Nordic Society Oikos

Damage Tolerance and Cotyledonary Resource Use in the Tropical tree superba Author(s): J. W. Dalling and Kyle E. Harms Source: Oikos, Vol. 85, No. 2 (May, 1999), pp. 257-264 Published by: Wiley on behalf of Nordic Society Oikos Stable URL: http://www.jstor.org/stable/3546491 Accessed: 15-04-2015 19:04 UTC

REFERENCES Linked references are available on JSTOR for this article: http://www.jstor.org/stable/3546491?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Wiley and Nordic Society Oikos are collaborating with JSTOR to digitize, preserve and extend access to Oikos.

http://www.jstor.org

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions OIKOS 85: 257-264.Copenhagen 1999

Damage toleranceand cotyledonaryresource use in thetropical treeGustavia superba

J. W. Dalling and Kyle E. Harms

Dalling, J. W. and Harms,K. E. 1999. Damage toleranceand cotyledonaryresource use in the tropicaltree Gustaviasuperba. - Oikos 85: 257-264.

Explanations for the extremelylarge seed size of some tropical forestplants are currentlylacking. In thisstudy we examinecotyledonary resource use of the tropical tree Gustavia superba, and test whethertolerance to damage of either seeds or establishingseedlings might be an importantfunction leading to, or maintaining,its large ( > 5 g) seed reserves.We found that seeds of Gustaviawere both tolerantof insectinfestation and were also capable of successfulgermination after removal of half of their cotyledonaryreserves. Simulated complete above-groundherbivory resultedin repeatedresprouting (up to 8 times).Resprout shoots were constructed of a small,but fixedproportion of remainingcotyledonary mass regardlessof seed size. In the absence of damage,cotyledon reserves were used foronward seedling growth; however,conversion of cotyledonresources was slow, lastingseveral months. Given high rates of damage to Gustaviaseeds and seedlingsin the field,and the apparent use of cotyledonaryreserves to tolerate or recover from it in growing house experiments,we concludethat cotyledonary seed size and morphologyin thisspecies is adaptive in survivingpre- and earlypost-germination hazards.

J. W. Dalling and K. E. Harms,Smithsonian Tropical Research Inst., Apartado 2072, Balboa, Panama (dalling/@bci.si.edi).

The adaptive value of large, persistentseed reserves trampling(Denslow 1980, Clark and Clark 1985, 1989, presentin some tropicalrain foresttrees remains elu- 1991, Armstrongand Westoby 1993, Andersson and sive. While large seed reserveshave long been thought Frost 1996, Harms and Dalling 1997). Large seed re- to be associated with seedlingestablishment require- servesmay also allow successfulemergence from deeply ments in shady understoreyconditions (e.g. Troup buriedcache sites (Vander Wall 1993), or seed reserves 1921, Salisbury 1942, Snow 1971, Foster and Janson themselvesmay providea rewardto dispersers(Steele et 1985; but see Kelly and Purvis 1993, Hammond and al. 1993, Andersson and Frost 1996, Grubb 1996). Brown 1995) shade persistencedoes not always corre- While evidenceis accumulatingthat each of these fac- late with seed size (Augspurger1984; but see Boot tors may contributesome selectiveadvantage to large 1996) and some of the smallest seeds ( < 1 mg) are seed size in some species,their relative importance has produced by shade-tolerantspecies (e.g. Metcalfeand seldom been assessed for individualspecies. Grubb 1995). We suggesttherefore that the adaptive In this study we examined the resource allocation value of large seed size may not be relatedto low levels patterns in seedlings of the tree Gustavia superba of irradianceper se. (H.B.K.) Berg ()in centralPanama, and Alternativeexplanations of large seed reservesare examinedwhether seed and seedlingdamage, and scat- that theyafford greater tolerance of pre- and post-ger- ter-hoardinghave selected for, or are importantin, minationhazards, such as partial seed predation(e.g. maintaininglarge seed size in this species. Using an Janzen1971, 1976,Dirzo 1984,Dalling et al. 1997),and experimentaland observationalapproach we assess the seedling damage from herbivores,falling litter,and applicabilityof the followingpredictions for Gustavia

Accepted 18 August 1998 Copyright(C OIKOS 1999 ISSN 0030-1299 Printedin Ireland - all rightsreserved

