Reprint Series1 January 1982, Volume 215, pp. 19-27 S C I E N C E

Neotropical Anachronisms: The Fruits the Gomphotheres Ate

Daniel H. Janzen and Paul S. Martin

Copyright 0 1982 by the American Association for the Advancement of Science

Neotropical Anachronisms:The Fruits the Gomphotheres Ate

Daniel H. .Janzen and Paul S. Martin

New World terrestrial biotas have longcontained a rich fauna of large herbi-vores. During the Pleistocene, untilaround 10,000 years ago, the NorthAmerican mammalian megafauna wascomparable in its number of genera oflarge mammals (those exceeding 40 kilo-grams in adult body weight) to that ofAfrica in historical time (1). Althoughquantitative estimates of prehistoric bio-mass cannot be obtained directly fromthe fossil record, the high carrying ca-pacity for domestic mammals of NewWorld ranges- a capacity similar to thatof African game parks--indicates thatthe Pleistocene biomass of native NewWorld large herbivores was high. Martin(2) estimated an average preextinctionbiomass for unglaciated North Americanorth of Mexico at 21 animal units persquare kilometer or 28.2 x lo6 metrictons on 7.8 x lo6 square kilometers (1unit = 1 cow plus a calf or 1 horse = 449kilograms). While patchily distributed,the megafaunai biomass of lowland Cen-tral America must have been compara-ble, exceeding 50 animal units per squarekilometer on favorable sites.

The number of species of large CentralAmerican Pleistocene herbivores inNeogene deposits of the last 10 millionyears greatly exceeds the number pres-ent in the past 10,000 years. Tapir, deer,peccaries, monkeys, and capybara occuras Pleistocene fossils, but the remains ofgomphotheres (mastodon-like probos-cidians), ground sloths, glyptodonts, ex-

SCIENCE, VOL. 215, 1 JANUARY 1982

tinct equids, Mixotoxodon, Toxodon,and other extinct large herbivorous ani-mals (Table 1) are more common. IfNeotropical ecologists and evolutionarybiologists wish to determine who eatsfruit, who carries sticky seeds, and whobrowses, grazes, tramples, and voids

by reconstructing the interaction be-tween an extant palm and its Pleistocenemegafauna. Without concerning our-selves with what caused the Pleistocenemegafaunal extinctions (3), we are con-sidering a portion of what happenedwhen roughly three-quarters of all thespecies and individuals of large mam-mals were suddenly removed from a drytropical region and its adjacent rain for-ests. The present-day analogy is a tropi-cal, forested African habitat stripped ofits elephants, rhinoceroses, zebras,elands, bush pigs, and other large herbi-vores and left alone for 10,000 years.

We focus on the trees that did not goextinct when their dispersal agents wereremoved. We do this because (i) tree-disperser interactions are not so tightlycoevolved that a reasonable natural his-tory consequence is extinction of oneimmediately following extinction of the

Summary Frugivory by extinct horses, gomphotheres, ground sloths, and otherPleistocene megafauna offers a key to understanding certain plant reproductive traitsin Central American lowland forests. When over 15 genera of Central American largeherbivores became extinct roughly 10,000 years ago, seed dispersal and subsequentdistributions of many plant species were altered. Introduction of horses and cattle mayhave in part restored the local ranges of such trees as jicaro (Crescentia alata) andguanacaste (Enterolobium cyclocarpum) that had large mammals as dispersalagents. Plant distributions in neotropical forest and grassland mixes that aremoderately and patchily browsed by free-ranging livestock may be more like thosebefore megafaunal extinction than were those present at the time of Spanishconquest.

that segment of the habitat that would other; (ii) if there is a large extinct Pleis-have been within reach of a variety of tocene megaflora in tropical America, itmegafaunal trunks, tusks, snouts, has so far escaped detection by paleobot-tongues, and teeth, the missing mega- anists; (iii) the plants that did go extinctfauna must be considered. cannot be directly studied; and (iv) we

There are prominent members of the are confronted with a number of puzzlinglowland forest flora of Costa Rica whose fruit and seed traits whose mystery dis-fruit and seed traits can best be ex- appears when interpreted in the light ofplained by viewing them as anachro- the extinct Pleistocene megafauna. Al-nisms. These traits were molded through though megafaunal extinction resulted inevolutionary interactions with the Pleis-tocene megafauna (and earlier animals)but have not yet extensively responded

Daniel H. Janzen is a professor in the Departmentof Biology, University of Pennsylvania, Philadelphia

to its absence. We first examine this 19104, and Paul S. Martin is a professor in the

evolutionary and ecological hypothesisDepartment of Geosciences, University of Arizona,Tucson 85721.

~34-8075/82/0101-~19$01.00/0 Copyright 0 1981 AAAS

major changes in intrahabitat plant spe-cies composition and population traits,10,000 years is too short a time to expectall the surviving trees to have come to anew evolutionary equilibrium with thesurviving animals and other plants.

A Reconstruction of the Fruiting of

Scheelea 12,000 Years Ago

We shall reconstruct an event from theCosta Rican lowlands about the time aportion of the megafauna vanished. To-ward the end of the dry season in thePacific coastal plain, at a time whennutritious forage is scarce, there is themajor peak in ripe fruit fall from the largeforest palm Scheelea rastrata. In thedense riparian palm groves and uplandmixed forest, the yellow egg-sizeddrupes fall by the thousands. The fruitfall attracts a herd of five gomphotheres

(Cuvieronius), members of the familyGomphotheriidae and more closely relat-ed to the extinct North American mas-todonts (Mammut) than to mammoths(Mammuthus) (4). They forage here dai-ly and consume about 5000 Scheeleafruits per day. The hard nut wall (bonyfruit endocarp) protects the large softseeds from the gomphotheres’ massivemolars and most of the nuts are defecat-ed intact. Below most palms, the groundis picked clean of the fallen fruit. Thepalm groves and individual palms areconnected by well-traveled trails alongwhich small piles of defecated Scheeleanuts are common. Such piles of nuts arealso scattered about in other areas wherethe gomphotheres browse. such as intree-falls, along river banks, and at forestedges.

