Вы находитесь на странице: 1из 18

Nordic Society Oikos

Post-Dispersal Predation on Isolated Seeds: A Comparative Study of 40 Tree Species in a


Southeast Asian Rainforest
Author(s): Geoffrey M. Blate, David R. Peart and Mark Leighton
Source: Oikos, Vol. 82, Fasc. 3 (Sep., 1998), pp. 522-538
Published by: Wiley on behalf of Nordic Society Oikos
Stable URL: http://www.jstor.org/stable/3546373 .
Accessed: 24/04/2013 14:01

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 support@jstor.org.

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 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
OIKOS 82: 522-538. Copenhagen1998

Post-dispersal
predationon isolatedseeds:a comparative
studyof
40 treespeciesin a SoutheastAsianrainforest

M. Blate, David R. Peart and Mark Leighton


Geoffrey

Blate,G. M., Peart,D. R. and Leighton, M. 1998.Post-dispersal


predationon
isolatedseeds:a comparative
studyof40 treespeciesina Southeast
Asianrainforest.
- Oikos 82: 522-538.

Manystudiesofpost-dispersal seedpredation havefocusedon density and distance


dependent mortality,whilerelatively fewhaveexamined thefatesof isolatedseeds.
Yet,scatter-dispersedseeds(sensuHowe)arecommonly deposited singlyor in small
groups.Furthermore, evenin specieswithhighly aggregated seeddistributions, the
fatesof themostwidelydispersed individualsmaybe criticalforrecruitment. We
comparedpredation rateson single,isolatedseeds,among40 speciesof treesin
lowlandtropicalrainforest at GunungPalung,WestKalimantan, Indonesia.Seeds
wereplacedalongfourreplicate transectsand monitored fordamagebypredators,
removaland germination in fourtrials,each lastingat least30 days.Tethering of
seedsdidnotaffect removal rates,indicating thatremovals wereattributable to seed
predatorsandnotmerely to physical disturbancebyanimalsor abioticfactors. After
30 days,mortality dueto seedpredation, averaged overspecies, wasmorethan50%;
amongspecies, predation lossesrangedfrom0 to 100%.Overtherangeofseedsizes
weexamined (0.1 g to 11.6g fresh weight)predation rateswerenegatively associated
withseedsizeandwiththethickness andhardness oftheseedcoat.Lowerpredation
on largerseeds is contrary to theoretical predictions and some priorempirical
findings,and maybe partially explainedby the scarcity of predators capableof
penetratingthephysical defenses oflargeseedswithhardseedcoats.Large,softseeds
withlowpredation ratesmayhavepoornutrition or maybe protected bychemical
defenses.Speciesdiffered greatlyin30-daygermination rates,ranging from0 to47%.
Somespecieswithlowpredation ratesalso had lowgermination rates;theimplica-
tionsfortheoverallriskofpredation duringtheseedstagearediscussed. Predation
rateswerenotassociated withspecies'natural dispersal mode(clumped vsscatter-dis-
persed),
contrary to theoretical predictions.Spinyrats(Maxomys spp.)werethemost
abundantseed eatingrodent.Caged spinyrats avoidedlarge,hard seeds and
preferredsoft,medium sizedseeds.The substantial ratesofpostdispersal predation
on isolatedseeds thatwe measuredmay be sufficient to influence strongly the
populationdynamics andlifehistory evolutionoftreesinthisrainforest community.

G. M. Blate, Yale School of Forestryand Environmental


Studies,New Haven, CT,
USA (presentaddress: TropicalForest Foundation,225 ReinekersLane, Suite 770,
Alexandria,VA 22314, USA [tff@igc.apc.orgj).- D. R. Peart, Dept of Biological
Sciences, DartmouthCollege, Hanover,NH 03755, USA. - M. Leighton,Dept of
Anthropology, Peabody Museum,Harvard Univ.,Cambridge,MA, USA.

Seed predationis a major source of mortalityin rain have an important role in population dynamics
forest trees. Pre-dispersal losses may reach 100% (Harper 1977) and natural selection(Janzen 1971); it
(Janzen 1969) and post-dispersalseed mortalityoften may also influencecommunitystructure(Clark and
exceeds 75% (Howe et al. 1985, Schupp 1988a, b), Clark 1984) and contributeto the maintenanceof
reaching 100% for some species (Chapman 1989). species diversity(Janzen 1970, Connell 1971, Grubb
Consequently,seed predationin tropical forestsmay 1977).

Accepted5 January1998
Copyright?) OIKOS 1998
ISSN 0030-1299
Printedin Ireland - all rightsreserved

