill, New York.

MacArthur, R.H. & Wilson, E.O. (1963) An equilibrium theoryof insular zoogeography. Evolution, 17, 373–87.

Macedo, R.H.F. (2002) The Avifauna: ecology, biogeography,and behaviour. The Cerrados of Brazil ecology and naturalhistory of a Neotropical Savanna (ed. by P.S. Oliveira &R.J. Marquis), pp. 242–265. Columbia University Press,New York.

McCrossin, M.L., Benefit, B.R., Gitau, S.N., Palmer, A.K. &Blue, K.T. (1998) Fossil evidence for the origin ofterrestriality among Old World higher primates. Primatelocomotion, recent advances (ed. by E. Strasser, J. Fleagle,A. Rosenberger and H. McHenry), pp. 353–396. PlenumPress, New York.

McDonald, H.G. (1997) Xenarthrans: pilosans. The Miocenefauna of La Venta, Colombia (ed. by R.F. Kay, R.H. Madden,R.L. Cifelli and J.J. Flynn) pp. 233–245. SmithsonianInstitution Press, Washington.

MacFadden, B.J. (1995) Pliocene mammals from Inchasi,Bolivia: the endemic fauna just before the Great AmericanInterchange. Bulletin of the Florida Museum of NaturalHistory 39, 87–140.

MacFadden, B.J. (2000) Cenozoic mammalian herbivores fromthe Americas: reconstructing ancient diets and terrestrialcommunities. Annual Review of Ecology and Systematics, 31,33–59.



MacFadden, B.J. & Shockley, B.J. (1997). Ancient feedingecology and niche differentiation of Pleistocene mammalianherbivores from Tarija, Bolivia: morphological and isotopicevidence. Paleobiology, 23, 77–100.

McNaughton, S. & Georgiadis, N. (1986) Ecology of Africangrazing and browsing mammals. Annual Review of Ecologyand Systematics, 17, 39–65.

McShea, W.J., Underwood, H.B., Rappole, J.H., eds. (1997)The science of overabundance: deer ecology and populationmanagement, p. 402. Smithsonian Institution Press, Wash-ington, DC.

Mack, A.L. (1993) The sizes of vertebrate-dispersed fruits: aNeotropical-Paleotropical comparison. American Naturalist,142, 840–56.

Marinho-Filho, J., Rodrigues, F.H.G. & Juarez, K.M. (2002)The Cerrado mammals: diversity, ecology, and naturalhistory. The Cerrados of Brazil: ecology and natural historyof a Neotropical savanna (ed. by P.S. Oiveira and R.J.Marquis), pp. 266–284. Columbia University Press, NewYork.

Marquis, R.J. (1984) Leaf herbivores decrease fitness of atropical plant. Science, 226, 537–39.

Marshall, L.G. (1988) Land mammals and the Great AmericanInterchange. American Scientist, 96, 380–388.

Mauricio, R., Bowers, M.D., Bazzaz, F.A. (1993) Pattern of leafdamage affects fitness of the annual plant Raphanus sativus(Brassicaceae). Ecology, 74, 2066–71.

Michener, C.D. (2000) The bees of the world, p. 913. JohnsHopkins University Press, Baltimore.

Milewski, A.V. & Diamond, R.E. (2000) Why are largeherbivores absent from Australia: a new theory of macro-nutrients. Journal of Biogeography, 27, 957–78.

Mittelbach, G.G., Steiner, C.F., Scheiner, S.M., Gross, K.L.,Reynolds, H.L., Waide, R.B., Willig, M.R., Dodson, S.I.,Gough, L. (2001) What is the observed relationship betweenspecies richness and productivity? Ecology, 82, 2381–96.

Morin, P.J. (1999) Community Ecology, p. 424. OxfordScience, Malden, MA.

Morrison, D.W. (1978) Influence of habitat on the foragingdistance of the fruit bat, Artibeus jamaicensis. Journal ofMammalogy, 59, 622–624.

Morrison, D.W. (1980) Foraging and day-roosting dynamics ofcanopy fruit bats in Panama. Journal of Mammalogy, 61,20–29.

Moulton, M.P. & Pimm, S.L. (1986) The extent of competitionin shaping an introduced avifauna. Community ecology (ed.by J. Diamond & T.J. Case), pp. 80–97. Harper and Row,New York.