Nut-rich dung is frequented by agoutis(Dasyprocta punctata) and other smallrodents that remove the nuts. They gnaw

some open and bury others, which aredisinterred when food is scarce. Occa-sionally, when an agouti finds an intactScheelea fruit, it eats the oily sweet pulpand discards the nut. The palm fruits thatescape the gomphotheres and agoutis areeaten by tapirs (Tapirus bairdii) and col-lared peccaries (Tayasu tajacu). Theseanimals chew off the pulp and spit outthe hard nuts. Some Scheelea fruits andnuts are taken by squirrels (Sciurus va-riegatoides) which prey on the seeds.

Insect seed predators (adult bruchidbeetles) oviposit on exposed nuts in thegomphothere dung. The larvae destroyvirtually all the seeds in the nuts left onthe ground surface. By ovipositing onnuts before the rodents get them, theseinsects even kill many of the seeds in thenuts buried by rodents.

The palm population occurs in riparianvegetation, dry hillsides, and woodedpatches in grassland and is largely main-

Fig. 1. Fruits (all to the same scale) in Santa Rosa National Park. Guanacaste Province, Costa Rica, that were probably eaten by Pleistocenemegafauna: (A) Crescentia alata (Bignoniaceae), (B) Enterolobium cyclocarpum (Leguminosae), (C) Sapranthus palanga (Annonaceae), (D)Annona purpurea (Annonaceae), and (E) Acrocomia winifera (Palmae) (19). The white portion of the rule in (B) is 15 centimeters long.

tained by the seed input from the gom-phothere dung. A seedling commonlyappears many kilometers from its parentyet in the vicinity of conspecific adults.There are even adults in habitats whereseedlings have extremely low survivalprobabilities because the gomphotheresgenerate repeated palm recruitment at-tempts in them. Many seeds are killed byseed predators, and most seedlings growfrom seeds that were missed by bothbruchids and agoutis because they weredeeply buried in dung or were carried farfrom the concentrations of seed preda-tors near the parent trees. Also, the

rodents fail to retrieve some of the nutsthey bury. The fruit phenology (that is,the timing of fruit fall within the day andseason), fruit nutrient content, nut shapeand hardness, seed crop size, germina-tion timing, and other reproductive traitsare molded and maintained by complexinteractions in which the gomphothere,with its huge stomach. massive molars,and peripatetic behavior, plays a centralrole.

Then the gomphotheres are gone. Thepalm fruits fall as usual; in a month asmany as 5000 accumulate below eachfruit-bearing Scheelea palm. The first

fruits to fall are picked up by agoutis,peccaries, and other animals that aresoon satiated. As the pulp rots off fallenfruits beneath the parent palm, the bru-chids oviposit on virtually all of theexposed nuts. The bulk of the seedsperish directly below the parent, Even ifthey escape the predators, the seedlingsfrom the undispersed seeds are over-shadowed by an adult conspecific, one ofthe strongest competitors in the habitat.In the next century the distribution ofScheelea begins to shrink. In severalthousand years the local distribution ofScheelea has reached a new equilibrium

Table 1. Missing large herbivores of Central America.

Scientific name

EdentataMegatheridae

Eremotherium(includingMega therium)

Mylodontidae

Megalonychidae

DasypodidaePampatheriumChlamytherium

GlyptodontidaeGlyptodon

RodentiaHydrochoeridae

Neochoerus

CarnivoraUrsidae

ArctodusTremarctos

NotoungulataToxodontidae

Toxodon

LiptoternaMacraucheniidae

MacraucheniaProboscideae

GomphotheriidaeHaplomastodon

CuvieroniusElephantidae

MammuthusPerissodactyla

EquidaeEquus

(Amerhippus)

ArtidactylaTayassuidae

Platygonus

CamelidaePaleolama

BovidaeBison

Common name

Size inanimal units

(1 = 440kilograms)-~

giant ground sloth 8

mylodont groundsloth

megalonychidground sloth

giant armadillo

glyptodont

2 to 4

1 to 2

1 to 2

1 to 2

giant capybara

extinct bear

toxodon

macraucheniops

gomphothere

mammoth

native horse

flat-headed peccary

extinct llama

extinct bison

0.3

1 to 1.5

3

2

5 to 8

Habitat Food

Lowland tropicalforest, savanna

Savanna

Forest

Savanna

Arid lowlandtropics, warmtemperate

Riparian forest

Forest, savanna

Savanna

Savanna

Tropical forest,savanna

Forest, savanna

Savanna, forestedge

Savanna, forestedge

Savanna

Savanna, forestedge

Leafy browse(39)

Grass (42),browse (43,44)

Browse

Omnivore(terrestrial)

Grass (23), fruit,carrion

Riparian andaquatic plants

Meat, fruits,foliage

Grass, lowbrowse

High browse

Fruits, browse

Grass, browse

Grass, browse,fruits

Grass, browse,fruits

Grass, lowbrowse

Grass, lowbrowse

Origin of fossil record

Guatemala (40), Panama (41)

Guatemala (40), Venezuela (45)

Nicaragua (46)

Venezuela (45)Guatemala (40)

Venezuela (45), Guatemala (40)

Venezuela (4.5)

Venezuela (45)

El Salvador (47), Nicaragua (46, 48)

Venezuela (45)

El Salvador (47), Brazil (49)

Costa Rica (50), Venezuela (51)

El Salvador (47)

Central America (23), Guatemala(40), Nicaragua (48), Costa Rica(19a), Venezuela (45)

Mexico (52)

Venezuela (45)

Guatemala (40), Nicaragua (48)

pattern that involves fewer habitat typesand a lower density of adult trees. Thepalm grows only in those microhabitatsso favorable that recruitment occurswith minimal seed disperal and escapefrom seed predators.