522 OIKOS 82:3 (1998)

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
Some seed predationstudiesin the tropicshave fo- Studies on individuallyplaced seeds are demanding,
cused on pre-dispersalpredation (e.g. Janzen 1969, because manyseeds mustbe monitoredoverlargeareas
Grieg 1993). Many more(e.g. Wilson and Janzen1972, in orderto obtainestimatesof predationrates.But they
Augspurger1981, 1983, Clark and Clark 1984, Becker are necessaryto investigatethe seed dynamicsof the
and Wong 1985, Coates-Estrada and Estrada 1988, many scatter-dispersed taxa that occur in rain forest
Schupp and Frost 1989, Schupp 1992, Bustamanteet communities(Howe 1989), as well as to quantifythe
al. 1993,Howe 1993,Terborghet al. 1993,Forgetet al. predationriskto isolated seeds in the "tails" of highly
1994) have been promptedby the Janzen-Connellhy- aggregated distributions.Three studies in Panama
pothesis(Janzen 1970, Connell 1971), which proposes (Sork 1987, Schupp 1988a, b, Forget and Milleron
thatpredationriskincreaseswithnonspecific seed den- 1991) examinedthe effectsof post-dispersalpredation
sityand/orthe proximityto parenttrees.Tests of the by vertebrates on isolated seeds of a singletreespecies,
Janzen-Connellhypothesishave naturallyfocused on i.e. at controlledor measured densitiesof < 1 m-2.
evaluatingthe relationshipbetweenpredationriskand Hammond (1995) measuredpredationrateson seeds of
eitherthe local densityof seeds or the distance from fourspeciesof treesin Mexico, and foundlowerpreda-
adult plants. However, the extremecase of very low tionrateson singlyplaced seeds thanon seeds placed in
density,effectively isolated seeds has been largelyig- groups of 5 or 10, in matureor late secondaryforest.
nored.Accordingto Janzen'searly(1970) formulation, However,to our knowledge,therehas been no commu-
theriskof predationwas assumedto fallalmostto zero nity-levelassessmentof predationon isolated seeds for
for the most widelydispersedseeds, i.e. those at the any plant community.A community-level approach
fringesof the seed distribution. allows us to evaluate the patternsof seed predation
Speciesdifferin thespatialpatternsof theirdispersed among species withwidelyvaryingseed morphologies
seeds. Far fromthe parentplant, the densityof wind- and sizes, and among larger taxonomic groupings
dispersedseeds declinescontinuouslywithdistance.In above the species level. Such comparisonsare more
contrast,some animal-dispersed seeds may be scattered
whenthe speciesco-occur,are subject
easilyinterpreted
individuallyor depositedin clumps,generallyfarfrom
to predationfromthe same fauna,and evaluatedusing
the parenttree (Howe 1989). Patternsof mortalityfor
the same methods of experimentation and
scatteredseeds may be quite different fromthose dis-
measurement.
persedin largerclumps(Whelan et al. 1991), especially
Rain forestsin SoutheastAsia are characterizedby
if densitydependentpost-dispersalseed predationis a
interspecificallysynchronizedfruitingevents (masts),
major source of mortality.We suggestthat assessing
occurringat intervalsof ca 5-7 yr (Whitmore1984).
predationriskforisolatedseeds maybe veryimportant,
Most of the time, i.e. between masts, the seed rain
both for seeds in the outer parts of distributions that
includesmanytaxa of treesand lianas thatgenerallydo
are aggregatedaround parenttrees,and forthose that
not participatein the major mastingevents.Our study
are normallyscatter-dispersed.
If the Janzen-Connell was conductedduringa 10-month, non-masting period.
hypothesisholds, so thatmost
recruitmentoriginatesfrom relativelyisolated seeds, We quantifiedthe rates of loss of dispersedseeds to
this does not imply that seed predation on isolated vertebrateseed predatorson experimentally placed, iso-
seeds is inconsequential.On the contrary,if isolated lated seeds for 40 species of seeds that were available
seeds are themain sourceof recruits, it becomesimpor- duringthe studyperiod. The fatesof seeds weremoni-
tantto evaluatethefactorsthatinfluencetheirsurvival. tored for damage or removal by predators,and for
Seeds of scatter-dispersed taxa may occur singlyon the germination.Germinationstatuswas includedbecause
forestfloor.This can resultfrombeingregurgitated or germinatedseeds are no longer susceptibleto seed
defecatedone seed at a time,as occurs withtoucan or predation.Time to germinationinfluencesseed preda-
hornbillregurgitationof several taxa in the Myristi- tion riskover the seed stage of the lifecyclesimplyby
caceae (Leighton1982,Howe 1989). Alternatively, defe- influencingthe time that seeds are available to
cations containingseveral seeds of a taxon may be predators.
scatteredby successivelayersof vegetationas theyfall The experiments includedboth scatter-dispersed and
from high in the canopy, e.g. gibbon defecations clump-dispersed taxa. We also measuredseed size and
(Thomas 1995). The fate of isolated seeds is clearlyof aspectsof seed morphologyforeach species,and exam-
centralimportanceto the populationdynamicsand life ined the relationshipbetween seed morphologyand
historyof scatter-dispersed species.Speciesdispersedby predationrisk.
animalsin clumps(e.g. seeds consumedby orang-utans Optimal foragingtheory (Charnov 1976) predicts
and deposited in large, multi-seededdefecations)are thatpredatorsshould preferseeds withthe greatestnet
intermediatebetweenthe extremesof scatteredseeds reward.Gross returnsto predatorsincreasewith seed
and seed shadowsconcentratedheavilyaround parents. size. However,net rewardfor a givenseed is likelyto
Even orang-utandefecationsmay sometimesinclude vary widely among potentialpredator species, which
only one or a fewnonspecific seeds when the disperser have differenthandling abilities,dependingon their
has been feedingon a diversityof species. size, strengthand typesof mouthparts.Whiletheverte-

OIKOS 82:3 (1998) 523

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
brate fauna at the studysite is documented(Blundell basins, and were thereforeconsideredspatiallyinde-
1996),the relativecontributions of vertebratespeciesto pendentwith respectto predationeffects.Seeds were
seed predationlosses are unknown.Consequently,we placed singly,and well spaced (see below), to assess
do not propose specifichypothesesfor relationships the predationrisk for isolated seeds. Between 8 and
betweenseed size and the predationrate imposed by 15 species were used on each transectfor each trial.
the predator community.However, for a given seed The specificseed taxa used varied over trialsbecause
size, it is logical to expect that predationrate would of phonological differencesamong the tree species.
decreasewithincreasingthicknessand hardnessof the Only three seed species were available, and used, in
seed coat (i.e. testa plus adherentendocarp). more than one trial. Differencesin crop size among
In addition to the field trials, where seeds were fruitingtrees and lianas also limitedthe number of
exposed to all seed predatorsin the community,we seeds available for each species. Consequently,the
conductedseparate experimentswhere seeds were ex-
numberof seeds per transectfor each species ranged
posed only to the most obvious potentialpredatorsin
from4 to 15 (mean n = 9.1). However,the numberof
the vertebratefauna at the studysite,spinyrats (Max-
species,species identities,and numberof seeds of any
omysspp.); theseseed eatingrodentsare locally abun-
given species, were always the same among the four
dant. Because of their potential importance as
replicate transectswithin any field trial. A total of
predators,we assessed the preferences of spinyrats for
seeds of varying size and morphologies, in cage 1499 seeds was placed in the fourfieldtrials.
experiments.
In summary,we addressedthe followingspecificre- Seed collectionand treatment
search questions: 1. What are the predationrates on All taxa used in theexperiments wereanimal dispersed,
isolated seeds due to terrestrial vertebrateseed preda-and all but one (Lithocarpussp.) had fleshyfruits.Seeds
tors,in the naturalhabitatat the timeof naturalseed werecollectedmainlyin the alluvialbenchand lowland
dispersal?2. What are the relationshipsbetweenpreda- sandstonehabitats,in two ways. First,some seeds were
tion risk and seed morphologicaltraits(seed weight, collectedfromthe forestfloor afternatural dispersal.
and thicknessand hardnessof the seed coat). 3. How Especially in the case of scatter-dispersed seeds, this
do seed morphologicaltraitsinfluencethepreference ofmethodwas ofteninefficient, because of thedifficultyof
the most abundantpotentialpredators,spinyrats? findingsingle seeds. Therefore,many taxa were col-
lected predominantlyfrom fruitsthat had not been
consumedby frugivores, but had fallenbelow fruiting
trees.We ensuredthat the fruitswere ripe and seeds
Methods were mature.Seeds were cleaned, wherenecessary,by
Studysite removingthe arils or pulp with a knife.Those with
signsof damage by invertebrates or fungiwerenot used
The studywas conductedin lowlandtropicalrainforest in the fieldtrials.
at the 15-km2Cabang Panti Research Site, ?13'S,
10007'E,in West Kalimantan,Indonesia in the900-km2 Seed morphological traits
Gunung Palung National Park. Annual rainfallis ca The wet weight(seed plus seed coat) and seed coat
4.5 m, but distinctdry periods occur duringFebruary thicknesswere recordedfor a subsampleof at least 5
and July-August.Several distincthabitat typesoccur randomly selected seeds of each seed taxon. Seed
withinCabang Panti (Leighton 1990); our studysites weight was determinedby
dividing the combined
werelocated on alluvial benchand adjacentsandstone-
weightof all seeds in the subsample(measuredto 0.01
derivedsoils, at 10-100 m elevationa.s.l. These habi-
g) by the numberof seeds. Seed coat thicknesswas
tats support the highestdiversityof tree species at
measuredto 0.05 mm usingVerniercalipers.If all seeds
Cabang Panti (M. Leightonand C. Cannon unpubl.).
in the subsamplecontainedan embryoand endosperm,
the seeds in the sample of that taxon were considered
viable. Seed coat hardnesswas recordedas "soft" if it
Predationrate trials could be penetratedeasily by a fingernail, "hard" if it
Experimentaldesign could not, and "veryhard" if it was difficult to pene-
Four field trials, each lasting a minimumof 30 d, trate with a knife.Seeds and fruitswere drawn and
were conducted between 11 September,1992 and 27 described,and a reference collectionof seeds,fruitsand
June, 1993. In each trial, "seeds" (i.e. the dispersal leaves was made forlateridentificationof thetaxa used
units, including the embryo, endosperm, testa and in the experiments. None of the seeds in the fieldtrials
sometimes adherent endocarp; see Table 1) were had spinesor otherunusual physicaldefense,although
placed along each of four replicatetransects.Tran- seeds of one specieswerecoveredwithdense hairs(see
sects ran along ridge trails separated by drainage Discussion).