Mueller-Dombois, D. (1972) Crown distortion and elephantdistribution in the woody vegetations of Ruhuna NationalPark, Ceylon. Ecology, 53, 208–226.

Niklas, K.L. (1997) The Evolutionary biology of plants, p. 449.University of Chicago Press, Chicago, IL.

Nogueira, M.R. & Peracchi, A.L. (2003) Fig-seed predation by2 species of Chiroderma: discovery of a new feeding strategyin bats. Journal of Mammalogy, 84, 225–233.

Nowak, R.M. (1999) Walker’s mammals of the world, 6th edn,p. 1936. Johns Hopkins University Press, Baltimore.

Oliveira-Filho, A.T. & Ratter, J.A. (2002) Vegetation physiog-nomies and woody flora of the Cerrado biome. The Cerrados

of Brazil: ecology and natural history of a Neotropicalsavanna (ed. by P.S. Oiveira and R.J. Marquis), pp. 91–120.Columbia University Press, New York.

Olmos, F., Galetti, M., Paschoal, M., Mendes, K.L. (1993)Habits of the Southern bamboo rat, Kannabateomys ambly-onyx (Rodentia, Echimyidae) in southeastern Brazil. Mam-malia, 57, 325–35.

Owen-Smith, R.N. (1982) Factors influencingthe consumptionof plant products by large herbivores. Ecology of tropicalsavannas (ed. By G. Huntley & B. Walker), pp. 359–404.Springer-Verlag, New York.

Owen-Smith, R.N. (1988) Megaherbivores: the influence of verylarge body size on ecology, p. 369. Cambridge UniversityPress, Cambridge.

Pacheco, M.A.W. (2001) Effects of flooding and herbivores onvariation in recruitment of palms between habitats. Journal ofEcology, 89, 358–366.

Patton, J.L., da Silva, M.N.F., Malcolm, J.R. (2000) Mammalsof the Rio Jurua and the evolutionary and ecologicaldiversification of Amazonia. Bulletin of the AmericanMuseum of Natural History, 244, 1–306.

Peres, C.A. (1999) The structure of nonvolant mammal com-munities in different Amazonian forest types. Mammals of theNeotropics, Vol. 3 (ed. by J.F. Eisenberg & K.H. Redford),pp. 564–81. University of Chicago Press, Chicago, IL.

Peres, C.A. (2000) Effects of subsistence hunting on vertebratecommunity structure in Amazonian forests. ConservationBiology, 14, 240–253.

Peres, C.A., Janson, C.H. (1999) Species coexistence, distribu-tion, and environmental determinants of Neotropical primaterichness: a community-level zoogeographic analysis. Primatecommunities (ed. by J.G. Fleagle, C.H. Janson, and K. Reed),pp. 55–74. Cambridge University Press, Cambridge.

Peters, R.H. (1983) The ecological implications of body size,p. 329. Cambridge University Press, Cambridge.

Pettet, A. (1977) Seasonal changes in nectar-feeding by birds inZaria, Nigeria. Ibis, 119, 291–308.

Pierson, E.D. & Rainey, W.E. (1992) The biology of flying foxesof the genus Pteropus: a review. Pacific island flying foxes:proceedings of an international conservation conference (ed.by D.E. Wilson & G.L. Graham) US Fish and Wildlife ServiceBiological Report, 90, 1–17.

Poonsward, P. & Tsuji, A. (1994) Range of males of the GreatHornbill Buceros bicornis, Brown Hornbill Ptilolaemustickelli and Wreathed Hornbill Rhyticeros undulates in KhaoYai National Park, Thailand. Ibis, 136, 79–86.

Potts, C.E. & Behrensmeyer, A.K. (1980) Paleoecology ofmiddle Siwalik sediments and faunas, Northern Pakistan.Paleaogeography, Palaeoclimatology, Palaeoecology 30,133–155.

Potts, C.E. & Behrensmeyer, A.K. (1992) Late Cenozoicterrestrial ecosystems. Terrestrial ecosystems through time(ed. by A.K. Behrensmeyer, J.D. Damuth, W.A. DiMichele, R.Potts, H-D. Sues and S.L. Wing), pp. 419–541. University ofChicago Press, Chicago, IL.