Now enter the biologists. assumingthat they are studying a coevolved sys-tem that approximates an evolutionaryequilibrium, They search the morpholog-ical and behavioral features of the exist-ing biota for adaptive meanings. Theystudy Scheelea nut wall thickness andhardness (5), size of fruits and dispersalagents (6, 7), the ratio of one- to two- tothree-seeded nuts (6, 8), the spatial pat-tern of seed predation (9), fruiting phe-nology (5, 9), seed predator satiation (5,6), and the balance between the fruitpulp reward and the seed content rewardto the foraging rodent (10). These inves-tigators notice the huge surplus of fallennuts that remain directly below the par-ent Scheelea and attribute it to contem-porary removal of dispersers by huntersor simply poor adjustment of seed cropsize to the disperser guild. If they wereworking in Africa, however, they would

notice the Scheelea-elephant interaction;in Central America they do not considerthe former scheelea-gomphothere inter-action. The investigators attend only tothe living fauna, although they take careto study native, not introduced, animalsin a seemingly natural habitat.

Researchers have regarded nut wallthickness as an evolutionary adaptiveresponse by Scheelea to the drilling abili-ties of bruchid larvae and the gnawingabilities of rodents. The main selectivepressure determining nut wall thickness,however, could well have been thecrushing force of a gomphothere’s mo-lars, and bruchids and rodents mightsimply have evolved to where they couldpenetrate this defense. The researchersassumed that the reward of fruit pulpshould exceed the work expended by arodent to get at the edible seed minus thevalue of that seed; throughout most ofthe evolutionary history of Scheelea,however, terrestrial rodents may havegotten fruit pulp only rarely. Coevolu-tion of rodents and Scheelea fruits wasassumed; the alternative hypothesis wasnot considered; the rodent is simply

Anacardiaceae

Annonaceae

BignoniaceaeBromeliaceae

EbenaceaeEuphorbiaceaeLeguminosae

Malpighiaceae

Moraceae

Palmae

RhamnaceaeRubiaceae

Sapotaceae

Tiliaceae

Spondias mombinSpondias purpureaSpondias radlkoferiAnnona purpureaAnnona holosericeaAnnona reticulataSapranthus palangaCrescentia alataBromelia karatasBromelia penguinDiospyros nicaraguensisHippomane mancinellaAcacia farnesianaAndira incrmisCaesalpinia coriariaDioclea megacarpaEnterolobium cyclocarpumHymenaea courbarilPithecellobium mangensePithecellobium samanProsopis julifloraBunchosia biocellataByrsonima crassifoliaBrosimum alicastrumChlorophora tinctoriaFicus spp.Acrocomia viniferaBactris guinensisBactris majorZizyphus guatemalensisAlibertia edulisGenipa americanaGuettarda macrospermaRandia echinocarpaManilkara zapotaMastichodendron capiriApeiba tibourbou

jobojobojobosoncoyasoncoyaanonapalancojicaropiiiuelapinuelapersimmonmanzanillohuisachealmendro deldivi diviojo de bueyguanacasteguapinol

monte

cenizeromesquitecerezonanceramonmot-ahigo, figcoyolbiscoyolbiscoyolnaranjillotrompilloguaitil blancomosqueta

nisperotempisquepeine de mico

making use of a food source that wassuddenly plentiful because of Pleisto-cene megafaunal extinction. Biologistsdid not suspect that flowering schedules,plant heights, leaf replacement rates,fruit crop size and phenology, or eventhe genetic structure of a palm popula-tion could now be seriously anachronis-tic if it was evolved to match the habitatsoccupied and type of population distribu-tion pattern that is generated by dispers-al through an extinct wide-ranging largemammal. If the fruiting traits of S. ros-trata are now in major part anachronis-tic, as we suggest, then much of itsinteraction with present-day animalsmay hardly be evolved, to say nothing ofcoevolved (11).

The Megafaunal Dispersal Syndrome

In the lowland deciduous forest ofGuanacaste Province, Costa Rica, thereare at least 39 species of trees or Largeshrubs (Table 2) that are reasonable can-didates for a reconstruction such as thatenvisioned for Scheelea palms and gom-photheres. These trees and shrubs dis-play a set of fruit and seed traits incommon-traits that are puzzling if ex-amined only in the context of the poten-tial native dispersal agents. We viewthese traits as part of the following mega-fauna1 dispersal syndrome.

1) The fruits are large and indehiscent(Fig. 1) and contain sugar-, oil-, or nitro-gen-rich pulp. The seeds they contain areobviously not dispersed abiotically asare the seeds in the large explosiveschizocarp of Hura crepitans (Euphor-biaceae) or the large samara-filled dehis-cent fruit of Swietenia macrophylla (Me-liaceae).

2) The fruits look, feel, and taste likethose eaten by large seed-dispersingmammals in Africa and have seeds andnuts of similar size, hardness, and shapeto those in African fruits that are eatenby large mammals.

3) The large nuts or seeds (Fig. 2) areusually protected by a thick, tough orhard endocarp or seed coat that usuallyallows them to pass intact by the molarsand through the digestive tract wheneaten by introduced large mammals suchas horses, cows, and pigs. Seed scarifi-cation in the animal digestive tract some-times occurs during dispersal. and somescarified seeds are digested.

4) If the seeds are soft or weak, theyare very small (as in figs) or imbedded ina hard core or nut like those in Spondias,Scheelea, and Hippomane. Fruits withsoft seeds may also contain seed-freehard sections in the pulp or core that

SCIENCE, VOL. 21.5

block occlusion of the molar mill, as inthe sweet and woody fruit of Guazumaulmifolia.

5) Different species bear ripe fruits atdifferent times of the year in a givenhabitat.

