Journalof Biogeography (1996) 23, 447-458

The ecologyand evolutionof the New Worldnon-pollinating figwasp communities

STUART A. WEST', EDWARD ALLEN HERRE2, DONALD M. WINDSOR2 and PHILIP R. S. GREEN' 'Departmentof Biology,Imperial College at SilwoodPark, Ascot,Berkshire SL5 7PY, U.K and 2SmithsonianTropical Research Institute, Apartado2072, Balboa, Republicof Panama

Abstract. We present data on several previously of pollinatorwasps and viable seeds. Some of the undescribedspecies fromsix genera of New World non- investigatedare parasitoidsof othernon-pollinating species. pollinatingfig . We show that many of thesespecies We examinethe importance of thevarious forms of spatial have a negativeeffect on the reproductivesuccess of both heterogeneityin theparasitism rate that can act to stabilise the pollinatorwasps and the host figs.Our resultssuggest the host-parasitoidinteraction. Finally, we discuss the that the two most abundant genera of non-pollinating factorsunderlying the large variation in theabundance and wasps, the Idarnes and the Critogaster,compete for the diversityof the non-pollinatingwasps both among and same pool of femaleflowers as the pollinatingwasps in the withinfruit crops. Urostigmaand Pharmacosyceafigs, respectively. Wasps from the genusAepocerus induce and develop withinlarge galls, Key words. , , parasites, , in the Urostigmafigs. By drainingresources from the fruit density dependence, spatial heterogeneity,community thesewasps may have a detrimentaleffect on theproduction structure.

success of the pollinatingwasps and also the host figs,by 1. INTRODUCTION reducingthe figs'ability to dispersepollen (West & Herre, The species that are only able to develop withinthe 1994). In contrast,species which merely gall the fruitwall fruitof fig trees are collectivelytermed fig wasps. These or unoccupiedovaries may have less obvious costs to their species include both mutualisticpollinators and parasitic hosts. non-.While the fig trees are completelydependent In thisstudy we describevarious aspects of the upon the pollinatorsfor the dispersal of pollen between of thenon-pollinating wasps associatedwith two subgenera fruit,the non-pollinators provide no apparentservice. Each of monoeciousNew Worldfigs (Ficus, subgenera Urostigma figspecies usually has a singlespecies specificpollinating and Pharmacosycea).We found threecommon genera of wasp species (Ramirez, 1970; Wiebes, 1979; Herre et al., non-pollinatingwasps associatedwith the Urostigmafigs in 1996b). These pollinatingwasps are all membersof the Panama: IdarnesWalker, Aepocerus Mayr and Physothorax chalcidoid family,, and have relativelysimilar Mayr.The Idarnescan be furthersplit into two very different lifecycles. The non-pollinatingwasps also generallyappear groups which we have referredto as Idarnes and Idarnes to be species-specificto a singlefig species (Gordh, 1975; (incerta)(see section2). The majorityof thespecies that we Ulenberg,1985; van Noort, 1991; Boucek, 1993; Machado, have examinedare undescribed.However, molecular work et al., 1996). However,a singlefig species may have several suggeststhat a distinctspecies of each of the four wasp associated non-pollinatingwasp species (e.g. Compton & groups is found in most of the Urostigmafig species Hawkins,1992). These non-pollinatingwasp speciesbelong (Machado et al., 1996; Table 2). While the differentspecies to severalchalcidoid families and show a largerange of life- withineach of these four groups appear to use the same cycles(Boucek, 1988, 1993). resourcesfor larval development,members of the different Verylittle is knownabout the biologyof any of thenon- groups tend to use differentresources (Table 1). All the pollinatingfig wasps. Basic questionsinclude: what are the Idarnes species develop withinfemale , which they resourcesutilized for larval growthby these wasps, and appear to competefor with the pollinatingwasps (West & does theirpresence have any detrimentalcost to theirhost Herre,1994; section3). The Aepocerusand Idarnes(incerta) fig?These questions are closely linked because the effect wasps developwithin much larger galls whichprotrude into that differentspecies have on the reproductivesuccess of the centreof the fruitand appear to arise fromfemale theirhosts will depend upon theirlarval diet. For example, flowersand possiblythe fruit wall (section2.2). The fourth species which are competitors or of the and finalgroup are the Physothorax,which are parasitoids pollinatorswill have a direct effecton the reproductive of the Aepoceruswasps.

( 1996 BlackwellScience Ltd 447 448 StuartA. Westet al.

050 ?

ctc

Big

0- ~~~~~~~~~~~~~~~~~+ -C -~~~~~~~~~~~~~~~cC O $ ct ~ ~ ~ c-c~ -o -o c~~~~~~ z~~~~c