524 OIKOS 82:3 (1998)

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
t
000 cd
o
=0)'TC4W)C
t
'I 0 C4 0 >0 C - 0 -0

*-~~~~~~~~~~~r
X0 *_
C> 00 o on VI on o t 00 cl rn m
CD v 't 'I 0

0)~ ~ ~ ~~~0
t
co - i_o o00N - N 0 _ C N 0 N
< -1 --- 1 1- 1?-- 1?? I ? ? ?? ? ? 1? ?

:czlv~ 1
000C( - 00 ~l NlCl0 -Cl C ClN
U0 O Cl C-0 C C 0> N 00 eC l> 0>C C 0

0j~
C C-t IO 4D Crx: 11 n I 14 1
1k 1kn k

QO c00M

00) c

;~~~~~~~~~~~~~r
cd
^)
d
knhmt
_ 1n-
C) _>O O t O
CIOr Nt M t- 0
oo "C O o - rn - -O o
00 r C)
O Ch
OO E 00.0 - Cl~l~lC000 0
Cl O 00 M C O
M l O
Cl Cl - -

cO 0.Cic CO

Cq ') en m C1-4 ft cn C< m 1 r C> 't fn O'rt C1 CIA Nt r m 00 Zt Nt

0)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Z = 4:
? '0? E6 ~ ~~~~~U
~E ~ | 'd~ U U Co n C'
-41Cl0ClCO 6
l-lN0
CO
0
Un n
00
0
cn
r~~~~~~~
00
00
n cn Un
-NO
1t
0c
V)
~ ~~~~t
0 (U
Vz
C
0
-
c U
0
U
0t
~
ZC
0 0
Z

Ctc~~cc~~c~l'Ctcz 't~ Cl
Cl
0
-
Cl~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~3C
l Z~cll Cc
-C-- 0 C C-Z dCl
=~ : 0 'I I

'00-Il Cl~~~~~~~~~~~~~~~~~~~~~~~~~~~~LJ
al L rl ~-
: _ * n b S
R ~~~~~~~~~~~~~~~~~~~~~~~~~~~C SCn sW sscnG(/

OIKOS~~~~~~~~~~~~~~~~~~~C
82:3 (1998
52

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
0)

C Cl
O W m O r 1N
Cl
Cl l00 E 'C r-
Cl >
0l ? a -
O \0 -
0- 0 CD- 00
? n 't 't o00 Tr Cl
rN 4 - 00 N- \O c -o
U AS 666666666666666666666666666o o oo

k C> t < tr C) C) "o cE<t-C1 m C) C) rn C o

t Cn
W~~~C
rn <C)
'00 N Co
Ot CI'
T Cm
Cl 00 m
It
~tC14c
\C Cl NP
M 00 ' 0
Cl N 0?
00 ooo 0
'00 >00
0
4; 66666666oo oo o o -6666 6A666 6o coi oC

V: = N _4N_ N<N
_e
N
N
_t~14 C4 CA C1 (1

oo oo o o o
Ua V\t oA or o-ooo Fi

- C l - C c- - r) "-ItC I-r T-t t Cl4 -n Cl Clr l -


Q
C> ) *- OIIN 0 ) C D ) C) 11 C) r- e t C) w 0 CN C C) 00 C)
(U Or 00 O~ 00 O7 00 N O 00 O 0 000 C0 0 0 (
CO.

0) ?

-o

r~~~~~~~c on
C un Co un cn U U U Ln oo Cono cn n 7F
Con to

Ct 0 o
*E

-.s z l Cll 4

CZ Cd~~~~~~~~~~~~~~~~~~~~~~~~~C
U~ ~ C c-C'Ce$ c- 4'0 $0 '0 '0e- 4 c-c- - -Crru0 Cl '0 - CO>

VD r,4 "t rn kn ~ ~ ~ x
H2 0IO 823(19)0

(z) od m

.9 X 4 U B U X Q *Q A Q < k Cn;;; en Cnv; ; c.S .4 .% W. r O O r

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
trialsat GunungPalung,West Kalimantan,Indonesia.Each ratwas
Table 2. Menus of seeds offeredto spinyratsin preference
offeredMenu A on the second night,and Menu B on the thirdnightaftercapture.For each menu,fourseeds of each taxon
were offeredtogether,at dusk. Each genus was representedby a singlespecies. See textand Table 1 fordetails.