Pough, F.H., Andrews, R.M., Cadle, J.E., Crump, M.L.,Savitzky, A.H., Wells, K.D. (2001) Herpetology, p. 612.Prentice Hall, Upper Saddle River, New Jersey.

Power, M.E. & Tilman, D. (1996) Challenges in the quest forkeystones. Bioscience. 46, 609–620.



Prins, H.H.T. & Reitsma, J.M. (1989) Mammalian biomass inan African equatorial rain forest. Journal of Animal Ecology,58, 851–61.

Proctor, M., Yeo, P., Lack, A. (1996) The natural history ofpollination, p. 479. Timber Press, Portland, Oregon.

Putz, F.E. (1984) How trees avoid and shed lianas. Biotropica,16, 19–23.

Rancy, A. (1999) Fossil mammals of the Amazon as a portrait ofa Pleistocene environment. Mammals of the Neotropics: theCentral Neotropics, Vol. 3, Ecuador, Peru, Bolivia, Brazil(ed. by J.F. Eisenberg and K.H. Redford), pp. 20–6. Unversityof Chicago Press, Chicago, IL.

Raven, P.H. & Axelrod, D.I. (1974) Angiosperm biogeographyand past continental movements. Annals of the MissouriBotanical Garden, 61, 539–673.

Remsen, J.V. Jr. (1990) Community ecology of Neotropicalkingfishers. University of California Publications in Zoology,124, 1–116.

Retallack, G.J. (2001) Cenozoic expansion of grasslands andclimatic cooling. Journal of Geology 109, 407–426.

Rich, P.M., Lum, S., Munoz E.L. & Quesada A.M. (1987)Shedding of vines by the palms Welfia georgii and Iriarteagigantea. Principes, 31, 31–40.

Richards, P.W. (1996) The tropical rain forest. 2nd edn, p. 575.Cambridge University Press, Cambridge.

Robbins, R.K. & Opler, P.A. (1997) Butterfly diversity and apreliminary comparison with bird and mammal diversity.Biodiversity II: understanding and protecting our biologicalresources (ed. by D.E. Reaka-Kudla & E.O. Wilson), pp. 69–82. Joseph Henry Press, Washington, DC.

Rodrigues, F.H.G. & Moneiro-Filho, E.L.A. (1999) Feedingbehaviour of the pampas deer: a grazer or a browser? IUCNDeer Specialist Group News 15, 12–13.

Sagers, C.L. & Coley, P.D. (1995) Benefits and costs of defensein a Neotropical shrub. Ecology, 7, 1835–1843.

Salas, L.A. (1996) Habitat of the lowland tapir (Tapirusterrestris L) in the Tabaro River valley, southern Venezuela.Canadian Journal of Zoology, 74, 1452–1458.

Salas, L.A, Fuller T.K. (1996) Diet of the lowland tapir (Tapirusterrestris L) in the Tabaro River valley, southern Venezuela.Canadian Journal of Zoology, 74, 1444–1451.

Schatz, G.E. (1990) Some aspects of pollination biology inCentral American forests. Reproductive ecology of tropicalforest plants (ed. by K.S. Bawa & M. Hadley), pp. 69–84.Parthenon Publishing Group, Paris.

Schenkel, R. & Schenkel-Hulliger, L. (1969) The Javanrhinoceros in Ujung Kulon Nature Preserve. Acta Tropica,26, 98–134.

Schluter, D. (2000) The ecology of adaptive radiations, p. 288.Oxford University Press, Oxford.

Schmitt, C.G. & D.C. Cole (1981) First record of black-leggedseriema (Chunga burmeisteri) in Bolivia. Condor, 83, 182–183.

Short, L. & Horne, J. (2001) Toucans, barbets and honeyguides,Ramphastidae, Capitonidae and Indicatoridae, p. 526.Oxford University Press, Oxford.

Sibley, C.G. & Monroe, B. (1990) Taxonomy and distributionof birds of the world, p. 1111. Yale University Press, NewHaven, CT.

Sick, H. (1993) Birds in Brazil, a natural history, p. 704.Princeton University Press, Princeton, NJ.