C' ct I, 0

o -~~~~~~~~~~~~~~~~~CdCdd' ~ ~ ~ - ~ 7

0c ;- = = ;! B v=, 3 _ o o o o o > ? o - cd

D 0

5ct 3 c*: 4Sa

bD ~ ~ ~ ~ ~ ~ 0I ct boEi

>d ddC

Q0 *d 0 0 0~~~~~~~~~~~~~~~~~b0b H d C-d

>00 HSm mm

ct ~ ~ ~ ~ ~ lcwl cec t 96,Junlo igorpy 3 4-5 Non-pollinatingfig wasps 449

In the Pharmacosyceafigs that we have studiedthere are 1989;Ware et al., 1993).These foundress wasps subsequently only two common types of non-pollinatingwasps, both die inside the fruit.The proportionof the flowersin each belongingto the genus CritogasterMayr. The two types fruitthat begin to develop is dependentupon the number can easily be distinguishedby the colour of the female's of foundressesthat entered that fruit (Herre, 1989). Fruits body and themorphology of themales (section2). As with thatare not pollinatedare usuallyaborted (e.g. Compton, the Urostigmawasps, a distinctspecies from each of these Ross & Thornton,1994). groupsis generallyfound in each of the Pharmacosyceafig The female flowerswithin the fruitshow continuous species (Machado et al., 1996). Both these types of variationin lengthof styleranging from those with ovaries Critogasterappear to competefor female flowers with the close to the hollow centreof the fruit;that is, close to the pollinatingwasps in a similarway to the Idarnes wasps stigmaticsurfaces (short-styled flowers) to thosewith ovaries (section3.2). close to thewall of thefruit, far from the stigmatic surfaces The firstaim of this paper is to answer fundamental (long-styledflowers). The ovariesof theshort-styled flowers questionsabout thebiology of thesenon-pollinating wasps. are closerto theovipositing foundress wasps. Of theflowers Initiallywe describein detail the naturalhistory of the figs that develop, the long-styledflowers tend, in general,to and wasps studied(section 2). Followingthis, we quantify develop as viable seeds,whereas the seeds developingfrom theeffects that the Idarnes, Critogaster and Aepoceruswasps short-styledflowers tend to be eatenby thewasps' offspring have on viable seed and pollinatorwasp productionand, (Herre, 1989). It should,however, be noted thatvariation therefore,the reproductivesuccess of theirhosts (section in stylelength is gradual and by no means bimodal, and 3). We thenuse theseresults to inferthe larval dietsof the that the characterizationof flowersas 'short' and 'long' wasps (section3). styledis an oversimplificationmade purelyfor convenience Our second aim is to show the ways in which non- (see also Verkerke,1986,1989; Bronstein, 1988a,b; Compton pollinatingfig wasps can be used to examinemore general & Nefdt,1990). biologicalquestions. For example,considerable theoretical Justbefore final ripeningof the fruittakes place, the and empiricalattention has been paid to the factorsthat winglessmales of thepollinating wasps chew theirway out may explainthe persistenceof host-parasitoidinteractions of the flowersin which they have developed. They then (reviewedin Hassell & Godfray,1992; Jones,Hassell & crawl around the interiorof the fruitsearching for flowers May, 1994). Parasitoids such as Physothoraxand their which contain femalewasps. The males chew open these Aepocerushosts may provide useful systems for the empirical flowersand matewith the females. The femalesthen emerge study of potentialstabilizing effects of various formsof fromtheir flowers and gatherpollen, before leaving through parasitoid aggregation at many differentspatial and a hole in the fruitwall chewedby the male wasps. temporalscales. In this studywe examinethe relationship betweenpercentage parasitism and host density,and the 2.2. The non-pollinatingwasps factors that may contributeto the stability of their interaction,using data from the fruitof a single crop Beforeexamining each genus in more detail we will first (section4). As a second example we considerpatterns at establishthe featureswhich are commonto the biologyof the communitylevel. Specifically,we discuss factorsthat all the non-pollinatingwasps thatwe have examinedhere. may influencethe abundance and diversityof the various First,unlike the females, all thespecies considered typesof non-pollinatingwasps across differentfig species in this study oviposit fromoutside the fruitwall (Table (section5). 1). Other studies have shown that some species of non- pollinatingwasps do enterthe fruitto oviposit(see Galil & Eiskikowitch,1969; Murray, 1989; Compton, 1993b). 2. BACKGROUND BIOLOGY Secondly,all the non-pollinatingwasps emergefrom their flowersat approximatelythe same timeas the pollinators 2.1. The figsand their pollinators and exitthe fruit through the hole chewedby thepollinator We sampled naturallyoccurring fig species in the vicinity males.The non-pollinatorwasps thereforedepend upon the of the Panama Canal. The figspecies are grouped in the pollinatingwasps to exitthe fruit. However, we suspectthat subgenus Urostigma,section Americana, and the subgenus the Aepocerusmales may be able to chew an exit hole in Pharmacosycea,section Pharmacosycea. The Urostigmafigs the absence of pollinatormales, as some otherspecies of are pollinatedby wasps belongingto the genusPegoscapus non-pollinatingwasps are able to do (e.g. Compton,Rasplus while the Pharmacosyceafigs are pollinated by wasps & Ware, 1994; Cook & Power, 1996). belongingto the genus . The Idarnes,Aepocerus and Physothoraxwasps are the Individualtrees of all thefig species may produceone to mostcommon non-pollinating genera found emerging from threefruit crops per year (Morrison, 1978; Milton et al., Urostigmafigs in Panama. The Idarneswasps can be split 1982;Windsor et al., 1989). At theinitiation of a fruitcrop, into two morphologicallyvery differentgroups (Boucek, thetree synchronously produces large numbers of receptive 1993; Machado et al., 1996). Those belonging to the fruit. Mated, pollen-bearingfemale pollinating wasps flavicollisand carmespecies-groups have winglessmales and (foundresses)arrive at the tree,enter these fruit, pollinate femaleswith very long ovipositors.We referto thisgroup the receptive,uniovulate flowers, probe the flowerswith as Idarnes.In contrast,the Idarnes belonging to the incerta their ovipositorsand attemptto lay eggs in the ovaries species-grouphave wingedmales and femaleswith relatively (Frank, 1984; Herre, 1989; van Noort, Ware & Compton, shortovipositors. We referto thisgroup as Idarnes(incerta).

? BlackwellScience Ltd 1996,Journal of Biogeography,23, 447-458 450 StuartA. Westet al.

TABLE 2. The non-pollinatingfig wasps associatedwith various speciesof Ficus (subgenusUrostigma). Idarnes represents species from the flavicollisand carmespecies groups, while Idarnes (i) representsspecies from the incertaspecies group. A distinctspecies of wasp fromeach of the genera(or groups)is associatedwith each speciesof figtree (Machado et al., 1996).

Fig Wasp genus species Ficus Idarnes Idarnes (i) Aepocerus Heterandrium Physothorax Doryctinae

bullenei + citrifolia + + + + + + colubrinae + + costaricana + + + dugandi + + + + + + obtusifolia + + + perforata + popenoei + + + trigonata + + + + turbonata + + ?

Occasionallywasps fromother genera are found in fruit, considerablylarger than the pollinatingor Idarneswasps. but due to the small numbersof such wasps that we have The males are wingedand so are able to mate withfemales collectedthey will not be consideredafter this section (but both fromtheir own and fromother fruit, in a similarway see also Tables 1 and 2). Examples includewasps fromthe to the Idarnes (incerta)males. In section 3.2 we consider generaHeterandrium Mayr whichgall thefemale flowers of the impact that the Aepoceruswasps may have on the severalUrostigma species, and some Braconidae:Doryctinae productionof pollinatorwasps and viable seeds. which appear to be parasitoidsof Idarnes (incerta)in E Althoughoccurring in many Urostigmaspecies (Table 2), citrifolia(pers. obs.; identifiedby S. van Noort) (Tables 1 Physothoraxare only common in the fruitof E dugandi and 2). (Table 3). We thereforeonly consider the Physothorax Female Idarneswasps ovipositat the same stage of fruit species associated with E dugandi. Several observations development as the pollinator wasps (Herre, 1989; suggestthat Physothorax are parasitoidsof Aepocerus. First, Bronstein,1991; West & Herre, 1994), and appear to use Physothoraxis placed in the ,a familythought the same attractantcues as thepollinators to findreceptive to consistmainly of parasitoidsof gall-formingfig wasps trees (E.A. Herre, unpublishedresults; Bronstein, 1991). (Boucek, 1993). Secondly,Physothorax and Aepocerusboth These wasps also appear to utilize and compete for the emergefrom the same type of large distinctive gall. Thirdly, a same pool of femaleflowers as thepollinating wasps across G-testshows that the presence of Physothorax and Aepocerus a range of figspecies (section 3; West & Herre, 1994). The were significantlypositively correlated between fruit of F ecologyof the Idarneswasps thereforeshares several basic dugandi (G= 14.43, P<0.001, n=84). We only found featureswith that of the pollinatingwasps (West & Herre, Physothoraxonce in a fruitthat containedno Aepocerus 1994). The femalesare approximatelythe same size as the (this fruitcontained only two Physothoraxindividuals). pollinatingwasps, and have ovipositorsover twice their Finally, Physothoraxfemales have considerablylonger body length.The males are smallerand wingless,possessing ovipositorsthan Aepocerusfemales, suggesting that they mandibleswith distinct teeth which they use in lethalmate lay theireggs in thefruit at a laterstage of fruitdevelopment competitionwith conspecific males (Hamilton, 1979; pers. (see also Godfray,1988; Compton, 1993a,b; Compton et obs.). The effectof Idarneson the productionof pollinator al., 1994). We will considerthe effectsof theseparasitoids wasps and viable seeds are consideredin more detail in on theirhosts in section4. section3.1. Each species of Pharmacosyceafig has two common The Idarnes(incerta) females oviposit at an earlierstage species of Critogasterwasp associated withit: one species of fruitdevelopment than the pollinating wasps. The larvae withgreen females and smalllouse-like males (similarto C. of thesewasps developwithin galls whichprotrude into the singularisMayr); and one specieswith yellow females and centreof thefruit, and arisefrom female flowers. The males largermales (similarto C. fiavescensMayr). Occasionally are wingedand so are able to matewith females inside their greenor yellowwinged males are also found.The females own fruit,and with females from other fruitwho they of both groups have ovipositorsseveral timestheir body encounterin the foliageof the natal tree(pers. obs.). length. Both morphological and molecular work has Female Aepocerusoviposit at a similar stage of fruit suggestedthat while the majority of Critogasterspecies are developmentas the pollinatingwasps, and will oftenlay associatedwith only one figspecies, at leastone Panamanian theireggs in a fruitbefore it has been pollinated.The galls Critogasterspecies is able to developin thefruit of different withinwhich the Aepocerusdevelop are the largestwithin figspecies (Boucek, 1993; Machado et al., 1996). any fruit.These distinctivegalls protrudeinto thecentre of The various Critogasterspecies have very similar life the fruit,and appear to arise fromeither the fruitwall or histories.Females arriveat and startprobing fruit several femaleflowers (see also Bronstein,1991). This large gall days before the pollinatingwasps, and then continueto size is reflectedin thesize of theadult wasps. Aepocerusare oviposit until a few days afterpollination has occurred