OIKOS 85:2 (1999) 257

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions that resultfrom either damage toleranceor dispersal Dalling and K. Harms pers. obs.). Under growing advantages: 1) cotyledonaryseed reservesare retained house conditions,even small fragments(1-2 g) of afterthe productionof a functionalshoot, and regard- excisedcotyledons are capable of producingfunctional less of initial seed size; 2) seeds remain capable of seedlings (Harms et al. 1997). Most fallen Gustavia germinationeven after insect infestation,or partial fruitscontain at least some seeds that have been dam- consumption;3) seeds are capable of producingfunc- aged by theclearwing moth Carmentaforaseminis (Sesi- tional resprout shoots after complete above-ground idae) (Harms and Aiello 1995). herbivory;4) seed reservesof deeplyburied seeds (simu- lating scatter-hoarding)are not entirelyused in reach- ing the soil surface; 5) cotyledonaryreserves are Effectsof insectseed damage on seed germination retained,rather than used to augmentonward seedling growthin the absence of physicaldamage; 6) resprout- All seeds for all experimentswere collected frombe- ing is common for naturallyregenerating seedlings in neath ca 20 treesalong Barbour Trail on BCI. In July the field. 1994, we collected 500 seeds from fallen, indehisced fruits.No evidenceof insectinfestation was apparent on the seeds at the time the studywas initiated.Seeds wereplaced in a growinghouse undershade conditions Methods (0.4-1.0 mol m-2 d -' PAR), and planted on the surface of ten germinationtrays, watered daily, and Studysite and speciesdescription censused weekly over the followingeight weeks for The studywas carriedout in seasonallymoist tropical germinationand signsof insectinfestation. forest on Barro Colorado Island (BCI), in central Panama. Rainfallon BCI averages2600 mm yr- 1, with a pronounceddry season fromJanuary to April(Wind- Effectsof seed size on initialshoot size sor 1990). The flora,fauna, geologyand hydrologyof the site are describedby Croat (1978) and Leigh et al. In May 1994,we collected50 seeds exhibitinga natural (1982). range in seed size, and accuratelymeasured their fresh Gustaviasuperba is a common sub-canopytree (10- masses. Seeds wereplanted in the growinghouse (con- 20 m tall) of lowland Tropical Moist and Wet Forest ditionsas above). Three weeks afterthe emergenceof (sensu Holdridge et al. 1971) of Central America and the initial sprout we excised the entireabove-ground northernSouth America(Croat 1978,Prance and Mori shoot above thecotyledons, dried it in the oven at 70'C 1979). On BCI, Gustaviais a common componentof for 72 h and measured the dry mass. Seedlings that 40-100 yr old secondaryforest on the easternside of developed two or more above-groundsprouts simulta- the island. During the mid-wetseason (May-August) neously or that showed signs of insectdamage to the adults of Gustaviaproduce indehiscentfruits of 150- cotyledonswere excluded(n = 1). 600 g containingseeds of variable number and size. Sork (1987) reports5-50 seeds/fruit,and a mass of 3-15 g/seed; Forget (1992) reports an average of 7 Effectsof partialcotyledon removal on seedling viable and an mass of 12.6 g/seed, seeds/fruit average size and resproutingcapacity (n = 50). Seeds appear morphologicallyundifferentiated (i.e. no distinctembryo is visibleupon dissection),and In June 1993, we randomlyassigned 90 seeds to three consist largelyof 2-4 (mode = 3) cotyledonsegments. categories:unmanipulated, 25% seed (cotyledon)mass Seed germinationis hypogeal,with the cotyledonsof removal and 50% seed mass removal. Seeds were re- scatter-hoarded,or otherwiseburied seeds remaining weighed and cut until the appropriateproportion of beneath the soil. For unburied seeds, the cotyledons cotyledonmass had been removed. Seeds were ran- turn green on exposure to light; however,since the domlyplaced on forestsoil in threelarge seedlingtrays cotyledonslack stomata,they are unlikelyto undertake (conditions as above) and watered daily. The initial much photosynthesis(K. Kitajima pers. comm.). cohort of seeds to germinatewas observedclosely to Sork (1985, 1987) and Forget(1992) foundthat seeds determinethe length of time required from the first of Gustavia are both dispersed (scatter-hoarded)and emergenceof an above-groundsprout until the first consumedby rodentson BCI. Forget(1992) foundthat pair of leaves had just fullyexpanded (i.e. theshoot was (Dasyproctapunctata) usually dig up and con- fullyfunctional). This time period (3 wk) was highly sume Gustavia cotyledonsafter germination.Cotyle- consistentamong seedlings. dons are sometimescarried away fromseedlings and Three weeksafter sprout emergence the entireabove- partial cotyledondamage occurs (Forget 1992). Seeds ground shoot above the cotyledonswas excised from are also commonlyfragmented before or during the the , and dried and measured as above. Since processof removalfrom the tough,indehiscent fruit (J. Gustavialeaves are nearlywhite when theyemerge and