Menu A Menu B

Genus Size Hardness Genus Size Hardness

Friesodelsia small hard Adenanthera small hard


Rourea small soft Artocarpus small soft
Aglaia medium soft Syzygium medium soft
Knema medium hard Santiria medium hard
Diospyros large hard Lithocarpus large hard
Elaeocarpus large v. hard Dracontomelon large v. hard

Seed dispersalsyndromes beforeplacement.Flagging was placed > 1 m above


Seed taxa were classed as scatter-dispersed
or clump- each seed location. Singleseeds at 5 m x 10 m spacing
dispersed,based on (1) theirpatternof occurrenceas did not provide unusual, concentratedresourcesfor
dispersedseed as observedin thisstudy,(2) theirusual predators,and the smallstakesused forrelocationwere
patternof occurrencein 1-M2 seed traps in the same of natural materials.Thus, these seed placementand
studysiteover severalyears(M. Leightonunpubl.) and markingprocedureswere unlikelyto resultin the at-
(3) the usual dispersalagent,as recordedin fieldcen- tractionof seed predatorsto the placementlocations.
suses of vertebratevisits to fruitingtrees (Leighton Although seeds with signs of damage by fungi or
1990). Taxa occurringmainlyas singleseeds afterdis- invertebrateswerenot placed in thefield,all seeds were
persal and known to be dispersed by birds were open to attackby theseagentsafterplacement.
classifiedas scatter-dispersed,while those occurring
mainly in clumps and known to be mainly primate Monitoring
dispersed were classified as clump-dispersed.Taxa All seeds used on a given transectwere placed in the
which did not clearlybelong to eithergroup, or for fieldon the same day (i.e. theyformeda "cohort",for
whichinsufficient informationwas available, were not the purposesof measuringremovalrates). For the first
assignedto eithergroup (Table 1). two weeks, seeds were checkedevery2 d for signs of
predation.Subsequently,remainingseeds werechecked
Seed placement every4-7 d for at least 1 monthin each of the four
trials.At each check date, we recordedseed presence,
For each transect,seeds were placed at 5-m intervals
conditionand germination status.Seeds wereexamined
along each of two lines that ran parallel to, on either
for signs of animal activity(e.g. toothmarksor seed
side, and 5 m away fromthe main transectline. Thus,
fragments)and placed in the followingcategories:nib-
for example,a 500-m trail segmenthad 200 potential
bled (<half-eaten), mostly-eaten(>half consumed)
seed locations (2 linesx 500 m lengthwith 5-m spac-
and eaten (if only fragmentsof the seed or seed coat
ing). Only slopes greaterthan 400 and very densely
could be found).Seeds thatwere absent,but forwhich
vegetatedgaps were excluded fromthe potentialseed
there was no visual evidence of predation, were
placementlocations.
classifiedas removed.
The total densityof seed rain at the researchsite A tethering experiment(see below and Results)indi-
duringnon-mastingperiods is quite low, ca 0.25 seeds cated thatremovalof seeds could be reliablyattributed
m- 2 yr- l (Leighton1990),so thattheadditionof even to seed predators.The above classes were then aggre-
a few seeds can increase local densitiessubstantially. gated into two categoriesforanalysis.Seeds thatwere
However,placementof thewell spaced,individualseeds intactor less thanhalfeaten wereclassed as remaining;
in theexperimental trialsincreasedtotalseed densityby those that were more than half eaten or missingwere
only ca 16% above this natural level, in local areas classed as lost to predation.
along each transect(ca 200 seeds added per 500 x 10-m
area). Tethering experiment
Seeds were assigned randomlyto locations within In trial4, some seeds of each of the 12 speciesin that
each transect,and each seed was droppedfroma height trialwere tethered,to determinewhetherseed removal
of 50 cm, to simulatenaturalseed fall.In the fewcases could reliablybe attributedto seed predation,rather
where it was necessary,a small amount of litterwas than othercauses (e.g. water,litterfallor passing ani-
moved to relocate the seed. The exact location was mals). Equal numbersof seeds of each species were
markedby 2 stakes(ca 15 cm long and 3 mmdiameter), tethered,or left untetheredas controls.Seeds in the
insertedintothe ground10 cm on each side of theseed. controlgroup were treatedas in trials 1-3.
Stakes were untreatedwooden skewersthat had been Tethered seeds were attached by a small drop of
exposed to the air at the fieldsite for severalmonths epoxy glue (Araldite Rapid epoxy adhesive;

OIKOS 82:3 (1998) 527

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
Table3. Predation ratesandmeanmorphological traits by3 ormorespeciesinseedpredation
represented
fortheplantfamilies
trialson 40 speciesofrainforesttreesat GunungPalung,WestKalimantan, Indonesia.Familiesarelistedinorderofincreasing
predation rates.Predation ratecalculated rateindex(MRI, see text).Hardnesswas a classvariableat thespecies
as mortality
levelwith1 = soft,2 = hard,and 3 = veryhard.All meansand standard errorsat thefamilylevelwereobtainedfromspecies
meanvalues.
Family No. species Totalno. Mean seed Mean seedcoat Mean seed Meanpredation
tested seeds weight(g) (mm)
thickness coat hardness rate(MRI)
Burseraceae 3 94 6.64 + 1.25 2.72 + 0.50 2.67 + 0.50 0.15 + 0.06
Lauraceae 3 119 4.5+ 0.51 0.68+ 0.12 1.33+ 0.14 0.28+ 0.05
Annonaceae 6 217 0.49+ 0.06 0.13+ 0.01 1.83+ 0.08 0.53+ 0.07
Myrtaceae 5 174 1.84 + 0.14 0.34 + 0.03 1.00 + 0.00 0.62 + 0.17
Connaraceae 3 101 0.36+ 0.06 0.20+ 0.00 1.00+ 0.00 1.22+ 0.26

Ciba-Geigy) to 2 m of finegauge monofilament nylon replicatetransect,theMRI providesmoreopportunities