6) Many of the fruits fall off the treeupon ripening or even well before theyripen; this is best described as behavioralpresentation of fruits to earth-bound dis-persal agents.

7) The fruits usually attract few or noarboreal or winged dispersal agents suchas bats, guans, or spider monkeys. Ifthese animals are attracted, as they areto figs or Spondias fruits, there is usuallya much larger fruit crop than they caneat.

8) In present-day forests, a high pro-portion of a tree’s fruit crop rots in thetree or on the ground beneath it withoutbeing tasted by any potential dispersalagent. This is true even in those nationalparks where sizable wild vertebrate pop-ulations may equal or exceed their pre-Columbian densities.

9) Peccaries, tapirs, agoutis, and smallrodents usually act as seed predators anddispersers of these trees; these animalsdo not act purely as dispersal agents,but at present they are often the onlyones.

10) The fallen fruits are avidly eaten byintroduced horses, pigs, or cattle (or bymore than one). Free-ranging popula-tions of these animals at carrying capaci-ty normally consume all of the fallen fruitin most trees’ crops. At least some of theseeds pass through the digestive tract ofthese animals and eventually germinate.The introduced large herbivores may re-enact many portions of the interactionthe trees had with the extinct mega-fauna.

11) The natural habitats (such as allu-vial bottoms or gentle slopes) of thesetrees are on the edges of grassland and inadjacent forest that are Iikely to be at-tractive to herbivorous megafauna andusually not on steep rocky outcrops andprecipitous slopes.

As we come to know more of thenatural history of the Costa Rican trees,more species will undoubtedly be addedto the list in Table 2. For example, insouthwestern Costa Rica in the lowlandevergreen rain forest of Corcovado Na-tional Park, at least the following havemost or all of the traits listed above:Achmaea magdalenae, Astrocaryumstandleyanurn, Calophyllum macrophyl-lum, Dusia macrophylata, Enallagmu la-tifolia, Elais melanococa, Hymenaeacourbaril, Parkiu pendula, Pouteriaspp., Raphia taedigera, Scheelea ros-trata, Simaba c e d r o n , Terminalia ca-

1 JANUARY I982

Fig. 2. Fruits and their seeds from Santa Rosa National Park that were probably dispersed byPleistocene megafauna. The seeds to the right of each fruit represent a normal quantity of seedsfound in each fruit. (A) Hypenaea courbaril (Leguminosae). (B) Acrocomia vinifera (Palmae).(C) Guazuma ulmifolia (Sterculiaceae). (D) Enterolobium cyclocarpum (Leguminosae). (E)Apeiba tibourbou (Tiliaceae) (19).

tappa, Theobroma sp., Zamia spp. Cou-marouna panamensis nuts come from atree common in many Panamanian andCosta Rican rain forests; the nuts aredispersed by contemporary mammals(12) and were probably dispersed bygomphotheres as well.

Certain species listed in Table 2 haveinstructive exceptions to the traits listedabove. Although Acacia farnesiana hasno sweet flavor or other attractant easilyperceptible to humans in the mesocarp ofits dry, pulpy, and indehiscent fruit, cat-tle and horses seek out and eat the fruits(13), just as do African big game animalswith African Acacia (14). Prosopis juli-flora (mesquite) is especially interestingin this context. In the arid southwesternUnited States, horses and cattle areknown to have aided in the dispersal ofmesquite seeds and the ripe pods ofvarious Prosopis species are often sweetand pleasant tasting to people. In Guana-caste, the ripe pods of P. juliflora areonly slightly sweet and somewhat astrin-gent. Horses and cattle in Guanacasteeat the pods but not as eagerly as dothese animals in northern Mexico, Tex-as, and southern Arizona. Because of thevery local and patchy distribution of P.juliflora in Guanacaste (landward mar-gins of mangrove swamps and high

beach dunes), it has had minimal contactwith livestock.

The relation between habitat and pal-atable fruit production is important. InGuanacaste, the species in Table 2 occuron relatively flat ground on terrain suit-able for large mammal movement. Onsteep rocky slopes in the dry tropicalforest (short-tree forest) of southern So-nora, terrain unsuitable for foraging oflarge mammals, Gentry (15) listed 32prominent woody species, none of whichhave fruits or seeds adapted for largemammal transport. These include Ceibaacuminata, Bursera simaruba, Willardiamexicana, Conzattia sericea, Caesal-pinia platyloba, C. standleyi, Cassiaemarginata, Lysiloma divaricata, L.watsoni, T a b e b u i a palmeri, T . chry-santha, Haematoxylon brasiletto, Jatro-pha platanifolia, J. cot-data, and Ipo-moea arborescens. On the adjacent flood-plains and arroyo bottoms there are spe-cies that have fruits adapted for mega-fauna1 dispersal: Sassafridium macro-phyllum, Vitex mollis, Guazuma ulmifo-lia, Pithecellobium dulce, P. mexica-num, P. undulatum, Prosopis chilensis,and Randia echinocarpa. Thus, in south-ern Sonora, where deciduous tropicalforest reaches its northern limit, at about28”N, the trees with hard seeds and

23

sweet fruits that are palatable to largemammals, including humans, are foundin canyon bottom habitats that wouldhave been the natural corridor for move-ment of the extinct megafauna, just asthey are for introduced livestock.

The diets of the extinct neotropicalherbivorous megafauna. Many largemammals (Table 1), including edentates,gomphotheres, notoungulates, and atleast some equids, were in contact withneotropical and subtropical floras fortens of millions of years, an ample periodfor the evolution of a plant-megafaunadispersal syndrome. On the basis of fieldstudies (13, 14, 16, 17), we assume that,just as contemporary large grazing andbrowsing mammals and some large car-nivores readily consume wild fruits anddefecate the seeds alive, the extinct onesdid as well.