Silva, J.M.C. (1995) Birds of the Cerrado region, SouthAmerica. Steenstrupia 21, 69–92.

Simms, E.L. & Rausher, M.D. (1987) Costs and benefits ofplant resistance to herbivory. American Naturalist, 130, 570–81.

Smallie, J.J. & O’Conner, T.G. (2000) Elephant utilization ofColophospermum mopane: posible benefits of hedging. Afri-can Journal of Ecology, 38, 352–259.

Smith, T.B. & Wayne, R.K. (1997) A role for ecotones ingenerating rainforest biodiversity. Science, 276, 1855–1857.

Sterndale, R.A. (1884) Mammals of India. Thacker, Spink, andCompany, Calcutta.

Steven, G.C. (1983) Atta cephalotes (Zompopas, Leaf-cuttingAnts) Costa Rican natural history (ed. by D.H. Janzen),pp. 688–691. University of Chicago Press, Chicago.

Stevenson, P.R. (2001) The relationship between fruit produc-tion and primate abundance in Neotropical communities.Biological Journal of the Linnean Society, 72, 161–78.

Stiles, F.G. (1981) Geographical aspects of bird-flower coevo-lution, with particular reference to Central America. Annalsof the Missouri Botanical Garden, 68, 323–51.

Strauss, S.Y. (1997) Floral characters link herbivores ⁄pollina-tors, and plant fitness. Ecology, 78, 1640–45.

Strauss, S.Y, Conner, J.K, Rush, S. (1996) Foliar herbivoryaffects floral characters and plant attractiveness to pollina-tors: implications for male and female fitness. AmericanNaturalist, 147, 1098–1107.

Strickland, D.L. (1967) Ecology of the rhinoceros in Malaya.Malayan Nature Journal, 20, 1–17.

Sunquist, M. & Sunquist, F. (2002) Wild cats of the world.University of Chicago Press, Chicago, IL.

Suominen, O. (1999) Impact of cervid browsing on theterrestrial gastropod fauna in the boreal forests of Fennos-candia. Ecography, 22, 651–658.

Sussman, R.W. & Kinzey, W.G. (1984) The ecological role ofthe Callitrichidae: a review. American Journal of PhysicalAnthropology, 64, 419–49.

Svenning, J-C. (2001) On the Role of microenvironmentalheterogeneity in the ecology and diversification of Neotropi-cal rain-forest palms (Arecaceae). Botanical Review 67, 1–53.

Talbot, L.M. (1960) A look at threatened species. Oryx, 5, 169–215.

Tangley, L. (1997) In search of Africa’s forgotten forestelephant. Science, 275, 1417–1419.

Terborgh, J. (1983) Five New World primates a study incomparative ecology, p. 260. Princeton University Press,Princeton, NJ.

Terborgh, J. & van Schaik, C.A. (1987) Convergence vs.nonconvergence in primate communities. Organization ofcommunities: past and present (ed. by J.H.R. Gee & P.S.Giller), pp. 205–25. Blackwell Scientific, Oxford.

Thomas, D.W. (1983) The annual migrations of three species ofWest African fruit bats (Chiroptera: Pteropodidae). CanadianJournal of Zoology, 61, 2266–2272.

Toon, A. & Toon, S. (2002) Rhinos, natural history andconservation, p. 72. Voyageur Press, Stillwater, Minnesota.

Tsuji, A., Poonswad, P., Jirawatkari, N. (1987) Application ofradio tracking to study ranging pattern of hornbills (Bucerot-idae) in Thailand. Proceedings of the Jean Delacour/Interna-tional Foundation for the Conservation of Birds Symposium on



Breeding Birds in Captivity. International Foundation for theConservation of Birds, Los Angeles, California, pp. 316–51.

Turkalo, A.K., Fay, J.M. (1995) Studying forest elephants bydirect observations: preliminary results from the Dzangaclearing, Central African Republic. Pachyderm, 20, 45–54.

Turner, I. (2001) The ecology of trees in the tropical forest,p. 298. Cambridge University Press, Cambridge.

Van Der Wal, R., Egas, M., Van Der Veen, A., Bakker, J. (2000)Effects of resource competition and herbivory on plantperformance along a natural productivity gradient. Journalof Ecology, 88, 317–30.