258 OIKOS 85:2 (1999)

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions begin expanding(i.e. withoutchlorophyll) we assume Field surveyof resproutingfrequency thatthese excised parts were constructed almost entirely out of the energyreserves of the seed. Observationsof In June 1994 we marked 100 newlyemerged seedlings one "albino" seed (completelylacking chlorophyll) (with cotyledons attached and <6 fully expanded seem to support this, as it also produced an initial leaves) along BarbourTrail on BCI. In August 1994 we seedling and resprout seedlings similar to those of marked an additional 100 seedlingsthat germinated normal . We continuedto census seedlingsafter later in the year along WheelerTrail on BCI and that excision of the shoot and harvestedeach successive, were in a similar developmentstage. In contrast to newlyexpanded resproutthat was produced.We con- subsequentyears (1995-1997), in 1994 many Gustavia tinued the experimentuntil no new resproutswere seeds germinatedon the soil surface without being produced (59 wk afterplanting). consumedor scatter-hoarded,and all the seedlingswe markedhad open cotyledonsattached to seedlingsand exposed above the soil. Seedlingsalong Barbour trail werere-censused in Augustand November1994, and in January,March, and June 1995 and at Wheelersite in Effectof seed burialon initialshoot size and October 1994 and in January,March, and June 1995. resproutingcapacity At each recensuswe recordedthe numberof remaining Since the scatter-hoardingrodents that disperse Gus- cotyledons,and the numberof resproutingseedlings. tavia seeds can buryseeds to deep soil layers(up to 8 cm; Smythe 1978, Hallwachs 1986) we examined the capacity of Gustavia seeds to establish and resprout Statisticalanalysis froma range of soil depths.In June 1993 we collected Analysesof variance (ANOVA) were performedusing 180 seeds (mean = 9.0 g, s.e. = 0.4), and buriedseeds at the MGLH proceduresin SYSTAT (Wilkinsonet al. six soil depths (0, 1, 2, 5, 10 and 15 cm). Seeds were 1992). Effectsof partialcotyledon removal and of seed buriedin 20 cm cm in diameter,20 depth pots groups burial on the initial shoot and subsequent resprout of 10 seeds per pot and threepots per treatment.Pots mass, were both analysed by repeated measures were filledwith sieved (5 mm mesh) forestsoil and ANOVA, usingGreenhouse-Geisser adjusted F-tests of placed in a growinghouse (conditionsas above). Pots betweensubjects factors. For the partial cotyledonre- were censused weeklyand the newlyexpanded above- moval experiment,results of analysesperformed on the ground shoots excised as above, dried and measured. initial shoot and two subsequentversus three subse- The experimentwas terminatedafter 61 wk when no quent resproutswere verysimilar, and the analysisfor new resproutswere produced. threeresprouts is presentedhere. For the seed burial experiment,analyses are performedon the per pot means of the initial shoot mass and threesubsequent resprouts.Kruskal-Wallis non-parametric ANOVA was Effectof completecotyledon removal on onward used to analyse timeto seedlingemergence. seedlinggrowth Effectsof completecotyledon removal on subsequent seedlinggrowth were analysed using analysis of covari- In July1997 we selected120 undamagedseeds (location ance (ANCOVA) of seedlingheight and seedlingleaf as above) to examinethe effect of cotyledonremoval on area. The covariates(leaf area and seedlingheight at subsequent seedlinggrowth. Seed size in this experi- the timeof treatmentapplication) were used aftertest- = = mentranged from 9.0 g to 32.1 g (mean 16.8, s.e. ing for an interactionbetween the covariateand treat- in 0.5). Seeds were assigned to treatments a stratified ment(the homogeneityof slopes assumption). random manner to ensure equality of mean seed masses, and planted into individual2-L pots of forest soil in a growinghouse (conditionsas above). In two cotyledonremoval treatments, seedlings were grown for Results eitherthree weeks or nine weeks aftershoot emergence Insectinfestation and seed germination and then seedlingheight and leaf area were measured and cotyledonswere completelyremoved. Six weeks Of the 500 seeds collected,463 germinatedwithin the later height and leaf area were remeasured and eightweeks of the study.The remaining37 seeds com- seedlingswere harvestedfor total drymass. Cotyledon pletelyrotted away, as a result of insect infestation removaltreatments were paired withcontrols in which and/orfungal infection. Insect infestation however was seedlingsretained their cotyledons throughout the ex- not necessarilyfatal; one or more of pupae or exit periment.Final cotyledon dry mass (at excision or tunnelsof the sesiidmoth (Carmentaforaseminis; Eich- harvest,depending on treatment)was also measuredto lin 1995) were found in 124 of the 463 (27%) of the determinerates of cotyledonresource use. seeds whichsuccessfully germinated.

OIKOS 85:2 (1999) 259

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions Table 1. Seed mass, initialseedling mass and resproutingcharacteristics of whole and cut seeds.