Hypotheses and Tests

Our evolutionary hypothesis can betested by comparing the array of fruitseaten and seeds dispersed by large mam-mals in Africa and Asia with the fruits oftropical America on the one hand andwith the fruits of New Guinea or tropical

Australia on the other; the latter twotropical land masses have never had amammalian fauna that would select for awell-developed megafaunal dispersalsyndrome. We can also test our hypothe-sis by reintroducing Pleistocene mam-mals such as horses (18) to the neotrop-ics and observing their response to thefruits and the response of the plant popu-lations to the mammals. Since the ex-periment has been running for 400 years,a number of the relevant tree populationsmay have already regained populationstructures that are more similar to thoseof the Pleistocene than they are to thoseof recent pre-Columbian times. Never-theless, on a very local scale the oppor-tunity exists for experimentation withtree population structures by the intro-duction of horses, as does the opportuni-ty to study horse responses to detailedfruit and seed traits.

The interaction between Costa Ricanrange horses and jicaro trees (Crescentiaalata) is an example. In Santa Rosa Na-tional Park (19), a horse population thatis usually on an unsupplemented dietranges freely through a portion of themixed deciduous forest and grasslandwhere there are Pleistocene fossil horseremains (19a). The contemporary horses

Fig. 3. Adult Crescen-tia alata with full-sized immature fruitduring the dry season.Naturally fallen ripen-ing fruits are visibleon the ground to theleft of the tree (19).

in Santa Rosa eat substantial amounts offallen fruit of jicaro as well as fruits ofmany other trees listed in Table 2.

In this park, as elsewhere in Mexicoand northern Central America (20), ji-caro grows above small patches of grassin diffuse, nearly monospecific stands.Scattered individuals also occur in theadjacent forest. Reaching a height of 3 to4 meters, jicaro has the spreading andshrubby shape of a savanna tree (Fig. 3).It would not look out of place in NairobiNational Park in Kenya.

The spherical fruits of jicaro (6 to 15centimeters in diameter) contain 200 to800 seeds that are similar in size andshape to broadened cantaloupe seedsand are embedded in a slippery, fibrouspulp. Although the seeds are stiff andsolid, they are more rubbery than hard.Toward the end of the dry season (Marchto May), and again in mid-rainy season(August to September), the still-greenhard fruits fall from the tree. After amonth or more the fruit turns brown andis ripe. There is a very thin layer of sugaron its outer surface at this time. Duringripening, the inner light-colored pulpchanges from one with a flat and slightlyastringent flavor to a slimy black massthat is quite sweet. Despite a penetratingfetid odor, the pulp is quite palatable tohumans (21). In horse-free habitats theindehiscent fruits lie on the ground androt in the rainy season, and fermentationof the fruit pulp kills the seeds. A fallingfruit occasionally cracks open on impact,but one of us (D.H.J.) has not foundseedlings to be produced as a result.When the jicaro tree is in or near forest,an occasional fruit is chewed open bysquirrels. These rodents remove theseeds from the fruit pulp and chew themup. This seed predation results in occa-sional seed dispersal, since the animalmay carry the fruit to a site better pro-tected from predators and drop someseeds along the way or leave some insidethe fruit. The vast majority of jicarofruits are not subject to this treatment.

When range horses are free to foragebelow the trees, they quickly consumethe crop of jicaro fruits. The hard fruitsare broken between the incisors (Fig. 4),an act that requires a pressure of about200 kilograms (22). The gooey pulp isscooped out with the tongue and incisorsand swallowed with little chewing. Formore than ten consecutive days, threecaptive and well-fed range horses ate thefruit pulp of 10 to 15 fruits in each of twomeals a day, one in the morning and onein the evening (22). A herd of 17 rangehorses broke and consumed 666 jicarofruits in one 24-hour period (22). Thepercentage of seeds that survive passage

SCIENCE. VOL. 215

through the gut of a horse is not known,but the dung becomes filled with viablejicaro seeds on the second day after thehorse starts to eat the fruits. About 97percent of these filled seeds germinateafter they are washed out of the horsedung and placed on moist soil or paper.Seeds washed out of the pulp and placedon moist paper also show 97 percentgermination. Sapling jicaro trees arecommonplace in horse pasturing areasinside and outside of Santa Rosa Nation-al Park, provided that the habitats arenot burned annually. Seedling and sap-ling jicaro trees are extremely rare inthose areas of the park where horses donot have access, even in grass and foresthabitats that have dense stands of adultsand are rarely burned.

These observations indicate that Pleis-tocene horses were an important part ofthe disperser coterie of Crescentia alata.Since the Pleistocene horse evolved inthe New World (23), there might even beelements of coevolution in the interac-tion of horses and jicaro fruits.

Today, jicaro and its congener (Cres-centia cujete) are widespread in the drierparts of Central America (20). This dis-tribution is probably the result of boththe immediate pre-Columbian distribu-tion and the post-Columbian spread ofCrescentia by introduced horses. In ad-dition, the hard fruits are used by hu-mans as household tools such as bowls,ladles, and rattles, and the trees aretherefore dispersed in this way too (21).At the time that the domestic horse wasintroduced, C. alata was very likely arelatively rare tree, occurring in smallpatches in relatively open vegetationsuch as marsh edges, along topographicbreaks, and on floodplains, just as it isnow in lowland Costa Rican habitats freeof horses. With essentially no seed dis-persal, the trees were limited to thosesites where populations could survivewith minimal seedling recruitment. Thereturn of horses after 10,000 years result-ed in intensified seed dispersal and hasundoubtedly resulted in the appearanceof more adult jicaro trees in many morekinds of habitats.

The postulated constriction of therange of C. alata after the extinction ofthe Pleistocene horse may affect otheranimals in the habitat. For example, nec-tarivorous bats would be affected by areduction in jicaro density. The flowersof C. alata are nocturnal, abundant, andheavily visited by four species of nectar-ivorous bats in Guanacaste deciduousforests (24), and are the only commonnectar source available to bats in thepark forests during several months of therainy season. The decline of the jicaro

Fig. 4. Range horse breaking a ripe fruit ofCrescentia alata between its incisors (19).

population would have strongly affectedthe population dynamics and structure ofthe many other plant species that arepollinated or dispersed by bats in theCentral American deciduous forest low-lands.