line,followingthe methodsof Schupp (1988a; see also for statisticalcomparisonsthan does the CPM.
Whelan et al. 1991). Glue was allowed to dryat least 6 Use of linear regressionon a logarithmicscale to
h beforeplacementin the field.Tetheredseeds were calculate the MRI assumes the daily probabilitythat
placed in the same way as controls,exceptthata tether any remainingseed will be attackeddoes not varyover
line was tied to a 20-30 cm x 1 cm ironwood stake the course of a trial. This is equivalentto assuming
drivencompletelyinto the ground. negativeexponentialdeclinein the numbersof surviv-
ing seeds. In addition, using transectsas replicates
Data analysis requiresthat therebe no consistentdifferences among
Two measures of predationrate were used for each transectsfor seeds of a given species. Plots of the
species.The first,the cumulativeprobabilityof mortal- transformed proportionsvs time were examinedvisu-
ity(CPM) was calculatedas theproportionof the total ally for time trends, and for consistentdifferences
seeds (those initiallyplaced) that had been lost to among the trailsused as transects;theseplots indicated
predation by day n of each trial. CPM provides a no consistenttrendsin eithercase.
simple,intuitivemeasureof the amountof seed preda- Because we foundno significant differencesbetween
tion that had occurredfor each species,at each sam- MRI values fortetheredand untethered seeds,thedata
plingdate. Because of therelativelylow samplesizes on fromtrial4 werepooled withthe data fromtrials1-3
each transect,data were pooled across the four repli- in overall analyses.Finally,because seeds fromonly a
cate transectsto betterestimateCPM on each sampling fewspeciesoccurredin morethan one trial,we did not
date. While suitable for descriptiveresults,CPM is compareindividualspecies'predationratesstatistically.
unsuitable for statisticalcomparisons,which require Rather,statisticalanalyses focusedon comparisonsof
replicateestimatesof predationrates. predation rates (MRI) according to morphological
The secondmeasure,the'mortalityrateindex'(MRI) characteristics:seed weight,seed coat thicknessand
allows forreplicateestimatesof predationrate (one for hardness.In the threecases (see above) wherea species
each transect),but sacrificesthe abilityto followtime was used in more than one trial, mean values were
trends in the cumulative probability of predation obtainedforthose species(over trials)beforestatistical
analyseswere done. Statisticalanalyseswere run using
(whichis possiblewithCPM). For each transect,MRI
SYSTAT software(SYSTAT Inc. 1992).
providesa singleestimateof the mean rate of seed loss
to predationover the first30 d, duringwhichmost of
the seed losses occurred.
For each species,MRI was obtainedby firstcomput-
Feedingtrials
ing the logarithm(base 10) of one plus the proportion
of seeds remaining,foreach day over the 30-d period. Experimentaldesign
Then, the slope of the line formedby plottingthese To determinewhetherthe most common seed eating
transformedproportionsagainst time was obtained, rodentshave preferencesfor particularmorphological
using linear regression.A separateslope was obtained characteristicsof seeds,feedingtrialswereconductedin
for each transectsample,i.e. therewere fourreplicate June 1993 with five spiny rats (Maxomys spp.; see
values foreach species.Finallytheseslopes werescaled below). Rats wereofferedtwelveseed species(fromthe
by multiplyingby - 100 to obtain the estimatesof 40 used in thefieldtrials),chosento representthreesize
MRI for that species. Thus, MRI is an index of the classes and two hardness classes. The six resulting
mean rate of predationloss per day over 30 d, but it is categories were (a) small-soft,(b) small-hard, (c)
not a true probabilityof predation.Because it can be medium-soft, (d) medium-hard,(e) large-hard,and (f)
calculatedas a singlenumberforeach species on each large-very hard. The last categorywas alteredfromthe

528 OIKOS 82:3 (1998)

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
Table 4. Germinationand morphologicaltraitsof seeds of 15 rain foresttree taxa at Gunung Palung, West Kalimantan,
Indonesia 30 d afterplacementin the field.Only genericnames are listed;one species per genus was tested(see Table 1 for
details).Only thosespecieswithsome germinationafter30 d are included;no germination occurredforthe remaining25 species
in the fieldtrials.

Family Genus No. of No. Percent Seed coat Seed coat Seed
seeds germinating germinating thickness(mm) hardness weight(g)

Lauraceae Cryptocarya 94 44 46.8 0.4 1 4


Connaraceae Rourea 41 13 31.7 0.2 1 0.5
Sapotaceae Palaquium 26 8 30.8 0.15 2 1.2
Leguminosae Fordia 40 11 27.5 0.1 1 0.5
Meliaceae Dysoxylum 57 13 22.8 0.1 1 6.3
Sterculiaceae Sterculia 36 7 19.4 0.1 1 0.3
Gnetaceae Gnetum 23 4 17.4 0.3 2 0.8
Myrtaceae Syzygium 174 28 16.1 0.34 1 1.8
Guttiferae Calophylluin 39 4 10.3 0.2 2 1
Annonaceae Friesodelsia 76 6 7.9 0.1 2 0.2
Leguminosae Adenanthera 60 4 6.7 0.1 2 0.2
Burseraceae Santiria 31 2 6.5 0.6 2 0.8
Connaraceae Agalea 27 1 3.7 0.2 1 0.5
Annonaceae Xylopia 31 1 3.2 0.15 2 0.7
Myristicaceae Knema 46 1 2.2 0.1 2 1.7

Sub-total 801 147 18.4 0.2 1.5 1.4


Others 698 0 0
Total 1499 147 9.8 0.2 1.5 1.4

originallyplannedlarge-soft to large-very
hard,because Trials
no speciesof large-softseeds wereavailable in the field On the firstnightof capture,ratsweregiventwo seeds
at the time. "Small" seeds were <0.7 g, "medium" of each of the 12 species (Table 2) so theywould be
seeds were between0.7 g and 3.0 g and "large" seeds familiarwitheveryspeciesused in the trialsto follow.
were > 3.0 g (freshweights,includingseed coat). On thesecondnight,each ratwas givenmenuA and on
The 12 seed specieswererandomlyassignedto one of thethirdnight,menuB. Seeds wereplaced haphazardly
two menus(Table 2). Each ratwas offeredbothmenus, on the papered sides of the tank floors,beforedusk.
one on each of two consecutivenights.Withineach Uneaten and partlyeaten seeds, as well as the seed
menu,fourseeds of each morphologicalcategory(e.g. coats of consumedseeds,were removedand talliedthe
small-hard)were offered.Thus, over two nights,each followingmorning.Betweenthe two feedingtrialsthe
rat was offered48 seeds (2 menusx 6 morphological rats were given 6-12 peanuts and ca 25 ml of water;
thesewere generallypartiallyconsumed.
classes x fourseeds in each class).

Rat captureand caging


Rats were capturedusing heavy duty wire mesh live- Results
traps, of a type commonlyused by rural villagersin
West Kalimantan.The trapswerebaitedwitha mixture Overall rates of seed predationon isolatedseeds
of peanutbutterand dryoats and placed in thefieldat
These rateswerecalculatedas the cumulativeprobabil-
dusk. Two verysimilarspinyrat specieswerecaughtin
ity of mortality(CPM).- Of the 1499 seeds placed, 2%
the traps,M. rajah and M. whiteheadi.We decided to
were removedon the firstday, 21% by day 5, 35% by
include these species in the trialsin the proportionin
day 10,47% by day 20 and 53% by day 30. The average
whichtheywere trapped.The fiveanimals used in the
removalrate over all species (based on species-specific
preferencetrialscomprisedfour M. rajah and one M. means) was 51% after30 d. In each trial,therewas a
whiteheadi.Each rat was weighed,sexed, identifiedto wide range among the species,in both the progressive
species (Payne et al. 1985) and placed in a tank (either ratesof seed removaland in the final(30-d) proportion
glass or wire mesh) measuring1 m x 50 cm x 40 cm. of seeds remaining.At the extremes,for some species,
Half the floorwas coveredwithleaves and the remain- no seeds were removedin 30 d, while for others,all
der coveredwithpaper. A metal can was providedfor seeds were removed(Table 1). Clear physicalevidence
shelter.Walls werecoveredwithwhitepaper to produce of consumption(i.e. partiallyeaten seeds or the seed
uniformilluminationwithinthetanks;lids wereof 1-cm coat residue from seeds completelyconsumed) was
wiremesh. For the feedingtrials,tankswere placed in noted in 26% of seed removal events.Removal rates
an open air hut at the fieldstation. were not significantlydifferent betweentetheredand