Treatment' Initial seed mass Initial shoot dry mass First resproutdry mass Average numberof (g ? s.e.) [n]2 (g + s.e.) [n] (g ? s.e.) [n] resprouts3(range)

Whole seeds 14.4+ 1.2 [30] 0.43+0.1 [11] 0.39+0.1 [11] 3.6 (2-8) 25% removed 13.1 + 0.9 [30] 0.41 + 0.1 [14] 0.23 + 0.0 [14] 2.7 (0-5) 50% removed 14.7 + 1.3 [30] 0.32 + 0.1 [20] 0.22 + 0.0 [19] 2.1 (0-6)

' Cut treatmentsare exact 25% and 50%10reductions in the initialseed masses. 2 Seed mass priorto cut. ' Excludingthe initialshoot.

Seedlingsize and resproutingcapacity wk) were slowerto emergethan seeds buriedin the soil at < 10 cm (median= 8 wk). Seed burialhad littleeffect There was a stronglinear relationshipbetween initial on eitherthe initialshoot size or resproutingcapacity seed mass and the initial shoot mass (r2 = 0.85; df= exceptat a depthof 15 cm. The initialshoot sizes were 1,37; F= 202.4; p < 0.001). Converting seed fresh significantlydifferent across the burial treatments(one- masses to dry mass (dry mass = 47.5% fresh mass) way ANOVA; df= 5,12;F= 3.3,p < 0.05), but post hoc indicates that 10% of the seed dry mass is used for testsshowed that only the mean shoot mass of plantsin constructionof the initialabove-ground shoot. the 15 cm depthtreatment was smallerthan for shoots All but one of the 90 seeds used in the partial of seeds buriedat depthsof < 5 cm. Similarly,although cotyledon removal experimentgerminated. Although therewas a significanteffect of burialdepth on thetotal the time taken fromplanting to the emergenceof an shoot mass produced (one-way ANOVA; df= 5,12; above-groundshoot variedfrom 2 to 21 wk, themedian F= 3.3,p < 0.05), thiswas onlysignificant for compari- timeto germination(4 wk) was the same for all treat- sons withthe 15 cm treatment. ments.Many of the seedlings(n = 42) producedmulti- ple resproutsconcurrently after one of the sequential resproutremovals. Although these representa large proportionof our data, theyare excludedfrom subse- Cotyledonremoval and seedlinggrowth quent analysessince the concurrentproduction of mul- The removal of cotyledonsthree weeks after shoot tiple resprout shoots is likely to affectpatterns of emergencehad a small, but significant,effect on leaf resourceallocation in a way thatdiffers from allocation area production (ANCOVA df= 1,93, F = 4.4, p < in individualswith sequentially produced shoots. 0.05), and whole plant biomass (ANCOVA df= 1,93, Both cut and uncutseeds werecapable of resprouting F= 7.2, p < 0.01), but not on seedlingheight 6 wk later multipletimes; uncut seeds produced up to eight se- (Fig. 2). In contrast,cotyledon removal 9 wk after quential above-groundresprouts, and cut seeds up to shoot emergencehad no effecton seedlingsize. How- six resprouts(Table 1). Justas the initialshoot size is ever, after9 wk only a quarterof the originalcotyle- related to seed mass, it appears that resproutsize is donary dry mass remained,as opposed to nearlyhalf relatedto the size of the remainingseed reservesof the afterthree weeks (Fig. 3). plant.Thus, as each sequentialresprout is producedthe Seedlings of Gustavia showed strongperiodicity in size of the subsequent resprout diminishes(Fig. 1, their growth;leaves were produced in flushesrather Table 2, withinsubjects 'shoot'). Whole seeds produced largershoots than seeds withhalf of thecotyledon mass 0.6- y = - removed (Tukey HSD, post hoc test,p <0.05), but S Wholeseeds 0.55 0.lOx n=11seeds interestingly,partial removal of seed reservesdid not 0.5 0 50 % removed y = 0.42 - 0. lox n=20 seeds alter the allocation patternof seed reservesto shoots (Table 2, Treatmentx shoot interactionis not signifi- 0.4 , cant; slopes parallel in Fig. 1). In effect,large seeds withpartially removed cotyledons behaved like smaller 0.3- 4.1 0 seeds. 0 .C 0.2 -

0.1 Effectof seed burial 0.0~ Seed burial,in contrast,did affectthe timing of seedling 2 3 4 Successiveshoot harvests emergence(Kruskal-Wallis test = 14.6,df = 5, p < 0.05). Seeds placed on the soil surfaceproduced shoots faster Fig. 1. Mass of sequentiallyharvested, above-ground shoots for seeds with intact cotyledons,and seeds with 50% of (median= 5 wk) thanseeds buried in thesoil (median= cotyledonmass removed.Data for seeds in the 25% removal 12 wk), and seeds buriedat 15 cm depth (median= 30 treatmentare omittedfor clarity.Error bars are 1 s.e.