Vanleeuwe, H. & Gautier-Hion, A. (1998) Forest elephantpaths and movements at the Odzala National Park, Congo:the role of clearings and Marantacea forests. African Journalof Ecology, 36, 174–182.

Vaughan, T.A. (1978) Mammalogy, 2nd edn. W.B. Saunders,Philadelphia, PA.

Wagner, T. (2000) Influence of forest type and tree species oncanopy-dwelling beetles in Budongo Forest, Uganda. Biotro-pica, 32, 502–514.

Wallace, A.R. (1852) On the monkeys of the Amazon. Proceed-ings of the Zoological Society of London, 20, 107–10.

Walsh, P.D. & White, L.J.T. (1999) What it will take to monitorforest elephant populations. Conservation Biology, 13, 1194–1202.

Warren, R.J. (1991) Ecological justification for controlling deerpopulations in eastern National Parks. Transactions of theNorth American Wildlife and Resources Conference, 56,56–66.

Warren, P.H. & Gaston, KJ. (1992) Predator-prey ratios: aspecial case of a general pattern? Philosophical Transactionsof the Royal Society of London B, 338, 113–30.

Webb, S.D. (1985) Late Cenozoic mammal dispersals betweenthe Americas. The Great American Biotic Interchange (ed. byF.G. Stehli and S.D. Webb), pp. 357–386. University ofChicago Press, Chicago, IL.

Weidensaul, S. (2000) The raptor almanac, p. 382. Lyons Press,Hong Kong.

Westerkamp, C. (1990) Bird-flowers: hovering versus perchingexploitation. Botanica Acta, 103, 366–71.

White, L.J.T. (1994) Factors affecting the duration of elephantdung piles in rain forest in the Lope Reserve, Gabon. AfricanJournal of Ecology 33, 142–150.

White, J.L. (1997) Locomotor adaptations in Miocene xenarth-rans. The Miocene fauna of La Venta, Colombia (ed. by R.F.Kay, R.H. Madden, R.L. Cifelli and J.J. Flynn) pp. 246–264.Smithsonian Institution Press, Washington, DC.

Whitmore, T.C. (1998) An introduction to tropical rain forests,2nd edn. Oxford University Press, Oxford.

Whitmore, T.C. & Silva, J.N.M. (1990) Brazil rain foresttimbers are mostly very dense. Commonwealth ForestryReview, 69, 87–90.

Wiles, G.J. & Glass, P.O. (1990) Interisland movements of fruitbats (Pteropus mariannus) in the Mariana Islands. AtollResearch Bulletin number 343. Smithsonian Institution Press,Washington DC.

Yap, S.K. & Chan, H.T. (1990) Phenological behaviour insome Shorea species in peninsular Malaysia. Reproductiveecology of tropical forest plants (ed. by K.S. Bawa & M.Hadley), pp. 21–35. UNESCO and Parthenon Publishing,Paris.

Yoshida, A. & Yamazaki, E. (1982) Micro-fossils. Tarijamammal-bearing formation in Bolivia (ed. by F. Takai),pp. 57–62. Research Institute of evolutionary biology,Tokyo.

Zimmer, C. (1998) At the water’s edge: macroevolution and thetransformation of life. Free Press, New York.

Zimmerman, D.A., Turner, D.A., Pearson, D.J. (1999) Birds ofKenya and Northern Tanzania, p. 576. Princeton UniversityPress, Princeton, NJ.

BIOSKETCHES

Cris Cristoffer is a Natural Resources Planner at Luke AirForce Base in Arizona. His primary research interest is indeveloping gestalts of natural communities and how theyfunction, so that their pasts can be revealed and theirfutures preserved. He has a special interest in comparingthe evolution and ecology of certain taxa of organisms,and in giving advice to park managers.

Carlos Peres, Reader in Tropical Ecology at the Centrefor Ecology, Evolution and Conservation, University ofEast Anglia, UK, has had a lifetime interest in the ecologyand conservation of wildlife populations in Neotropicalforests. His studies on forest vertebrate species andassemblages focus on responses to different forms ofanthropogenic disturbance including hunting, forestfragmentation and wildfires.



Elephants versus butterﬂies: the ecological role of large ...

Documents

Transcript of Elephants versus butterﬂies: the ecological role of large ...