OIKOS 82:3 (1998) 529

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
a) untetheredseeds in trial 4 of the field experiment
12 (Friedman'stest,blockedby species,df = 1, P = 0.39).
Evidence of attack by invertebrates was seen on 40 of
10 - the 1499 seeds over the first30 d. Of these40 seeds, 18
had been removedfromtheirplacementlocations by
8 day 30.
Of the familiesrepresentedby 3 or more species,
Burseraceaehad thelowestpredationrate(measuredas
a 4 MRI) and Connaraceae thehighest(Table 3). MRI was
not significantlydifferentbetween scatter-dispersed
2 .. taxa (mean 0.60, n = 23) and clump-dispersedtaxa
(mean 0.57, n = 10; t-test.P > 0.05).
0
0 1 2 3 4 5
Seed coatthickness(mm)
Germination
b)
Some germinationoccurredby day 30 for 15 species
(38% of all species; Table 4). Among these 15 species,
8- the percentageof seeds germinating by day 30 ranged
from2% to 47%. For all 40 species combined,only
9.8% of the seeds placed germinatedby day 30 (Table
6-
4). Afterday 30, germination continued,but at an even
slowerrate. By day 40, overallgerminationwas 10.2%
4 (in the threetrialswhereseeds were monitoredfor at
a) least 40 d), and by day 80, 20.6% of seeds had germi-
nated (in theone trialwhereseeds werefollowedfor80
d).
0-
Soft Hard V. Hard
Relationsamongseed morphological
traits
Seed coathardness
c) There were severalstatisticalrelationshipsamong seed
morphologicaltraits. First, seed coat thicknesswas
positivelycorrelatedwithseed weight(Spearman'srank
4. correlationr = 0.62, P < 0.01; Fig. la). Second, seed
weight differedsignificantly among hardness classes,
withveryhard seeds beingmuchlargerthaneitherhard
or soft seeds (ANOVA, df= 2,37, F= 7.0, P = 0.003;
Tukey's test;Fig. lb). Finally,seed coat thicknesswas
2- greatestin veryhard seeds, intermediate in hard seeds
and lowestin softseeds (ANOVA, df= 2,37, F = 56.6,
P < 0.0001; Tukey's test;Fig. lc).
0

Relationbetweenseed morphological
traitsand
Soft Hard V. Hard
predationrate
Seed coathardness
The effectsof seed morphologicaltraitson the rate of
Fig. 1. Relationsamongmorphological characters
for40 spe-
cies of seedsof rainforesttreesat GunungPalung,West seed loss to predatorswereanalyzedin a linearstatisti-
Kalimantan, Indonesia.(a) Seedweight vs seedcoatthickness cal model, with MRI as the dependentvariable and
(n= 40). Seed weight includestheseedcoat. Seedcoat thick- weight,hardnessand seed coat thicknessas indepen-
nesswas positively correlated withseed weight(Spearman's dent variables, plus all interactionterms. Backward
rankcorrelation r= 0.62,P < 0.01). (b) Seed weightvs seed
coat hardness.Values are means+ 1 S.E. n = 19 for soft, stepwiseregressionwas used to eliminatethose vari-
n = 17forhardandn =4 forveryhardseeds.Seedweight was ables contributing minimallyto total explainedvaria-
different
significantly amonghardness classes(ANOVA,df= tion in MRI. The distributionof residuals indicated
2,37,F= 7.0,P = 0.003).See textfordescription of hardness
classes.(c) Seed coat thicknessvs seed coat hardness.Seed that assumptionsof linear models were not seriously
coat thickness differed amonghardnessclasses violated.The resulting
significantly model retainedall threeindepen-
(ANOVA,df= 2,37,F= 56.6, P < 0.0001). dent variablesand the interactiontermsweightx seed

530 OIKOS 82:3 (1998)

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
Table5. Analysis rateindex(MRI, seetext)as thedependent
tableforthelinearmodelwithseedmortality
ofvariance variable,
as theindependent
seedweightand seedcoat thickness
and seedcoat hardness, variables.
Source Sum-ofsquares DF Mean-square F P
Hardness 0.00131 2 0.00066 4.51010 0.01236
Weight 0.00107 1 0.00107 7.36131 0.00736
Thickness 0.00064 1 0.00064 4.38787 0.03770
Hardnessx Weight 0.00156 2 0.00078 5.3627 0.00552
ThicknessxWeight 0.00115 1 0.00115 7.91951 0.00547
Error 0.02448 168 0.00115

coat thicknessand weightx hardness(Table 5). Over- litterfall.Alternativeexplanations of the similar re-
all, predationrate(measuredas MRI) decreasedsignifi- moval ratesbetweentetheredand untethered seeds (e.g.
cantlyas seed weight,hardnessand seed coat thickness thatattractionof predatorsto theglue or monofilament
increased.Small seeds withsoftand thinseed coats had lines balanced higher removals of untetheredseeds)
the highestremovalratesof all seeds in thetrials,while seem implausible.We conclude that tetheringwas not
large seeds with extremelyhard and thick seed coats necessaryto obtain good estimatesof removalrates.It
were least commonlytaken (Fig. 2). remainspossible that some of the seeds removedby
The interactionsof seed coat hardnessand thickness predatorswerenot immediately killedor consumed,but
withseed weightindicatethatpredatorsresponddiffer- rathercached forfutureuse, in whichcase some could
entlyto thephysicalimpediments of seed coat hardness ultimatelyescape predation.However,seeds cached by
and thickness,for seeds of different sizes. The main rodentsare generallylarge and hard, and such seeds
effectsand the interactionsare apparentin Fig. 2a, 2b, had verylow removalrates(e.g. two speciesof Canar-
and 2c. Relativelyfew large seeds were removed,irre- ium,with0% and 0.05% of seeds removedafter30 d;
spectiveof hardnessand seed coat thickness.Although Table 1, Fig. 3b, 3c). Thus, seed caching probably
seeds withthinand/orsoftseed coats were,on average, contributednegligiblyto the removal of seeds in the
taken rapidly,predationon the seeds withthesemor- fieldtrials.In a substantialfraction( > 1/4)of thecases
phological attributeswas concentratedin the small where seeds were taken, visible evidenceof predation
seeds. In contrastto larger seeds, small seeds had a (parts of seeds or seed coats) was foundat the marked
wide range of removal rates, and that variationwas locations. In the remainingcases, we inferthat seeds
associated withseed coat thicknessand hardness(Fig. wereeitherconsumedentirely, moved beforeconsump-
2a, 2b). Small seeds with hard seed coats generally tion,or leftresiduesthat were not easily recognized.
experiencedlowerpredationratesthansmall seeds with Relativelyfew seeds showed evidence of attack by
softseed coats (Fig. 2a). invertebrate seed predatorsover a 30-d period(only40,
or 2.6% of seeds placed). Of these 40 seeds, 18 (45%)
wereremovedby day 30, similarto the overallremoval
Feedingtrials rate of 53% of all seeds placed. Thus, it appears that
attackby invertebrates may not have a major effecton
In the experiments withcaptiverats,no large seeds (of the likelihoodof removalby vertebratepredators.
eitherthe hard or very hard categories)were eaten;
large seeds were thereforeexcludedfromthe statistical
analysis.There was a significantdifferenceamong the
proportionof seeds eaten in the remainingfour mor- traits
Seed size and morphological
phologicalclasses (Friedman'stest,blockedby ratindi-
vidual, df= 3, P = 0.039; Table 6). The percenteaten The strongrelationsamong the morphologicalcharac-
was similar for soft seeds (both small and medium teristicsof the seeds were not surprising.From normal
sized) and medium sized hard seeds, but small hard allometricrelationships, large seeds would be expected
seeds were eaten in much lower proportions. to have thickerseed coats. Seeds combiningthesetraits
(e.g. Canariumspp., Elaeocarpussp., Table 1) shouldbe
mechanicallymost difficult fora seed predatorto con-
sume.Such speciesdid have low removalrates(Table 1,
Discussion Fig. 2). Finally, very hard seed coats were also very
thick,and wererestricted to the largestspeciesof seeds
Role of predatorsin seed removal
(Fig. 2a, 2c).
The tetheringexperimentindicatedthat the seeds that Althoughcorrelated,the threemorphologicaltraits
were moved afterplacementin the fieldwere actively (seed coat thickness,hardnessand size) werenot totally
removedby animals,ratherthan incidentally disturbed predictablefrom one another, and each contributed
by animal activityor by abiotic factorssuch as rain or significantlyto thestatisticalmodelrelatingmorpholog-