260 OIKOS 85:2 (1999)

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions Table 2. Repeated measuresANOVA of cut treatment(whole 100l seeds, 25% cotyledonremoval and 50% removal) on the size of sequentiallycut resproutshoots. Only data for seedlings that produced threeresprouts are included. g 80-

Source df MS F ~60 Betweensubjects 0 Cut treatment 2 0.17 3.6* X 40- Error 21 0.05 0 U Withinsubjects 20- Shoot 3 0.44 30.8*** Shoot x Cut treatment 6 0.03 2.2 0f. , , i Error 63 0.01 0 3 6 9 12 15 Timesince germination (weeks) * P<0.05, *** P<0.001. Fig. 3. Meanpercentage ( + 1 s.e.) of remainingcotyledonary drymass for seedlings with cotyledons removed after 3, 9, and thancontinuously, and theinterval between flushes was 15 wk.The initialdry mass was calculatedassuming a 52.5%Yo seedmoisture content. Line drawn by eye. sufficientthat many seedlings produced no newleaves betweenthe timeof cotyledonremoval and harvest. For the3-wk cotyledon removal treatment, significantly moreseedlings failed to flushnew leaves (10 out of 25) portionof seedlingsflushing new leaves was not differ- 9-wk than for the 3-wkcontrol treatment (3 out of 26; ent betweenthe cotyledonremoval and control treatments(6 out of 18 and 11 out of 31, respectively). Yates-correctedx2 = 11.1, df= 1, p <0.001). The pro-

300 3 weeks = 9 weeks Cotyledonaryuse and resproutingin the field a) n=_9 Naturallyrecruiting Gustavia seedlings in 1994 showed relativelyhigh survivorship.Ten and twelve months 200. after the initial censuses, 60% and 53% of seedlings were still alive at Wheeler and Barbour sites, respec- tively(Table 3). Most seedlingslost theircotyledons gradually.At Barbour site 31% of seedlings,and at 100- Wheelersite 28% of seedlingslost one or more,but not all theircotyledons in the firsttwo months.Cotyledons eitherrotted away, weremobilized by the plant,or were 0 _J I lost. Cotyledon loss presumablyoccurred because the Cut Control Cut Control brittleattachment between stem and cotyledon was broken, or because cotyledons were excised. Few seedlings resproutedin the firsttwo months when 20 3 weeks 9 weeks b) cotyledonswere still attached, but after10 months16% and 27% of the survivingseedlings at the two siteshad resproutedat least once. U~~~~

Discussion

10 Gustavia appears to be another example of a large seeded tree that is remarkablytolerant to insect seed predators(Janzen 1971, 1976, Forget 1992). Resultsfor this species appear to be broadly similarto those for Cut Control Cut Control Prioriacopaifera (Dalling et al. 1997); seed germination Treatment appeared to be relativelylittle affected by infestationby Fig.2. a) Meanleaf area and b) meantotal biomass (+ 1 s.e.) of seedlingsin two treatmentgroups. In the cut treatmentmoth larvae, and seeds with as much as half of their cotyledonswere excisedfrom the plant 3 or 9 wk after cotyledonaryreserves removed germinatedas well as above-groundshoot emergence.In the controltreatment, intact seeds. Seed damage tolerance,however, may be cotyledonswere left on the plantand could contributeto the resultof morphologicalor chemicaladaptations as seedlinggrowth. Open columns indicate values at thetime the cut treatmentswere applied, and greycolumns values 6 wk much as size alone. In oaks (Quercus spp.), for exam- aftertreatment application. ple, chemicaldefences of the seed reservesare concen-