? BlackwellScience Ltd 1996,Joumal of Biogeography,23, 447-458 Non-pollinatingfig wasps 451

TABLE 3. Examples of the prevalenceof the differentwasp generain differentcrops (or trees)of various Urostigmafigs. Different crops fromthe same speciesare displayedseparately (e.g. citrifoliaI and citrifoliaII). The figspecies is followedby the numberof fruitsampled fromthat crop (N) and the numberof those fruitthat containedeach wasp genera.The percentagesof fruitcontaining each wasp genera are givenin parentheses.Idarnes represent species from the flavicollis and carmespecies groups,while Idarnes (i) representsspecies from the incertaspecies group. The prevalenceof the differentnon-pollinating wasp generavary enormouslyboth betweendifferent species of Ficus, and betweendifferent crops of the same species. However,the Idarnes are consistentlythe most common typeof non-pollinating wasp.

Wasp genera

Fig species Ficus N (pollinators) Idarnes Idarnes (i) Aepocerus Physothorax

bullenei 34 34 (100%) 4 (12%) 0 0 0 citrifoliaI 20 20 (100%) 0 0 0 0 citrifoliaII 51 51 (100%) 37 (73%) 34 (67%) 0 0 columbrinae 28 28 (100%) 5 (18%) 0 0 0 costaricanaI 10 10 (100%) 1 (10%) 0 0 0 costaricanaII 8 8 (100%) 8 (100%) 0 3 (38%) 0 dugandiI 55 55 (100%) 26 (47%) 12 (22%) 35 (64%) 4 (7%) dugandiII 150 138 (92%) 68 (45%) 0 124 (83%) 60 (40%) obtusifolia 40 40 (100%) 40 (100%) 0 10 (25%) 1 (3%) perforata 56 56 (100%) 1 (2%) 0 0 0 popenoeiI 28 26 (93%) 20 (71%) 0 0 0 popenoeiII 44 44 (100%) 20 (45%) 0 0 0

(D.M. Windsor,unpublished observations). The Critogaster pollinatorwasp and viable seed productionin Urostigma appear to competewith the pollinatorsfor femaleflowers figs,and Critogasterwasps on pollinatorwasp and viable in a similarway to the Idarnes (section 3.3). As the two seed productionin Pharmacosyceafigs. Fruit fromwhich types of Critogasterhave very similarbiologies we have only the male wasps had alreadyemerged from their seeds groupedthem together when examiningtheir effect on the werecollected late in theripening cycle. The fruitwere then productionof pollinatorwasps and viable seeds. cut open and the numberof foundresseswithin each fruit recorded.Each fruitwas subsequentlysealed betweentwo Petridishes, and all the wasps wereallowed to emergeinto 3. THE EFFECT OF NON-POLLINATING the Petridishes beforebeing frozen.Later, the numberof WASPS ON POLLINATOR WASP AND viable seeds and the speciesof each wasp withineach fruit VIABLE SEED PRODUCTION (includingthe wasps that emergedfrom it) were recorded. To understandfully the effectof the non-pollinatingwasps This techniqueallowed thenumber of pollinatorwasps and on viable seed and pollinatorwasp productionwe have to seeds thatdeveloped within each fruitto be relatedto both determinetheir larval diets. More specifically,we need to the numberof foundressesthat enteredthe fruitand the know if the non-pollinatingwasps are developingdirectly numberof non-pollinating wasps thatalso developedwithin at the cost of pollinatorwasps, or viable seeds, or merely the same fruit.As we mentionedearlier, this is important drainingresources from the fig? Theoretically, it is possible because the numberof foundressesthat enter a fruitto to answer this by looking for correlationsbetween non- oviposithas a largeeffect on thenumber of pollinatorsand pollinator presence, pollinator wasp and viable seed viable seeds produced in a fig (Herre, 1989). Exceptions production.However, attempts to determinethe larval diets were E costaricana and E yoponensis,in which only of the non-pollinatingwasps in this way have led to pollinatingand non-pollinatingwasps werecounted. inconclusiveresults (Bronstein, 1991). This may be due to We analysed the data fromeach fig species separately themany confounding factors that influence the production usingthe GLIM statisticalprogram (GLIM 3.77,Numerical of pollinatorwasps and viable seeds. For example, the AlgorithmsGroup, Oxford, 1985; Crawley, 1993). To numberof foundresses(pollen-bearing female pollinating determinethe effectof the non-pollinatingwasps we wasps) that enter a fruitto oviposit and the resources comparedthe viable seed and pollinatorwasp production available to thatfruit have a largeeffect on the numberof across fruitswith variable parasitismrates. In order to pollinatorsand viable seeds producedin a fig(Herre, 1989, controlfor the confoundingfactors that influenceviable 1996; Bronstein,1992; West& Herre, 1994). These factors seed and pollinatorproduction we carriedout an ANCOVA will cause large differencesin viable seed and pollinator withcrop and foundressnumber as factors. wasp productionboth betweendifferent trees and between differentfruits on a treeand should thereforebe controlled 3.1. The effectof Idarneswasps in Urostigmafigs forstatistically. We collected data fromseveral fig species in order to After controllingfor the effectsof variable foundress examine the effectof Idarnes and Aepoceruswasps on numbersand between-cropdifferences there was a significant

? BlackwellScience Ltd 1996,Joumal of Biogeography, 23, 447-458 452 StuartA. West et al.

TABLE 4. ANCOVA on numberof pollinatorsand viable seeds produced in individualfig fruitsfor five species of Ficus (subgenus Urostigma).The crop sampled from,number of foundressesof each fruitsampled, and the numberof Idarneswasps reared out of each fruitsampled were used as covariates.Ficus speciesis followedby numberof crops sampledper species (n), total numberof fruitsampled (N), followedby theproportion of totalvariance in thenumber of pollinatorsor viable seeds explainedby each of thecovariates. *, P<0.05; ***, P<0.01. Also givenare the slopes of the relationshipsbetween Idarnes presence and the numberof pollinatorsor viable seeds when the effectsof crop and foundressnumber are controlledfor. In all cases therewas a significantnegative correlation between the numberof Idarnesreared from a figfruit and the numberof pollinatingwasps. Further,in all cases therewas no significantrelationship between the numberof Idarnesreared from a figfruit and the numberof viable seeds.

Numberof Idarnes Fig species n N Crop foundresses Idarnes slopes

% Total variancein pollinatorproduction explained F bullenei 3 31 12 37*** 12* -2.47 F citrifolia 3 73 1 24*** 14*** -1.24 F costaricana 4 27 14 19 21* -1.09 F obtusifolia 3 63 27*** 1 12*** -1.61 F trigonata 4 74 3 59*** 7*** -0.75 % Total variancein viable seed productionexplained F bullenei 3 31 26* 1 2 0.84 F citrifolia 3 70 18*** 2 2 -0.22 F obtusifolia 3 39 59*** 2 3 -0.85 F trigonata 2 47 8 16 1 0.20

TABLE 5. ANCOVA on numberof pollinator wasps and viableseeds produced in individualfig fruits of Ficuspopenoei (subgenus Urostigma). Fruitwith one and morethan one foundresswere analysed separately. The crop sampledfrom, number of foundressesof each fruitsampled, and the numberof Idarnes wasps reared out of each fruitsampled were used as covariates.The numberof foundressesis followedby numberof crops sampledper species(n), total numberof fruitsampled (N), followedby the proportionof total variancein the numberof viable seeds explainedby each of the covariates.*, P<0.05; ***, P<0.01. Also givenare the slopes of the relationshipsbetween Idarnes presenceand the numberof pollinatorsor viable seeds when the effectsof crop and foundressnumber are controlledfor. In fruitwhich containedonly one foundressthe Idarneswere negativelycorrelated with viable seeds, and showed no correlationwith pollinator wasps. However,in fruitwhich contained more than one foundressthe Idarneswere negativelycorrelated with pollinator wasps, and showed no relationshipwith viable seeds. Therefore,in thesefruit the Idarneshad a similareffect to otherfig species.