OIKOS 82:3 (1998) 531

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
ical traits to removal rate in the field experiments Table 6. Resultsof seed predatorpreference trialsat Gunung
(Table 5, see also Fig. 2). Some large seeds (e.g. Palung, West Kalimantan, Indonesia. Data are numbersof
seeds in each hardnessand size class eatenby individualspiny
Dysoxlum) had thin,softseed coats while some small rats (Maxomys spp.) over 2 nights.Each rat was provided
seeds (e.g. Santiria) had thick,hard seed coats. Only with 8 seeds of each categoryon each night.Rat no. 5 was
Maxomys whiteheadi.the remainingrats were M. rajah. See
Table 2 forseed speciesoffered.Hard seeds weresignificantly
a) less preferredthan soft seeds (Friedman's test,df= 3, P=
0.039).
2.5 -
0 soft
0 0 ~~~~hard Rat individual Small Medium
_ 1.0- 0. v. hard
- 0 X Microcus Soft Hard Soft Hard
1.5- 3 2 6 6
0
0~~~~~ 2 8 5 8 5
1.0 3 4 0 6 4
4 1 0 1 4
c 0.5 ja 5 8 0 5 4

0.0 I I Totals 24 7 26 23
-1.2 -0.6 0.0 0.6 1.2 Mean 4.8 1.4 5.2 4.6
Log seed weight

one species (Microcus sp.) had unusual structuralfea-


b)
turesof the seed coat (otherthan thicknessand hard-
2.5-
Athin
ness) that appeared likelyto influenceseed predators.
Microcusseeds were covered in dense coarse hairs ca
2.0- 0 thick
A ~~ ~~00 v. thick 1.5 cm long. Predationrate (MRI) for Microcus was
1.5- X Microcus low comparedto the othersmall, softseeds (Fig. 2a);
only 18% of Microcus seeds had been removed by
0~~~~~~~~~~~
1.0 A predators30 d afterplacementin the field(Table 1).
In some cases, the morphologicalcharacteristics af-
fectingpredationrates were consistentat the family
level.Burseraceaeseedsweregenerallylarge,withthick,
C) ~-1.0 A -0.6 hard seed coats (Table 3), traitsassociated with low
0.0 0.6
seed mortalityrate. In contrast,Connaraceae seeds
a
were small,withthin,softseed coats, traitsassociated
1 .5 A withhighmortalityrate (Tables 3 and 5, Fig. 2).

AX AA
Influenceof seed size and morphology
on
predationrate
Log seed
coaeticknes It has been argued several times that larger seeds
should be more susceptibleto vertebratepredation.
Harper et al. (1970) suggestedthat large seeds should
Fi.2.0Rlto-n ewenpeainrt harphlgia be preferentiallytaken,because of the higherrewards
for seed predators.Similarly,the more formalmodels
of optimalforagingtheory(Charnov 1976) predictthat
large seeds should be preferredover small ones, given
equal handlingtime. Sork (1987) noted that although
large seed size facilitatesestablishmentof seedlings
variablesfor40 speciesof seeds of rainforesttreesat Gunung underlow lightconditions,it may also increasethe risk
Palung,West Kalimantan,Indonesia.Predationratemeasured of vertebratepredation.Putz and Appanah (1987) fur-
as mean 30-d mortalityrate index (MRI, see text) over four thersuggestthat large seeds are less likelythan small
replicatetransects.See Table I for sample sizes, Table 5 for seeds to be hiddenby surfacelitterand soil, and thus
statistical tests. (a) Predation rate vs seed weight, symbols
coded by seed coat hardness.(b) Predationratevs seed weight, more apparent to visuallyorientedpredators.Boman
symbolscoded by seed coat thickness.(c) Predationrate vs and Casper (1995) found higherpredation rates on
seed coat thickness,symbolscoded by seed weight.Microcus larger seeds (> 1 cm length)than on seeds < 1 cm
sp. was the only species withunusual morphologicalfeatures
likelyto affectpredators;its seeds were covered with dense lengthin hardwoodforestin the northeastern USA. In
coarse hairs (see text). cage experiments,Adler (1995) found that Central

532 OIKOS 82:3 (1998)

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
a)

1.0I

0.9

0.8

0.7

E 0.6
-5-ficrocus
0.5 - ] \
c +~~~~~~~~~~~~~~~~~~~Gnetum
o -A- ~~~~~~~~~~~~~~~~~~~~~~
-x Dysoxylum
t:04
: 0.4 -+- ad, > ^ * * * ffi Ternstromia mag.
--SyZy gum1
0.3 -o-riesdelsia 2
X Agalea
-

0.2

1 13
0.1
0. I I I I I I

0 1 2 4 5 6 8 10 12 14 16 18 22 24 26 28 30
Timeelapsed
(days)
Fig. 3.