OIKOS 85:2 (1999) 261

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions trated in the apical region of the acorn around the Graham 1983,Hallwachs 1986) burialdid not affectthe embryoresulting frequently in only partial consump- seedlingsize or the resproutingcapacity of the plant. tion of the seed (Steele et al. 1993). In Gustavia,the Seed burialat 15 cm howeverdid affectboth the timing seed consistsof two or threefleshy cotyledon segments, of shoot emergenceand the size of the initial shoot, withno clearlydifferentiated region containing a devel- though burial at depths this great may be rare under oped embryo.Presumably meristematic tissue is spread naturalconditions. throughoutthe cotyledonsbecause even small cotyle- How long are cotyledonaryreserves retained by the donary fragmentsare capable of generatingseedlings seedling?Our study suggeststhat the importanceof (Harms et al. 1997). cotyledonreserves for resprouting may be limitedto the Like otherlarge seeded species on BCI (Harms and firstfew weeks following shoot emergence.Comparison Dalling 1997), the expansion of the initial seedlingof of seedlinggrowth with and withoutcotyledon removal Gustaviarequired use of only a fairlysmall proportion showed that cotyledonscontributed slightly to subse- of the total cotyledonaryreserves (10% for the above- quent seedlinggrowth three weeks aftershoot emer- ground shoot). This phenomenonhas also been ob- gence,but the contributionshad mostlydisappeared by served for oaks, in which the cotyledonshave been 15 wk post-emergence.Cotyledon use may in part be considered 'over-sized' for the germinationprocess determinedby the seedling's light environment,with perhaps more rapid assimilationof seed reservesin (Sonesson 1994, Anderssonand Frost 1996). In fact, deep shade, or perhaps cotyledonspersist longer if among Gustaviaseeds of widelyvarying mass, and in seeds are buried, and less prone to desiccation. seeds wherecotyledonary reserves were removedprior Nonetheless,the contributionof cotyledonsduring the to germination,the same proportionof seed reserves firstfew weeks of shoot emergencemay be criticalas were allocated to formingthe initial fullyexpanded this would be predictedto be the period when the seedling.This apparentpattern of partial seed reserve probabilityof damage is highest;tissue loss to herbi- allocation may simplyresult from constraints on the vores in tropical foreststends to be much higheron rate at which seed reservesare mobilized, but also youngleaves than on toughenedfully expanded leaves suggestsan adaptiveexplanation whereby seed reserves (Coley and Barone 1996). are retainedin case the newlyformed seedling is dam- Detailed studiesof seed fate of Gustaviacarried out aged. In supportof this view we found that simulated by Sork (1985, 1987) and Forget (1992), however,do herbivory(removal of the entireabove ground shoot) not reportobservations of browsingand resproutingin resulted in rapid resprouting.Furthermore, repeated this species. They may simplyhave missed the phe- resproutingwas possible as each new shoot used the nomenon;in the growinghouse resproutseedlings grew same, consistentproportion of remainingcotyledonary veryquickly, often becoming fully expanded within six reserves. weeks of the original cut. Certainly,browsing and We can exclude the possibilitythat large seed size in resproutingappear to be a likelyexplanation of Sork's Gustavia is simply the result of resource needs to (1987) findingthat seedlingsexposed to mammalswere emergefrom deep burial followingscatter-hoarding, as shorterthan those withinmammal enclosures.Since suggestedby Vander Wall (1993), as our experiment Sork's recensuses of seedlings were conducted only showed that withinthe range of naturalscatter-hoard- every 5-6 months,evidence of resproutingmay have ing (1-8 cm burial depths;Smythe 1978, Hopkins and been difficultto detect.

Table 3. Survival,cotyledon loss, and resproutingfrequency for naturallyrecruiting Gustavia seedlings at two sites on BCI. Barbour (BAR), n= 100 seeds, and Wheeler(WHE), n= 100 seeds.

Census period

Site Jun94 Aug 94 Nov 94 Jan 95 Mar 95 Jun95

Surviving BAR 100 93 64 62 57 53 seedlings WHE - 100 92 76 63 60 With cots' BAR 100 65 0 0 0 0 WHE - 100 57 0 0 0 Mean cots2 BAR 2.6 1.4 0 0 0 0 WHE - 2.4 1.3 0 0 0 Resprouts3 BAR 0 3 4 9 13 14 WHE - 0 1 10 12 17 Resprout(%)4 BAR 0 3 6 11 16 16 WHE - 0 1 13 17 27

'Number of seedlingswith cotyledons attached. 2 Mean numberof cotyledonsegments attached. 3Total numberof seedlingswith resprout shoots. ' Percentof seedlingsalive in each census period resprouting.