Numberof Numberof Idarnes foundresses n N Crop foundresses Idarnes slopes

% Total variancein pollinatorproduction explained 1 4 37 8 1 0.03 >1 4 78 3 36*** 14*** -2.24 % Total variancein viable seed productionexplained 1 4 37 35*** 8* -0.54 >1 4 78 42*** 8 0 -0.12

negativecorrelation between the presenceof Idarnes and the flowersthat are available forthe developmentof wasp pollinatorwasp productionin five species of fig studied offspringwith her own eggs(Herre, 1989). These pollinated (Table 4). In contrast,there was no significantcorrelation flowersthat do not receivea pollinatoror Idarnesegg are betweenthe numberof Idarnesand seed productionin the thereforeable to develop as viable seeds. The numberof fourof the fivespecies in whichseeds werecounted (Table wasps thata fruitis capable of producingdoes not plateau 4). untilat least threeor fourfoundresses have entered(Herre, The importanceof controllingfor confounding variables 1989). Idarnesonly have a negativeeffect on theproduction in these analyseswas well demonstratedby the data from of pollinator wasps in these fruitin which pollinator F popenoei.In E popenoeithe relationshipbetween Idarnes productionhas saturated (Table 5). The majorityof F and both pollinatorwasp and viable seed productionwas popenoeifruit receive more than one foundress(Herre, 1993) dependentupon the numberof foundressesthat entereda and so the generaleffect of Idarnespresence was reduction fruit(Table 5). This resultcan be explainedby considering of the pollinators,as in the otherfig species. the effectthat the resources(eggs and pollen) providedby The importanceof controllingfor confounding variables variablenumbers of foundresseshad on fruitdevelopment. was furtherdemonstrated by thedata fromF trigonata.An If only one foundressentered a fruitthen the Idarnes analysis of this data withoutthe inclusion of crop and were negativelycorrelated with viable seeds and showed foundressnumber as factorsshowed no correlationbetween no relationshipwith pollinatorwasps (Table 5). A single Idarnes presenceand pollinatorwasp production(F1,73 = foundressin E popenoeiis not capable of saturatingall of 2.33, P>0.05). However, as shown in Table 4, when the

? BlackwellScience Ltd 1996,Journal of Biogeography, 23, 447-458 Non-pollinatingfig wasps 453

numberof foundressesentering a fruitand betweencrop the Critogasterare also likely to be competitorsof the differenceswere controlled for there was a significant pollinators,rather than their parasitoids. First, Muller negativecorrelation between the number of Idarnesand the (1887; see Boucek, 1993) has alreadylisted the Critogaster numberof pollinatorsin a fruit.The mostimportant factor species as gall producers in the Ficus subgenus confoundingthe effectsof Idarnesin thiscase was likelyto Pharmacosycea.Secondly, it has been noticedthat females be thatthe productionof pollinatorwasps in E Trigonata will arriveat and probe fruitwith their ovipositors several varies enormouslywith the number of foundressesthat days before the pollinatingwasps (section 2.2). If these entera fruit(Herre, 1989; Table 4). This confoundingeffect femalesare ovipositingbefore the arrivalof the pollinators of foundressnumber was so greatbecause the fruitof E thenthey are unlikelyto be theirparasitoids. trigonataare very large and require several foundresses beforepollinator wasp productionplateaus (Herre, 1989). 3.3. The effectof Aepoceruswasps in an Urostigmafig We have shown that Idarnes wasps have a negative (E dugandi) correlationwith the number of pollinatorwasps developing in a fruit(except for E popenoeifruit which contained only Finally,we considerthe effect of Aepoceruswasps on viable one foundress).This suggeststhat the Idarnes are either seed and pollinator wasp production in E dugandi. competitors,parasites or parasitoids of the pollinating Aepocerus develop in 60% of fruitand are commonly wasps. parasitisedby Physothorax (sections 2.2 and 4). To calculate Additionalobservations strongly suggest that Idarnes are the numberof Aepocerusthat were laid in each fruitwe not directlyparasites or parasitoidsof thepollinators (West summedthe numberof both Aepocerusand Physothorax & Herre,1994). Aepoceruscan preventfruit which received wasps thatemerged. no pollinatorwasps frombeing aborted (section 3.3). We We examinedthe fruitcrop of a singletree and found observedIdarnes develop in these unpollinatedfruit in E thatfruit which did not containAepocerus produced more popenoei,E dugandiand F trigonatain Panama. This has viable seeds (F1,30=1.53, P>0.05) and significantlymore also been observedto occur in E pertusa in Costa Rica pollinatorwasps (F1,31= 12.65,P<0.01) thanfruit which did (Bronstein,1991). Aepocerusoccur in muchlower numbers containAepocerus (Table 7). A largersample size would be per fruitand in verydifferent galls than the Idarnes(Table likelyto showthat the presence of Aepocerus has a significant 3; section2.2). Therefore,the Idarnes are not parasitoidsof negativeeffect on theproduction of viable seeds. Therewas theseother non-pollinating wasps. no significanteffect of increasednumbers of Aepocerusin These results,coupled with the fact that the Idarnes a fruiton pollinatorwasp (F131=1.97, P>0.05) or viable wasps usuallyemerge from the same layerof flowersas the seed (F1,30=0.002,P>0.05) production. pollinatingwasps (Herre, 1989), suggestthat the Idarnes Aepocerusdrain resourcesfrom the fig tree in at least utilize and compete for the same pool of flowersas the two ways. First,Aepocerus can preventabortion of both pollinatingwasps. This idea is furthersupported by two unpollinatedfruit, and fruitwhich were developinga few observationssuggesting that Idarnes reproductionmay pollinators(pers. obs.) (see also Galil & Eisikowitch,1968; actuallybe greaterin the absence of the pollinatingwasps. Bronstein,1991; Compton, 1993b; Cook & Power, 1995). First, the Idarnes wasps occur in significantlygreater These fruitwould normallybe aborted so that the tree numbers in unpollinated fruit that were not aborted could divertthe resourcesto otherfruit, where they could (Bronstein,1991; West & Herre,1994). Secondly,the Idarnes possiblybe used more profitably(Herre, 1989, 1996; West wasps of E popenoei appear to develop preferentiallyin & Herre,1994). This is likelyto be whyfruit which did not flowersthat do not contain pollinators in fruitwhere contain Aepoceruscontained more pollinatorwasps than pollinatorproduction has notsaturated (Table 5 and above). fruitwhich did containAepocerus. Secondly, the Aepocerus wasps will drain resources from the fig for their own development. 3.2. The effectof Critogasterwasps in Pharmacosycea figs 4. POPULATION DYNAMICS, PHYSOTHORAX We examinedthe effectof Critogasterwasps on pollinator PARASITOIDS AND THEIR AEPOCERUS wasp and viable seed production for three species of HOSTS Pharmacosyceafigs. Critogaster exhibit similar patterns to theIdarnes. After controlling for the number of foundresses A major questionin the dynamicsof any host-parasitoid that entered a fruit,and between crop differencesthe systemis what factorspromote the stabilityof the inter- presence of Critogasterand pollinator wasp production acting populations (Pacala, Hassell & May, 1990)? A werenegatively correlated in all threespecies of figstudied host-parasitoidinteraction is termedstable if the host and (Table 6). In contrast,there was no significantcorrelation populations densities remain roughly steady, and both between the number of Critogasterwasps and seed speciesare able to co-existover ecologicaltime. Parasitoid productionin the two speciesin whichseeds werecounted aggregationat certainpatches has been suggestedas an (Table 6). importantfactor in stabilizinghost-parasitoid interactions As withthe Idarnes this negative effect of the Critogaster (Pacala et al., 1990). on the productionof pollinatorwasps suggeststhat the Parasitoidaggregation may occur fortwo reasons.First, Critogasterare eithercompetitors, parasites or parasitoids parasitoidsmay aggregatein patchesof highhost density; of the pollinatingwasps. Two observationssuggest that and secondly,parasitoids may aggregatein certainpatches

? BlackwellScience Ltd 1996,Joumal of Biogeography, 23, 447-458 454 StuartA. Westet al.