Jicaro fruits are not the only fruitsreadily eaten by introduced horses. Asimilar interaction takes place betweenthe fruits of Enterolobium cyclocarpum(guanacaste) (25), Guazuma ulmifolia(guacimo), and Pithecellobium saman(cenizero) and horses and cattle.

Additional Considerations

Partial loss of dispersal agents. Al-though some frugivores may be littlemore than fruit thieves (26) or depositthe seeds in lethal sites, a tropical treeusually has a complex seed shadow pro-duced by several quite different types ofanimals (12, 27). Extinction of the Pleis-tocene megafauna would eliminate someof a tree’s disperser coterie and therebyexcise part of the tree’s seed shadow.For example, two bat-generated seedshadows (28) of Andira inermis (Table 2)contained many fruits that fell below theparent tree and were passed over bypigs, cattle, and horses, perhaps becauseof antibiotic compounds in the fruit pulp.The seeds in such fallen fruits are killedby the larvae of weevils (28), and thefallen and wasted seeds were viewed bybiologists as a cost of having a sloppyseed disperser and perhaps as due to thetree’s being in an area where the human-disturbed bat populations are lower thanthose to which the fruiting behavior ofthe tree is genetically adjusted. Howev-

er, we suspect that during the Pleisto-cene the fallen fruits would have beenpicked up by foraging gomphotheres,toxodons, and other animals that dis-persed the nut-encased, soft seeds moreeffectively, and perhaps to quite differentplaces.

Bats and other aerial or arboreal verte-brates would generally have taken theirshare of a fruit crop before it was avail-able to the terrestrial megafauna, andtherefore megafaunal extinction shouldhave had little direct effect on them orthe seed shadows that they generate.However, monkeys, squirrels, guans,and curassows, animals that forage forfruit both on the ground and in the treecrown, would have had more opportuni-ty to harvest fruits after the megafaunaextinction. Some increased seed dispers-al by these groups could be expected andthis might have compensated in part forthe loss of the larger dispersers.

Response by seed predators. Verte-brate seed predators such as agoutis,peccaries, and small rodents experi-enced a substantial increase in their foodsupply after the megafaunal extinction.As food availability increased, so shouldtheir populations, habitat coverage, andspecies density.

Arthropod fruit eaters and seed preda-tors were also affected by megafaunalextinction. Three species of Cleogonusweevils feed on the ripening fruit ofAndira inermis, and their larvae developin the fruit pulp and seeds of fallen fruits(28). If fruits were removed from belowAndira trees by large vertebrates, therewould not be the sizable weevil popula-tions that there are at present. The densi-ty of Zabrotes interstitialis bruchids, andthus their intensity of seed predation onseeds of Cassia grandis, is greatly in-creased when the fruits are left on thetrees until they rot (29). When a Pithecel-lobium saman fruit crop falls, its primaryinsect seed predator, Merobruchus co-lumbinus, has just left the fruits (30); wesuspect that the risk of being eaten by alarge mammal (now extinct) accounts forthe insects’ rapid exits. Ripe fruits arerotted by their occupant microbes as away of defending this resource againstlarge herbivores (31); a major selectivepressure for such microbial behavior dis-appeared when the Pleistocene Neotrop-ical megafauna disappeared. Likewise,other associates of large mammals, suchas dung beatles (Scarabaeidae), ticks,horse flies (Tabanidae), cowbirds, andvampire bats, must have been depletedby the loss of the Pleistocene megafauna.

Vegetative defenses against an extinctmegafauna, The extinct tropical Pleisto-cene herbivores consumed substantial

1 JANUARY 1982

Fig. 5 (left). Spines, 7 to 11 centimeters long, on the underside of the petioie of the leaf of saplingAcrocomia vinifera (19). Fig. 6 (right). Desmodium (Leguminosae) beggar’s-ticks stuck tothe forelegs of a free-ranging horse on the edge of the Costa Rican rain forest (38).

amounts of browse as well as fruits andseeds. We expect that some “function-less” but potentially defensive vegeta-tive traits exhibited by trees in modernhabitats are Pleistocene anachronisms.Spininess of African plants developed asa defense against large herbivores (32).There are numerous New World spinyplants in habitats where causal herbi-vores are missing. Spines on palm trunksare probably important in keeping climb-ing rodents from getting at developingfruits (for example of Bactris spp. andAstrocaryum spp.), but the long spineson leaves of Bactris and Acrocomia {Fig.5) cannot be explained this way. Anattempt to explain the spines withoutvisualizing large browsing mammals aspart of the interaction has led to con-struction of a model in search of a realis-tic selective pressure (33). In Santa RosaNational Park and elsewhere in CentralAmerica, prominent spines on the trunksand sometimes leaves of Hura crepitans,Ceiba pentandra (saplings only), Ceibaaesculifolia Acrocomia vinifera, Bom-bacopsis quinatum, Xanthoxylum setu-losum, and Chlorophora tinctoria (sap-lings only) are defenses of trees, espe-cially young trees, against a browsingmegafauna. Although such mechanicaldefenses may be diminishing because ofthe relaxation of selection for them, theyhave not yet disappeared. The recurvedthorns on the twigs and leaves of Mimo-s a guanacastensis, Pithecellobium pla-tylobum, Acacia riparia, A . tenuifolia,and Mimosa eurycarpa could easily havedeterred ground sloths or gomphotheres.The same applies to the needle-sharp tips

26

of the leaves of the understory shrubJacquinia pungens, which is leafy inCosta Rica only during the dry season(34). On well-armed deciduous foresttrees, the spines are commonly best de-veloped within 4 to 6 meters of theground in the neotropics just as they areon African trees. In open vegetation insouthern Sonora, we observed that theshrubby cymbal-spine acacia, Acaciacochliacantha, is extremely thorny.Nearby taller conspecific trees growingin regenerated low forest are almost en-tirely unarmed.