262 OIKOS 85:2 (1999)

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions Paradoxically,Sork (1987) and Forget (1992) both Acknowledgements- We thankthe SmithsonianTropical Re- found that seedlingdamage and mortalityin Gustavia searchInstitute for providing facilities, the Smithsonian Insti- tutionMellon program (JWD), weremainly caused by rodentsthat dig up PrincetonUniversity (KEH), germinating and the SmithsonianTropical Research Institute, Office of seedlingsexpressly to consume the cotyledons.Sork Educationfor financial support. Rafael Aizpr-da, Katia Silvera (1987) foundthat seedlingscontinued to be dug up by and ArturoMorris provided essential help in the fieldand greenhouse.Kaoru predatorsfor more than a year aftergermination; For- Kitajima,Jessica Eberhard and David Clark made helpfulsuggestions and commentson the get (1992) foundthat agoutis (Dasyprocta punctata) ate manuscript. or nibbled on Gustaviacotyledons up to four months after seedlings germinated.If cotyledon reservesare particularlyattractive to some herbivores,then why would theybe retainedfor so long by the plant rather References than mobilized?Perhaps a trade-offexists between the Andersson,C. and Frost, I. 1996. Growthof Quercusrobur advantages of resproutingfrom cotyledons as against seedlings after experimentalgrazing and cotyledon re- moval. - Acta Bot. Neerl. 45: 85-94. the additionalpredation risks associated withretaining Armstrong,D. P. and Westoby,M. 1993. Seedlingsfrom large them. Older seedlings that lack cotyledonsare also seeds toleratedefoliation better: a test using phylogeneti- capable of resprouting,which suggeststhat at some cally independentcontrasts. - Ecology 74: 1092-1100. Augspurger,C. K. point in ontogenya shiftoccurs in the location of 1984. Light requirementsof neotropical treeseedlings: a comparativestudy of growthand survival. storedreserves used forresprouting from cotyledons to - J. Ecol. 72: 777-795. roots. Boot, R. G. A. 1996. The significanceof seedlingsize and Alternatively,if little resproutingfrom cotyledons growthrate of tropicalrain foresttree seedlings for regen- eration in canopy openings. - In: Swaine, M. D. (ed.), does occur in the field,then large seed size in Gustavia Ecology of tropical foresttree seedings.UNESCO Paris may simplyrepresent a dispersalreward, as suggested and ParthenonCarnforth, pp. 267-282. by Steele et al. (1993) for oak species. Since Gustavia Clark, D. B. and Clark, D. A. 1985. Seedlingdynamics of a tropicaltree: impactsof herbivoryand meristemdamage. seeds are primarily dispersed by scatter-hoarding - Ecology 66: 1884-1892. agoutis (Forget 1992), large seeds may indeed have a Clark, D. B. and Clark, D. A. 1989. The role of physical dispersal advantage. Hallwachs (1994), found that damage in the seedlingmortality regime of a neotropical rain forest.- Oikos 55: 225-230. agoutiscached proportionallymore acorns and coconut Clark, D. B. and Clark, D. A. 1991. The impact of physical chunksas weightincreased, and hoarded heavierunits damage on canopy treeregeneration in tropicalrain forest. furtheraway. Our fielddata on Gustaviahowever do - J. Ecol. 79: 447-457. Coley, P. D. and Barone, J. not suggestany dispersal A. 1996. Herbivoryand plant advantage,perhaps because defensesin tropicalforests. - Annu. Rev. Ecol. Syst. 27: the year we carried out our studywas atypical. 1994 305-335. appeared to be an extremelygood fruitingyear for Croat, T. R. 1978. Flora of Barro Colorado Island. - Stan- fordUniv. Press, Stanford,CA. Gustaviaon BCI, and had thethird highest total annual Dalling, J. W., Harms, K. E. and Aizprfia,R. 1997. Seed seed fall to sixty0.25-M2 seed traps recordedbetween damage toleranceand seedlingresprout ability of Prioria 1986 and 1997 (S. J. Wrightunpubl.). This super-abun- copaifera('El Cativo'). - J. Trop. Ecol. 13: 617-621. Denslow, J. S. 1980. Notes on the dance of fruitin 1994 presumablymeant that many seedling ecology of a large-seededspecies of Bombacaceae. - Biotropica 24: Gustaviaseeds escaped dispersal/predationand germi- 146-156. nated on the soil surface.In contrast,Forget's (1992) Dirzo, R. 1984. Herbivorya phytocentricoverview. - In: Dirzo R. and study was conducted in 1990, which had the lowest Sarukhdn J. (eds), Perspectivesin plant population ecology. Sinauer, Sunderland,MA, pp. 141- total fruitfall1986-1997. Whereas > 50% of unburied 165. seedlings survived one year in this study, none of Eichlin,T. D. 1995. A new Panamanianclearwing moth. - J. Forget'sunburied seedlings survived > 2 mo. Lepid. Soc. 49: 39-42. Forget, P.-M. 1992. Seed removal and seed fate in Gustavia We have suggestedthat toleranceto cotyledondam- superba(Lecythidaceae). - Biotropica24: 408-414. age and the retentionof cotyledonaryreserves after full Foster, S. A. and Janson,C. H. 1985. The relationshipbe- leaf expansionare generalfeatures shared by manytree tween seed size and establishmentconditions in tropical woody plants. - Ecology 66: 773-780. specieswith seed reserves> 5 g and hypogealgermina- Grubb, P. J. 1996. Rainforestdynamics: the need for new tion (Harms and Dalling 1997,Dalling et al. 1997). The paradigms.- In: Edwards,D. S., Booth, W. E. and Choy, primaryfunctions of large seed reservesin thesespecies S. C. (eds), Tropical rainforestresearch: current issues. Kluwer,Dordrecht, pp. may thereforebe for 215-233. dispersaland for damage toler- Hallwachs,W. 1986. Agoutis(Dasyprocta punctata): the inher- ance rather than for the tolerance or avoidance of itors of guapinol (Hymenaea courbaril:Leguminosae). - unfavourablelight conditions. The abilityto manipu- In: Estrada, A. and Fleming,T. H. (eds), Frugivoresand seed dispersal.Junk, late Gustaviaseed size without Dordrecht,pp. 285-304. affectingeither germina- Hallwachs, W. 1994. The clumsydance betweenagoutis and tion successor the seed-sizeto seedlingsize relationship. plants: scatterhoardingby Costa Rican dryforest agoutis. indicates that this species would be ideal for further - PhD dissertation,Cornell Univ. Hammond,D. S. and V. K. investigationsof the dispersalversus establishment Brown, 1995. Seed size of woody ad- plantsin relationto disturbance,dispersal, soil typein wet vantagesof large seed size in scatter-hoardedspecies. neotropicalforests. - Ecology 76: 2544-2561.