TABLE 6. ANCOVA on numberof pollinatorsand viable seeds produced in individualfig fruitsfor threespecies of Ficus (subgenus Pharmacosycea).The crop sampledfrom, number of foundressesof each fruitsampled, and the numberof Critogasterwasps rearedout of each fruitsampled were used as covariates.Ficus species is followedby numberof crops sampled per species (n), total numberof fruit sampled (N), followedby the proportionof total variancein the numberof pollinatorsor viable seeds explainedby each of the covariates. *, P<0.05; ***, P<0.01. Also givenare the slopes of the relationshipsbetween Critogaster presence and the numberof pollinatorsor viable seeds whenthe effectsof crop and foundressnumber are controlledfor. In all cases therewas a significantnegative correlation between the numberof Critogasterreared from a figfruit and thenumber of pollinatingwasps. Further,in all cases therewas no significantrelationship betweenthe numberof Critogasterreared from a figfruit and the numberof viable seeds.

Numberof Critogaster Fig species n N Crop foundresses Critogaster slopes

% Total variancein pollinatorwasp productionexplained ? glabrata 1 20 63*** 11* -3.31 F insipida 21 580 36*** 6*** 1* -0.35 F yoponensis 1 22 70*** 13*** -1.01 % Total variancein viable seed productionexplained F glabrata 1 20 9 5 -0.19 F insipida 21 580 63*** 1 0 -0.12

TABLE 7. The averagenumber of pollinatorwasps and viable of the Physothoraxparasitoid on its Aepocerushost in E seedsin thirty-threefruit from a singlecrop of E dugandi.Fruit dugandi.Data from the fruitcrop of a single tree was are dividedbetween those in whichAepocerus where absent or analysedin two ways. First,we examinedif the parasitism present.Standard errors are givenin parenthesis.Fruit in which Aepoceruswere present contained fewer pollinator wasps and viable ratewas significantlycorrelated to the densityof hostsin a seeds(see section3.3). fruit(i.e. is there HDD?). Secondly,we determinedthe relativepossible importance of HDD and HDI in stabilizing Pollinatorwasps Viableseeds the host-parasitoidinteraction. The values of HDD and HDI were obtainedusing a Turbo Pascal programwritten Aepocerusabsent 154.7(24.0) 74.0(20.0) by Pacala. Aepoceruspresent 58.5 (12.5) 46.5 (7.6) Analysisof proportiondata, such as theparasitism rate, resultsin non-normalerror distribution, non-constancy of thevariance and loss ofinformation of the size ofthe sample irrespectiveof host density.These two typesof parasitoid fromwhich each proportionwas calculated(Crawley, 1993). aggregationhave been termed host density dependent To avoid theseproblems we analysedthe data witha general (HDD) and host densityindependent (HDI) (Pacala et al., linearmodel analysis of deviance, assuming binomial errors, 1990). in the GLIM statisticalpackage (GLIM 3.77, Numerical Parasitoid aggregationis stabilizingbecause it leads to Algorithms Group, Oxford, 1985). The number of competitionbetween individual parasitoids for hosts and a Physothoraxin a fruitwas used as the responsevariable reductionin the average parasitoid searchingefficiency. and the total numberof Aepocerusand Physothoraxin a The reductionin theaverage parasitoid searching efficiency fruitas thebinomial denominator. The data werefound to caused by aggregationincreases as the average parasitoid be overdispersed,which may resultin overestimationof densityincreases. This reductionin searchingefficiency significancelevels (Crawley,1993). To account for thiswe at high parasitoid densitiesintroduces temporal density rescaledthe deviance by an appropriateheterogeneity factor dependenceinto the systemwhich can potentiallystabilise (the ratio of Pearson's X2to the degreesof freedom).The thehost-parasitoid interaction (Free, Beddington & Lawton, heterogeneityfactor (HF) is givenin the text.Hypothesis 1977; Hassell & Pacala, 1990). testingwas carriedout using X2approximations. Pacala et al. (1990; Hassell & Pacala, 1990; Pacala & Physothoraxparasitized 21.0% of Aepocerus.There was Hassell, 1991; Hassell et al., 1991) have recentlydeveloped a positive correlationbetween the proportion of galls a usefulframework within which the relativeimportance parasitized by Physothoraxand the number of galls in of HDD and HDI can be determinedfrom the spatial a fruit(X2 =13.98, P<0.001, HF=2.66, n=41). Such a heterogeneityin the parasitismrate of fieldsamples. This relationshipis termed positive density dependence. A involves the calculation of the coefficientof variation possible explanation for this patternis that parasitoids squared (CV2 = variance/mean2)of the distributionof aggregated at fruitcontaining high densities of hosts. searching parasitoids weighted for hosts within any Alternativelythe parasitoidsand theirhosts may respond particularpatch. This variation can then be partitioned to the same figcues. into the HDD and the HDI components.The relationship Host densitydependent variation in the parasitismrate/ between these differentcomponents is given by: CV2= aggregation(HDD = 2.55) was greaterthan host density HDD*HDI -1. Pacala et al. (1990) developeda simplerule independentvariation in the parasitismrate/aggregation that if CV2>1 thenthe host-parasitoidinteraction will be (HDI= 1.41). The total CV2 for thisdata set (CV2=2.59) stable (but see below). exceeded unity,so the heterogeneityin parasitismhas the We examinedthe spatial variationin the parasitismrate potentialto stabilize the interactingpopulations. In fact

? BlackwellScience Ltd 1996,Journal of Biogeography, 23, 447-458 Non-pollinatingfig wasps 455