External seed dispersal. Contempo-rary beggar’s-ticks (Desmodium spp.)stick tightly to the hair of domestic horses(Fig. 6). Although they failed to adhereto the sleek coat of an adult captive tapir,or to that of a paca, collared peccary,and white-lipped peccary, experimentsand observations by D.H.J. in SantaRosa National Park show that the burfruits of Pisonia macrunthocarpa, Des-modium spp. , Krameria cuspidata,Triumfetta lappula, Aeschynomene sp.,Petiveria alliacea, and Bidens ripariastick tightly to the denser coats of horsesand cattle. Except for Pisonia and Kra-meria, these plants are herbaceous; theydepend on early colonization of open ornearly open ground for survival. Withthe loss of a megafauna we suspect thatmany of these plants declined severely indensity and some even suffered localextirpation, as the once open and well-trampled habitats were reforested and asseeds were no longer dispersed by largeshaggy beasts such as gomphotheres,toxodons, and ground sloths.

Discussion

In this addition to current evolutionarythought about the equilibrium state ofcontemporary neotropicai habitats, wepropose an answer to the riddle of whycertain trees produce far more ediblefruits than their current dispersal agentswill remove, produce fruits that are noteaten by contemporary dispersal agents,bear fruits that resemble those eaten byAfrican megafauna, and bear fruits avid-ly eaten by introduced livestock. Theseare traits of a megafaunal dispersal syn-drome that has not been evolutionarilyeradicated after the extinction of thedispersal agents 10,000 years ago. Analternative hypothesis is that these treesare not closely coevolved with particularfrugivores and that the system is justvery inefficient, as has been suggestedfor a Panamanian rain forest tree (35).

The fate of fruit crops in African gamepreserves is instructive in consideringthese two hypotheses. Observations byD.H.J. in Uganda and Cameroon forestssuggest that it is indeed a rare eventwhen the intact animal fauna does notconsume all of the fallen fruit crop. Forexample, in a portion of Kibale Forestnear Fort Portal, Uganda, where all theelephants had been killed, the fruits ofBalanites wilsoniana (100 to 150 gramsand 10 to 15 centimeters long) wereabundant and rotting on the ground be-low parent trees. The fruits of B. wilson-iana contain a 40-gram nut and are aboutthe same size and flavor as sapotaceousfruits of the Costa Rican rain forestwhich often lie rotting in large numbersbelow parent trees. Balanites fruits areswallowed by elephants (36, 37) and inthe portions of Kibale Forest where ele-phants were numerous, all the fallenBalanites had been immediately andthoroughly removed by them. In thisportion of Kibale, there are germinatingB. wilsoniana seeds in elephant dungalong forest trails.

Even if our hypothesis were to berejected because it could be shown thatin certain truly pristine neotropical habi-tats the extant animals can fully processthe annual fruit fall, the intriguing matterof the fate of those seed species thatwere dispersed by Pleistocene mammalsis not explained. Even if most populationstructures are now adjusted to the loss ofthe dispersal megafauna, we do not thinkthat this is likely to be the case withevolutionary or coevolutionary equilib-ria. We doubt that those trees with life-spans of 100 to 500 years have experi-enced sufficient generations since thePleistocene to replace the syndrome thatis no longer highly functional. Let us

SCIENCE.VOL.215

assume that the agouti was once a trivialdispersal agent and figured primarily as aseed predator. With the removal of thePleistocene megafauna, the agouti sud-denly has the opportunity for a variety ofevolved and coevolved interactions.However, it may well not have yet ex-ploited the opportunity (11). It may shiftits day-to-day activities in waysendipitously serve the dispersal

that ser-needs of

certain species ofeven though no

tree moderatelyevolution has

well,taken

place in plant or animal.Our discussion has focused on neo-

tropical plants and animals, but it can begeneralized to the sweet-fleshed largefruits of the Kentucky coffee bean Gym-nocladus dioica and honey locust Gledit-sia triacanthos (Leguminosae), osage or-angemina

Maclura (Moraceae), pawpaw Asi-(Annonaceae), and persimmon

Diospyros (Ebenaceae). When there wasa megafauna available to disperse theirseeds, such genera may have been dens-er andtreme

had much wider ranges. The ex-spininess of various New World

extra-tropical shrubs that are found inmoist as well as arid regions has not beenwelland

explained. The vesicatory ripe fruitsweak-walled nuts of Gingko biloba

might even have been evolved in associ-ation with a tough-mouthed herbivorousdinosaur that did not chew its food well.

References and Notes

1. P. S. Martin, Nature (London] 212, 339 (1966).2. Science 179, 969 (1973).3 . En Valen, Proc. N. A m . Paleo. Conf.

Congr. E. (1969), p. 469; L. Medway, AsianPersprct. 20, 51 (1979).

4. B. Kurten and E. Anderson, Pleistocene Mam-

SCIENCE, VOL. 215, i JANUARY 1982

mais of North America (Columbia Univ. Press, 30. -9 Trap. Ecol. 18, 162 (1977).New York. 1980). 31. -,Am: Nat. 111, 691 (1977).

5. D. H. Janzen. Principes 15, 89 (1971); R. A. 32. A. S. Boughey, Ohio J. Sci. 63, 193 (1963); W.Kiltie, thesis, Princeton University (1980); T. T. L. Brown, Ecology 41, 587 (1960); A. I. DaggStruhsaker and L. Leland, Biotropica 9, 124 and J. B. Foster. The Giraffe: Its Biolopv.(1977).

6. L. IX.’ Heaney and R. W. Thorington, J. Mam-mal. 59, 846 (1978).

Behavior and Ecology (Van N&trand Reinh&i;New York, 1976).