OIKOS 85:2 (1999) 263

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions Harms, K. E. and Aiello, A. 1995. Seed-boringby tropical Prance, G. T. and Mori, S. A. 1979. Lecythidaceae.- Flora clearwingmoths (Sessiidae): aberrant behavior or wide- Neotrop. Monogr. 21. spread habit? - J. Lepid. Soc. 49: 43-48. Salisbury,E. J. 1942. The reproductivecapacity of plants. - Harms, K. E. and Dalling, J. W. 1997. Damage and herbivory Bell, London. tolerance throughresprouting as an advantage of large Smythe,N. 1978. The naturalhistory of the CentralAmerican seed size in tropicaltrees and lianas. - J. Trop. Ecol. 13: (Dasyproctapunctata). - Smithson.Contrib. Zool. 481-490. 257: 1-52. Harms, K. E., Dalling, J. W. and Aizprua, R. 1997. Regener- Snow, D. W. 1971. Evolutionaryaspects of fruit-eatingin ation fromcotyledons in Gustaviasuperba (Lecythidaceae). birds. - Ibis 113: 194-202. - Biotropica29: 234-237. Sonesson, L. K. 1994. Growth and survivalafter cotyledon Holdridge,L. R., Grenke,W. G., Hatheway,W. H., Liang, T. removal in Quercus robur seedlings grown in different and Tosi, J. A., Jr. 1971. Forest environmentsin tropical naturalsoil types.- Oikos 69: 65-70. lifezones. - PergamonPress, New York. Sork, V. L. 1985. Germinationresponse in a large seeded Hopkins,M. A. and Graham,A. W. 1983. The speciescompo- Neotropicaltree species, Gustavia superba (Lecythidaceae). sition of soil seed banks beneath lowland tropical rain- - Biotropica 17: 130-135. forestsin North Queensland, Australia. - Biotropica 15: Sork, V. L. 1987. Effectsof predationand lighton seedling 90-99. establishmentin Gustaviasuperba. - Ecology 68: 1341- Janzen, D. H. 1971. Escape of juvenile Dioclea magacarpa 1350. (Leguminosae)vines frompredators in a deciduous forest. Steele,M. A., Knowles,T., Bridle,K. and Simms,E. L. 1993. - Am. Nat. 105: 97-112. Tannins and partial consumptionof acorns: implications Janzen,D. H. 1976. Reductionof Mucuna andreana(Legumi- fordispersal of oaks by seed predators.- Am. Midl. Nat. nosae) seedlingfitness by artificialseed damage. - Ecology 130: 229-238. 57: 826-828. Troup, R. S. 1921. The sylvicultureof Indian trees.Vol. 1. - Kelly, C. K. and Purvis,A. 1993. Seed size and establishment OxfordUniv. Press,Oxford. conditionsin tropicaltrees; on the use of taxonomicrelat- Vander Wall, S. B. 1993. A model of cachingdepth: implica- edness in determiningecological patterns.- Oecologia 94: tions for scatterhoardersand plant dispersal.- Am. Nat. 356-360. 141: 217-232. Leigh,Jr., E. G., Rand, S. A. and Windsor,D. M. 1982. The Wilkinson,L., Hill, M. and Vang, E. 1992. SYSTAT: Statis- ecology of a tropical forest:seasonal rhythmsand long- tics,Version 5.2 Ed. - Evanston,IL. termchanges. - SmithsonianInst. Press,Washington. Windsor,D. M. 1990. Climate and moisturevariability in a Metcalfe,D. J. and Grubb, P. J. 1995. Seed mass and light tropical forest:long-term records from Barro Colorado requirementfor regenerationin South-East Asian rain Island, Panama. - Smithson.Contrib. Earth Sci. 29: 1- forest.- Can. J. Bot. 73: 817-826. 145.

264 OIKOS 85:2 (1999)

This content downloaded from 130.126.55.129 on Wed, 15 Apr 2015 19:04:48 UTC All use subject to JSTOR Terms and Conditions