with HDD>2, there is sufficienthost densitydependent variationin wasp species diversityacross the differentfig heterogeneityfor this to potentiallystabilize the interaction species. on its own (Pacala & Hassell, 1991). The Idarnes are consistentlythe most abundanttype of Care should, however,be taken in makingconclusions non-pollinatingfig wasp in figsof the subgenusUrostigma fromthis analysis. First, we have examineddata fromonly throughoutthe Central and South America (E.A. Herre, a single fruitcrop. Second, the CV2 frameworkmakes unpublishedresults; Gordh, 1975; Bronstein,1991; West& fundamental assumptions about the biology of the Herre, 1994). In Panama, the Idarnes occur in a higher interactionsbetween host and parasitoid(Hassell & Pacala, proportionof fruit(Table 3) and in highernumbers per 1990; Taylor,1993). The assumptionswere summarizedby fruitthan any othergroup of non-pollinatingwasps. This Hassell & Pacala (1990) as: (1) the parasitoid species in success of Idarnes wasps in exploitingthe fig pollinator question are specialistson one host species; (2) the host mutualismmay be explainedby the fact that these wasps and parasitoidshave discretegenerations; (3) parasitoids appear to be utilizingthe same pool of flowersthat the encountertheir hosts at random; (4) the host-parasitoid pollinatorsuse (section3). The Idarneswasps exploitflowers interactionis of predominantimportance to the overall in whichthe pollinators grow and in whichthe fig apparently dynamics;factors such as competitionbetween hosts and cannot differentiatebetween a non-pollinator and a parasitoid interferenceare neglected. In addition, the pollinator.In addition,Idarnes wasps arriveat a receptive underlyingmodel assumes no parasitoid redistribution tree at the same time as the pollinators(Bronstein, 1991; withinthe season, somethingthat has more recentlybeen West& Herre,1994), and appear to use the same attractant shownto be importantin modulatingthe stabilizingpower cues as the pollinatorsto findreceptive trees (Bronstein, of HDD aggregation(Rohani, Godfray& Hassell, 1994). 1991; E.A. Herre,unpublished results). Therefore, a figtree It would be extremelyuseful to explorehow thelevel and cannot attractits pollinatorwasps withoutalso attracting formof heterogeneityin the parasitismrate changes with its Idarnes non-pollinatingwasps. It would be extremely spatial scale and over time. Variationat differentspatial instructiveto know if the abundance of other genera of scales and over time may have importantconsequences non-pollinatingwasps, such as Critogasterand Philotrypesis, for the population dynamicsof a host-parasitoidsystem can be similarlyexplained. Our results(section 3) suggest (Hassell & Pacala, 1990; Taylor,1993). However,long-term that at least withCritogaster this may be the case. fielddata documentingsuch patternsare scarce (but see Wasps such as Aepocerusare able to short-circuitthe fig's Hails & Crawley,1992; Redfern,Jones & Hassell, 1992; abilityto abort unpollinatedfruit, and are thereforeable Jones,Hassell & Pacala, 1993). Fig wasps may provide a to existindependent of the pollinator(see also Compton, usefulsystem for detailed field studies designed to examine 1993b). We mighttherefore expect these wasps to be far suchquestions because theycan easilybe sampledat several more prolificthan theyare (Table 3). However,a possible naturalspatial scales (e.g. fruit,branch, tree). factor limitingthe abundance of these wasps is their The observed pattern of parasitism may also have parasitoids.Theoretical studies, laboratory experiments and importantimplications for the behaviour of thehost wasps. biological control programmeshave demonstratedthat An individual female may be selected to change the parasitoidsmay play an importantrole in theregulation of distributionof her progenyacross patches (in this case theirhosts (Hassell & Godfray,1992). In addition,modern fruit)as a responseto parasitoidattack (Thompson, 1986a; methods of analysing detailed field data have made it Godfray, 1987, 1994). For example, positive density possible to estimate the degree to which a particular dependentrates of parasitismmay favour hosts that lay parasitoiddepresses its host populationlevel (e.g. Joneset smaller numbersof eggs in each patch. This individual al., 1993). Althougha moredetailed field study is required, behaviourwould then be reflectedin the distributionand the data presented in section 4 are suggestive that structureof the population.Population structurehas been Physothoraxmay play an importantpart in depressingthe shownto be themajor determinant of the sex ratios and male populationsof theirAepocerus hosts. matingstrategies for both pollinating and non-pollinatingfig More generally,the abundance of any of the non- wasps (Hamilton, 1979; Frank, 1985; Herre, 1985, 1987, pollinatingwasp species could also be influencedby the 1993; Herre et al., 1996a; Murray, 1987, 1989). fruitingphenology of theirhost trees (see also Cook & Consequently,by influencingtheir host's distribution, Power, 1996). For example, if a tree is sufficiently parasitoidssuch as Physothoraxmay have played a part in asynchronousin itsfruiting phenology then non-pollinating determiningtheir host's reproductivestrategies. wasps would be able to findfruit suitable for oviposition on the treefrom which they have emerged.In thiscase we mightexpect these wasps to becomemore common as they 5. COMMUNITY STRUCTURE go throughseveral generations on a singletree. This may The abundanceand diversityof non-pollinating wasps varies explain why a sample froman asynchronouscrop of a E enormouslyat many differentlevels: betweendifferent fig costaricanatree (Table 3: costaricanaI) had a much lower species(Herre, 1989; Compton& Hawkins, 1992; Tables 2 prevalenceof Idarnesand Aepoceruswasps than a sample and 3), betweensamples of the same figspecies at a single taken several weeks later from the same crop (Table 3: location (Table 3) and betweendifferent fruit on the same costaricanaII). It would be relativelyeasy to testthis idea tree.In thissection we willfirst discuss some of the factors furtherby exploitingthe factthat there exists geographical that are likelyto be controllingthe relativeabundances of variationin thefruiting phenology of certainfig species (C. the differenttypes of wasps. Then we will go on to discuss Smith,this issue).

? BlackwellScience Ltd 1996,Joumal of Biogeography, 23, 447-458 456 StuartA. Westet al.

4 we demonstratedthat non-pollinating fig wasps may have a large detrimentaleffect on the reproductivesuccess of both the figsand theirpollinating wasps. Specifically,we have shownthat both Idarnesand Critogasterhave a direct c)0~~~~ ~ U negativeeffect on pollinatorwasp production,which in turn effectsthe figs'ability to dispersepollen (sections3.1 and .~_I ~ p 3.3). In contrast,Aepocerus are likely to have only an o indirecteffect on pollinatorwasp and viableseed production, by drainingresources from theirhost tree (section 3.2). Nonetheless,these results emphasize the importanceof consideringnon-pollinating wasps in studiesthat investigate ,. 0 1 2 3 4 5 6 7 pollinatorwasp and viableseed production (see also Pellmyr, z Numberof gallingspecies 1995). Giventhat non-pollinating wasps have a largedetrimental FIG. 1. The relationshipbetween galling wasp and parasitoid effecton thereproductive success of theirhost figs, they are speciesdiversity for the non-pollinatingfig wasps of ten speciesof likelyto have been an importantselective pressure shaping Urostigmafigs. The square pointsrepresent two species.Data from many aspects of fig biology. Indeed, it has been only two samples (fortyfruit in each sample) of each figspecies were includedin this analysisto controlfor the large effectthat demonstratedthat the tendinghomopterans on E sur sample size has on species diversity(Compton & Hawkins, 1992; benefitthe treeby reducingthe numberof non-pollinating Memmot, Godfray& Gauld, 1994). The diversityof parasitoid wasps developingin fruit(Compton & Robertson,1988). speciesis positivelycorrelated with the diversityof gallingspecies Future work could examine the possibilitythat non- (X2 = 7.70,p<0.0O1, n-10I). pollinatingwasps may have influencedthe evolution of factorssuch as fruitmorphology, fruiting phenology and We now turn to the variationin the diversityof non- even dioecy (see also Frank, 1989; Patel, Hossaert-McKey pollinatingfig wasps across the differentfig species. This & McKey, 1993; Compton et al., 1994). topic has recentlybeen addressedusing the South African It is perhapspuzzling that figs have not merelyevolved fig species (Compton & Hawkins, 1992; Compton & van thickor toughfruit walls which could excludethe ovipositors Noort, 1992; Hawkins & Compton, 1992; Compton et al., of these'evil' non-pollinatingwasps. A possiblesolution to 1994). Compton & Hawkins (1992) consideredthe factors thisproblem may be thatthe evolutionof such a fruitwall that influencethe diversityof the gall-formingwasps (e.g. wouldconflict with other important roles the fruit wall must Idarnes or Aepocerus) and their parasitoids (e.g. play. For example,the fruitwall mustbe able to allow the Physothorax).They have shownthat, across figspecies, the transpirationof enoughwater to keep the figcool enough diversityof SouthernAfrican galling wasps is positively forthe pollinator larvae to survive(Patino, Herre & Tyree, correlatedwith tree height and wasp taxon size (Compton 1994). Alternatively,a fruit that is toughenough to exclude & Hawkins, 1992). They suggest that these results non-pollinatorsovipositing from the outside may not be demonstrate the importance of both phylogenetic particularlyattractive to frugivores. constraintsand ecological factorsin determiningspecies We have also shown that the parasitoidsmay play a diversityof the gallingwasps. Phylogeneticconstraints are significantpart in reducingthe populationsof theirgalling likelyto play a part in the evolutionof non-pollinating host wasps (section4). Justas non-pollinatingwasps may wasp communitiesbecause the high host-specificityand have been an importantselective pressure shaping many probable cospeciation between figs and their wasps aspects of figbiology, these parasitoids are likelyto have (Thompson, 1986b; Herre et al., 1996b; Machado et al., affectedmany aspects of their host's biology (Godfray, 1996) is likely to limit the number of galling species 1994). For example,we have discussedhow parasitoidsmay associatedwith any figspecies (Compton & Hawkins,1992; have influencedthe evolution of their hosts' population Hawkins& Compton,1992). Tree heightmay be important structure,and thereforetheir reproductive strategies (section because largertrees will produce more fruitand may be 4). These observationsadd to the many ways in which easier for non-pollinatingwasps to find (Compton & parasiteshave been implicatedto influencethe biologyof Hawkins, 1992). Compton & Hawkins (1992) also showed theirhosts (e.g. Price, 1980; Hamilton, 1980; Hamilton & that the diversityof parasitoid species was strongly Zuk, 1982). correlatedwith galling wasp diversityacross the African figspecies. We have found the same relationshipbetween ACKNOWLEDGMENTS parasitoidand gallerdiversity across the Urostigmafigs from Panama (Fig. 1). These resultssuggest that the variability of We thankMaritza Lopez, Monica Meija, PatriciaEscobar, available hosts is the most importantfactor determining Sara Dent and Mir Rodriguezfor usefuland enthusiastic parasitoiddiversity (Compton & Hawkins, 1992). assistancein the field.Jean-Yve Rasplus and Simon van Noort kindlyhelped identifythe wasp species. Jennifer McCabe, JamesCook, Charles Godfray,Egbert Leigh Jr, 6. CONCLUSIONS HefinJones, Steve Compton, Mike Bonsall, Dieter Ebbert, In this study we have describedvarious aspects of the Dan Racey and Pej Rohani generouslyprovided discussion ecology of severalnon-pollinating species. First, and commentson the manuscript.S.A.W. also thanks