7. N. Smythe, Am. Nar. 104, 25 (1970).8. ~~~;~~radford and C. C. Smith, Ecology 58,667

9. D. E. Wilson and D. H. Janzen, ibid. 53, 954(1973).

527 (1942). - -16. D. Y. Alexandre, Terre Vie 32, 47 (1978); B. D.

Burtt, J. Ecol. 17, 351 (1929); R. M. Laws,

10. C. C. Smith, in Coevolution of Animals and

Oikos 21, 1 (1970); D. H. Janzen, Ecology, in

Plants, L. E. Gilbert and P. H. Raven, Eds.(Univ. of Texas Press, Austin, 1975), pp. 53-77.

11. D. H. Janzen, Evolution 34, 611 (1980).12. F. J. Bonaccorso, W. E. Glanz, C. M. Sandford,

Rev. Biol. Trop. 28, 61 (1980).13. M. J. C. Basanez, Biotica 2, 1 (1977); D. H.

Janzen, unpublished observations.14. M. D. Gwynne, East Afr. Wild. J. 7, 176 (1969).15. H. S. Gentry, Carnegie Inst. Washington Publ.

42. M. Salmi, Acta Geog. 14, 314 (1955).43. J. Semper and H. Lagiglia. Rev. Cicnt. Znvesr.

Mus. Hist. Nat. San Rafael (Mendoza) 1, 89

press.17. D. H. Janzen, Oikos, in press.18. H. Ryden, America’s Lost Wild Horses (Dutton,

New York, 1978).19. Santa Rosa National Park, on the Pacific coastal

plain in extreme northwestern Costa Rica, inGuanacaste Province, has a mix of deciduousforest, old abandoned cattle pastures, and ever-green riparian forest (0 to 350 m in elevation).The photographs in Figs. 1 to 5 were takenthere.

gresskeport, 21srsession (1960), part 4”, p. 1.54.46. W. D. Page, in Early Man in America, A. L.

Brian, Ed. (Occasional Paper 1, Department of

(1968).

;3;th;;Tology, Umversity of Alberta, 1968), pp.

47. R. A. <&ton and W. K. Gealey, Geol. Sot. Am.

44. R. M. Hansen, Paleobiology 4, 302 (1978).

Bull. 60, 1731 (1949).

45. J. Rovo v Gomas. International Geologic Con-

33. G. B. Williamson, Principes 20, 116 (1976).34. D. H. Janzen, Biotropica 2, 112 (1970).35. H. F. Howe, Ecology 61, 944 (1980).36. B. D. Burtt. J. Ecol. 17, 351 (1929).37. D. H. Janzen, in Ecology of Arboreal Folivores,

G. G. Montgomery, Ed. (Smithsonian Institu-tion Press, Washington, D.C., 1979), pp. 73-84.

38. The photograph was taken m Corcovado Na-tional Park, a lowland rain forest on the OsaPeninsula, southwestern Costa Rica.

39. C. S. Churcher, Can. J. 2001. 44, 985 (1966).40. M. 0. Woodburne, J. Mammal. 50, 121 (1969).41, C. L. Gazin, Smithson. Inst. Annu. Rep. 4272

(1956), p. 341.

19a.L. D. Gomez (personal communication) reportsthat Pleistocene horse remains are common inthe Costa Rican lowlands, particularly in Guana-caste Province.

20. J. Rzedowski and R. McVaugh, Contrih. Univ.Michigan Herb. 9, 1 (1966).

21. I. C. Mercado, master’s thesis, University ofSan Carlos, Guatemala (1975); J. F. Mortion,Econ. Bot. 22, 273 (1968).

22. D. H. Janzen, unpublished data.23. B. Patterson and R. Pascual, Q. Rev. Biol. 43,

409 (1968).24. D. J. Howell and D, Burch, Rev. Biol. Trop. 21,

281 (1974); E. R. Heithaus, T. H. Fleming, P. A.Onler. Ecolopv 56. 841 (1975).

25. 0: H.’ Janzen: Ecology 62, 587 and 593 (1981);Biotropica 13 (Suppl.), 59 (1981).

26. H. F. Howe and G. A. Vande Kerckhove,Ecology 60, 180 (1979).

27. D. H. Janzen, Annu. Rev. Ecol. Syst. 10, 13(1979).

28. D. H. Janzen et al., Ecology 56, 1068 (1976)29. D. H. Janzen, ibid. 52, 964 (1971).

48. T. R. Howell, Geol. Sot. Am. Spec. Pap. 121.143 (1969).

49. G. G. Simpson and C. P. Coute, Bull. Am. Mus.Nut. Hist. 112, 129 (1957).

50. M. J. Snarskis, H. Gamboa, 0. Fonseca, Vincu-/OS 3. 1 (1977): L. D. Gomez (oersonal communi-cation) reports mastodon remains from the prov-inces of San Jose. Alaiuela. and Guanacaste.

51. A. L. Bryan, R. M. Casamiquela, J. M. Crux-ent. R. Gruhn. C. Ochsenius. Science 200. 1275(1978).

52. A. Mones, An. Inst. Nat. Antropol. Hist. 7, 119(1970-1971).

53. This study was supported by NSF grants DEB77-4889 and 80-l 1558 to D.H.J. and NSF grantsto P.S.M. Servicio de Parques Nacionales deCosta Rica provided habitats, field support, andexperimental animals and plants. W. Freeiandand D. McKey aided in making African observa-tions. R. M. Wetzel and I. Douglas-Hamiltonaided in locating references. The manuscriptwas constructively criticized by J. Brown, W.Hallwachs. D. J. Howell, and G. G. Simpson. Itis dedicated to W. A. Haber who first toldD.H.J. that horses eat Crescentia fruits and toG. Vega who first made D.H.J. aware of thesame fact.

2 7