? BlackwellScience Ltd 1996,Joumal of Biogeography, 23, 447-458 Non-pollinatingfig wasps 457

E.A.H. for providingthe financialsupport to attend the Galil, J. & Eisikowitch,D. (1968) On the pollinationecology of Ficus symposium in Bergen. Finally, we thank Steve Ficus sycomorusin East Africa.Ecology, 49, 259-269. Compton forinviting this contribution to the proceedings Galil, J.& Eisikowitch,D. (1969) Furtherstudies on thepollination ofthe Ficus symposium. This workwas supportedby NERC ecologyof Ficus sycomorusin East Africa.Tijd. Ent. 112, 1-13. Godfray,H.C.J. (1987) The evolutionof clutch size in invertebrates. studentships(S.A.W. and P.R.S.G.) and the Smithsonian Oxf Surv.Evol. Biol. 4, 117-154. Tropical ResearchInstitute (E.A.H. & D.M.W.). Godfray,H.C.J. (1988) Virginityin haplodiploid populations: a studyon figwasps. Ecol. Entomol.13, 283-291. Godfray,H.C.J. (1994) Parasitoids.behavioural and evolutionary REFERENCES ecology.Princeton University Press, Princeton, NJ. Gordh, G. (1975) The comparativeexternal morphologyand Boucek, Z. (1988) 6. Family Agaonidae. AustralianChalcidoidea systematicsof the neotropicalparasitic fig wasp genus Idarnes (),A biosystematicrevision of genera of fourteen (Hymenoptera:Torymidae). U. Kansas Sci. Bull. 50, 389-455. families,with a reclassificationof species, pp. 156-209. CAB Hails, R.S. & Crawley,M.J. (1992) Spatial densitydependence in International,Wallingford, UK. populationsof a cynipidgall formerAndricus quercuscalicis. J. Boucek, Z. (1993) The generaof chalcidoidwasps fromFicus fruit Anim.Ecol. 61, 567-584. in the new world.J. Nat. Hist. 27, 173-217. Hamilton,W.D. (1979) Winglessand fightingmales in figwasps Bronstein,J.L. (1988a) Fruit productionin a monoecious fig: and otherinsects. Sexual selectionand reproductivecompetition consequencesof an obligatemutualism. Ecology, 69, 207-214. in (ed. by M.S. Blum and N.A. Blum), pp. 167-220. Bronstein, J.L. (1988b) , antagonism, and the AcademicPress, London. fig-pollinatorinteraction. Ecology, 69, 1298-1302. Hamilton,W.D. (1980) Sex versusnon-sex versus parasite. Oikos, Bronstein,J.L. (1991) The nonpollinatingwasp fauna of Ficus 35, 282-290. pertusa:exploitation of a mutualism?Oikos, 61, 175-186. Hamilton,W.D. & Zuk, M. (1982) Heritabletrue fitness and bright Bronstein,J.L. (1992) Seed predatorsas mutualists:ecology and birds:a role forparasites? Science, 218, 384-387. evolution of the fig/pollinatorinteraction. -plant Hassell, M.P. & Godfray,H.C.J. (1992) The populationbiology of interactions(ed. by E. Bernays).CRC Press,Boca Raton. insectparasitoids. Natural enemies(ed. by M.J. Crawley),pp. Compton,S.G. (1993a) An associationbetween epichrysomallines 265-292. BlackwellScientific Publications, Oxford. and eurytomids(Hymenoptera: Chalcidoidea) in southernAfrican Hassell,M.P. & Pacala, S.W. (1990) Heterogeneityand thedynamics figwasp communities.Afric. Entomol. 1, 123-125. of host-parasitoidinteractions. Phil. Trans.Roy. Soc. Lond B, Compton, S.G. (1993b) One way to be a fig.Afric. Entomol. 1, 330, 203-220. 151-158. Hassell, M.P., May, R.M., Pacala, S.W. & Chesson, P.L. (1991) Compton,S.G. & Hawkins, B.A. (1992) Determinantsof species The persistence of host-parasitoid associations in patchy richnessin southernAfrican fig wasp assemblages. Oecologia, environments.I. A generalcriterion. Am. Nat. 138, 568-583. 91, 68-74. Hawkins, B.A. & Compton, S.G. (1992) African fig wasp Compton,S.G. & Nefdt,R.J.C. (1990) The figsand figwasps of communities:undersaturation and latitudinalgradients in species Ficus burtt-davyi.Mitt. Inst.Allg. Bot. Hamburg,23a, 441-450. richness.J. Anim.Ecol. 61, 361-372. Compton,S.G. & van Noort, S. (1992) SouthernAfrican fig wasps Herre,E.A. (1985) Sex ratioadjustment in figwasps. Science,228, (Hymenoptera: Chalcidoidea): resource utilisation and host 896-898. relationships.Proc. KoninklijkeNederlandse Akademie van Herre, E.A. (1987) Optimality,plasticity, and selectiveregime in Wetenschappen,95, 423-435. figwasp sex ratios.Nature, 329, 627-629. Compton, S.G. & Robertson,H.G. (1988) Complex interactions Herre,E.A. (1989) Coevolutionof reproductivecharacteristics in betweenmutualisms: ants tendinghomopterans protect fig seeds twelve species of new world figs and theirpollinator wasps. and pollinators.Ecology, 69, 1302-1305. Experientia,45, 637-647. Compton,S.G., Rasplus, J. & Ware,A.B. (1994) Africanfig wasp Herre, E.A. (1993) Population-structureand the evolution of parasitoid communities.Parasitoid communityecology (ed. by virulencein parasites of fig wasps. Science, 259, B.A. Hawkinsand W. Sheehan),pp. 343-370. OxfordUniversity 1442-1445. Press,Oxford. Herre, E.A. (1996) An overviewof studies on a communityof Compton,S.G., Ross, S.J. & Thornton,I.W.B. (1994) Pollinator Panamanian figs.J. Biogeogr.23, 593-607. limitationof fig tree reproduction on theisland of Anak Krakatau Herre,E.A., West,S.A., Cook, J.M., Compton,S.G. & Kjellberg, (Indonesia). Biotropica,26, 180-186. F. (1996a) Fig wasp matingsystems: pollinators and parasites, Cook, J.M. & Power, S. (1996) Effectsof within-treeflowering sex ratio adjustment and male polymorphism,population asynchronyon the dynamicsof seed and wasp productionin an structureand itsconsequences. Social competitionand cooperation Australianfig species. J. Biogeogr.23, 487-493. in insectsand arachnids.Vol. I. The evolutionof matingsystems Crawley,M.J. (1993) GLIM for ecologists.Blackwell Scientific (ed. by J. Choe and B. Crespi). CambridgeUniversity Press, Publications,Oxford. Cambridge. Frank,S.A. (1984) The behaviorand morphologyof the figwasps Herre, E.A., Machado, C.A., Bermingham,E., Nason, J.D., Pegoscapus aseutusand P jiminezi:descriptions and suggested Windsor,D.M., MacCafferty,S.S., van Houten,W. & Bachmann, behaviorsfor phylogenetic studies. Psyche, 91, 289-307. K. (1996b) Molecular phylogeniesof figsand theirpollinating Frank, S.A. (1985) Hierarchialselection theory and sex ratios.II. wasps. J. Biogeogr.23, 521-530. On applyingthe theory,and a testwith fig wasps. Evolution,39, Jones, T.H., Hassell, M.P. & Pacala, S.W. (1993) Spatial 949-964. heterogeneityand the populationdynamics of a host-parasitoid Frank, S.A. (1989) Ecological and evolutionarydynamics of fig system.J. anim.Ecol. 62, 251-262. communities.Experientia, 45, 674-680. Jones,T.H., Hassell,M.P. & May,R.M. (1994) Populationdynamics Free, C.A., Beddington,J.R. & Lawton, J.H. (1977) On the of host-parasitoidinteractions. Parasitoid communityecology inadequacyof simple models of mutual interference for parasitism (ed. by B.A. Hawkins and W. Sheehan), pp. 371-394. Oxford and predation.J. anim.Ecol. 42, 779-784. UniversityPress, Oxford.

? BlackwellScience Ltd 1996,Joumal of Biogeography, 23, 447-458 458 StuartA. Westet al.

Machado, C.A., Herre, E.A., McCafferty,S.S. & Bermingham, Rohani,P., Godfray, H.C.J. & Hassell,M.P. (1994) Aggregationand E. (1996) Molecular phylogeniesof fig pollinatingand non- the dynamicsof host-parasitoidsystems: a discrete-generation pollinatingwasps and the implicationsfor the origin and model with within-generationredistribution. Am. Nat. 144, evolutionof the fig-fig wasp mutualism.J. Biogeogr.23, 531-542. 491-509. Memmot,J., Godfray, H.C.J. & Gauld, I.D. (1994) The structure Taylor,A.D. (1993) Heterogeneityin host-parasitoidinteractions: of a tropical host-parasitoidcommunity. J. anim. Ecol. 63, 'aggregationof risk'and the 'CV2>1 rule'. TREE, 8, 400-405. 521-540. Thompson, J.N. (1986a) Oviposition behaviour and searching Milton,K., Windsor,D.M., Morrison,D.W. & Estribi,M.A. (1982) efficiencyin a natural population of a braconid parasitoid.J. Fruitingphenologies of two neotropicalFicus species. Ecology, anim.Ecol. 55, 351-360. 63, 752-762. Thompson,J.N. (1986b) Patternsin coevolution.Coevolution and Morre,P.D. (1994) The yucca expediency.Nature, 386, 588-589. systematics(ed. by A.R. Stone and D.L. Hawksworth),pp. Morrison,D.W. (1978) Foraging ecology and energeticsof the 119-143. ClarendonPress, Oxford. frugivorousbat Artibiusjamaicensis. Ecology 59, 716-723. Ulenberg, S.A. (1985) The phylogenyof the genus Murray,M.G. (1987) The closed environmentof thefig receptacle Coquerlin relationto its hosts, CeratosolenMayr (Agaonidae) and its influenceon male conflictin the Old World figwasp, and Ficus L. The systematicsof thefig wasp parasites of the Philotropensispilosa. Anim.Behav. 35, 488-506. genus ApocryptaCoquerl (ed. S.A. Ulenberg).North Holland Murray,M.G. (1989) Environmentalconstraints on fightingin PublishingCo., Amsterdam. flightlessmale figwasps. Anim.Behav. 38, 186-193. van Noort, S. (1991) The systematicsand phylogeneticsof the Pacala, S. & Hassell, M.P. (1991) The persistenceof hostparasitoid Sycoecinae (Agaonidae, Chalcidoidea, Hymenoptera). PhD Thesis,Rhodes University,Grahamstown, South Africa. associations in a patchy environment.II. Evaluation of field van Noort, S., Ware, A.B. & Compton, S.G. (1989) Pollinator- data. Am. Nat. 138, 584-605. specificvolatile attractants released from the figsof Ficus burtt- Pacala, S., Hassell, M.P. & May, R.M. (1990) Host-parasitoid davyi.Suid-Afrikaanse Tydskrif vir Wetensap.85, 323-324. associationsin patchyenvironments. Nature, 344, 150-153. Verkerke,W. (1986) Anatomy of Ficus ottoniifolia(Moraceae) Patel, A., Hossaert-McKey, M. & McKey, D. (1993) Ficus- syconiaand its role in the fig-figwasp symbiosis.Proc. K ned. pollinatorresearch in India: past, presentand future.Curr. Sci. Akad. Wet.89, 443-469. 65, 243-253. Verkerke,W. (1989) Structureand functionof the fig.Experientia, Patino, S., Herre, E.A. & Tyree, M.T. (1994) Physiological 45, 612-622. determinantsof Ficus fruittemperature and implicationsfor Ware, A.B., Kaye, P.T., Compton, S.G. & van Noort, S. (1993) survival of pollinator wasp species: comparativephysiology Fig volatiles:their role in attractingpollinators and maintaining throughan energybudget approach. Oecologia,100, 13-20. pollinatorspecificity. Plant Syst. Evol. 186, 147-156. Pellmyr,0. (1995) Windowsof fruitproduction in yuccas?No. A West,S.A. & Herre,E.A. (1994) The ecology of the New World critiqueof Jameset al. Oikos,72, 145-147. fig-parasitisingwasps Idarnesand implicationsfor the evolution Price, P.W. (1980) Evolutionarybiology of parasites. Princeton of the fig-pollinatormutualism. Proc. R. Soc. Lond B, 258, UniversityPress, Princeton, NJ. 67-72. Ramirez, B.W. (1970) Host specificityof fig wasps (Agaonidae). Wiebes,J.T. (1979) Co-evolutionof figsand theirinsect pollinators. Evolution,24, 681-691. A. Rev. Ecol. Syst. 10, 1-12. Redfern,M., Jones,T.H. & Hassell, M.P. (1992) Heterogeneityand Windsor,D.M., Morrison,D.W., Estribi,M.A. & de Leon, B. densitydependence in a field study of a tephritid-parasitoid (1989) Phenologyof fruitand leaf productionby 'strangler'figs interaction.Ecol. Ent. 17, 255-262. on Barro Colorado Island, Panama. Experientia,45, 647-653.

? BlackwellScience Ltd 1996,Journal of Biogeography, 23, 447-458