Americanspinyrats (Proechimyssemispinosus)prefer- seeds of different sizes in our cage experiments (Table


entiallyate large seeded species. Empirical studies in 6).
old fields(Mittelbachand Gross 1984) and in deserts It is not surprisingthat large seeds withthick,hard
(Podolsky and Price 1990) also supportthe contention seed coats would generallybe leftby seed predators,
that rodentsgenerallypreferrelativelylarge seeds. because of the difficulty or impossibilityof handling
However, despite theoreticalargumentsand many them.For example,Terborghet al. (1993) foundthat
empiricalresultsto the contrary,seed predatorsin our theverylarge(25-30 g), physicallywell-protected seeds
fieldexperimentpreferredrelativelysmall seeds (seed of Calatola venezuelana(Icacinaceae), sufferedno pre-
weight < 0.5 g freshweight),at least withinthe range dation at all. Our observationssuggestthatthe animals
of seed weightsrepresentedby the 40 species studied that eat such seeds are specialistscapable of dealing
(0.1 g to 11.6 g freshweight).Similarresults,i.e. higher with veryhard seed coats. The tuftedgroundsquirrel
predationrates on small and mediumsized seeds than (Rheithrosciurus macrotis), a large squirrel with ex-
on largeseeds,wereobtainedby Osunkoya(1994), in a tremelypowerfuljaws, readily eats the large, hard
rain foreststudyin northeastern Australia,whereseeds Canariumseeds; almost ail recorded observationsof
of 12 species rangingin mean size from0.06 g to 35.2 tuftedground squirrelsat the researchsite are in or
g (freshweight) were placed in the field. However, below the canopy of fruitingCanariumtrees(G. Blate
Osunkoya's (1994) study differedfrom ours in that and M. Leightonpers. obs.). Yet, even such specialist
seeds of all specieswereplaced together,so thatpreda- seed predatorsdo not impose a highmortalityrate on
tors were presentedwithmany more seeds, and many the taxa with large,hard seeds (Table 1, Fig. 2), per-
alternativeseed sizes to choose among, at each loca- haps because such specialistpredatorsdo not occur at
tion. The preferenceof seed predatorsfor small and high abundances.
medium sized seeds found by Osunkoya (1994) also In contrast,thelow removalratesof largeseeds with
matches the preferenceof spiny rats for small and soft,thin seed coats cannot be attributedto physical
medium sized seeds, when they were presentedwith defenses. Furthermore,it seems unlikely that large

OIKOS 82:3 (1998) 533

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
b)
1.0

0.9

0.8 A I

0.7 -
Cananiucf. apertum

g 0.6 -x+ A denanthera

-\- Knerma
O 0.5 - - Cryptocarya
1
0 \ -\-i-Palaguium
.0 - Co\b-ryntocarya
2
C). o --x--Dysoxvlum
Aft\ x
0

_ o
-a---TeT-t ama
A
0.3 - Fissistigma
- Syzygium 2
0.2

Id
0.t
0.0 -|Ht- -l1-1--
0 2 4 7 8 9 10 11 12 13 14 16 18 20 22 24 26 28 30
Timeinfield(days)
Fig. 3.

seeds would, in general,be visuallymore cryptic,or to be taken by invertebratethan by vertebrateseed


provide less effectivechemical cues than small seeds. predators.Anotherqualificationrelatesto the focusof
The fact that some fairlylarge seeds withno physical this study on isolated seeds. It is possible that the
protection(e.g. Stemonurussp. 2, Cryptocaryaspp.) trendswe found in predationrates,accordingto seed
wereremovedat quite low ratessuggeststhatthesetaxa size and morphology,and accordingto plant family,
may be eitherunrewardingnutritionally, or protected may be differentwhenseeds occur at highdensities,for
by effectivechemicaldefenses.Grubb (1996) presents example near the parent plant in species with highly
data fromrain forestspecies indicatingthat nitrogen aggregatedseed distributions.
concentrationdoes generallydecline with increasing
embryo size. Grubb (1996) also found that nitrogen
concentrationtendedto increasewiththe thicknessof
the seed coat, suggestingthat more nutritionallyre- Germination
wardingseeds may be betterprotectedphysicallyfrom When exploringthe implicationsof seed predationfor
the attacksof seed predators. population dynamics,it is importantto considernot
Generalizationsabout theeffectsof seed size mustbe only the predator-inducedmortalityrate of seeds, but
qualifiedbecause the range of seed sizes we examined, also the lengthof time that seeds are vulnerableto
althoughit coverstwo ordersof magnitude,spans only predators.Seeds of differentspecieshave verydifferent
part of the potentialrange.Metcalfeand Grubb (1995) germinationrates (Table 4). For the seed stage of the
reportedseeds rangingin dryweightfrom5 x 10-6 g lifecycle,both the predationrate and the durationof
to 50 g (i.e. coveringsevenordersof magnitude)in rain the seed stage contributeto the likelihood of seed
forestin Singaporeand Malaysia, exclusiveof orchids predation.Ng (1978) notedthatthe seeds of rainforest
and parasiticplants. It is possible that predationrates canopy trees that germinatedwithin 12 weeks after
may eventuallydecline with decreasingseed size, for dispersalweremorelikelyto escape seed predationthan
verysmall seeds. Verysmall seeds are also more likely those with more delayed germination. Similarly,

534 OIKOS 82:3 (1998)

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
c)

1.0

0.9

0.8 1- * * * . . * * * * *
-u-~~~~~~~~~~~~~~~Canariu
Elacarus
0.7 XX\ aa Adenanthera
-.-Syzygium 4
0.6 Kx Stemonurus mal.
E
-o- Santiri'a
-a-Syz;ygium 3
O
0.5
\*
XXX
(4oV;
Gymnacranthera
X
Stemonurusumb.
t 0.4 -+ Ternstroemia
sp.
- Roureopsis
0.3 - Baccaurea

0.2

0.1

0.0 -
0 1 2 4 5 6 8 10 12 14 16 18 20 22 23 24 25 26 27 28 29 30
Timeinfield(days)
Fig. 3.

Geldenhuys(1993) foundthat the seeds of Podocarpus Clearly, then, germinationrates (and the specific
facaltus,whose hard sclerotestadelayedgermination by conditionsforgermination) interactwithseed predation
up to one year,sufferedmore post-dispersalpredation rates to influencethe population dynamicsof plant
than the rapidlygerminatingseeds of P. latifolius,a species. Similarly,in the process of natural selection,
relativewitha thin,leatheryseed coat. traitsthat influencegerminationare likelyto interact
Althoughlong dormancyis more common in very withthosethatinfluence to seed predation.
susceptibility
small than in large seeds (Harper et al. 1970,Putz and These issueshave apparentlynotbeenexploredin depth,
Appanah 1987,Swaine and Whitmore1988),thisgener- eithertheoretically A comparativestudy,
or empirically.
alization does not apply to the shorttermgermination monitoringthe fateof seeds of manyspeciesup to the
rates found in our study.The large seeds at Gunung germinationstage, and includingseedlingmortalityof
Palung withthick,hard seed coats (e.g. Canariumspp., the early germinantswould be informative in this re-
Ternstroemia spp., Lithocarpusspp.), wereactuallyslow spect. At the population level, studies of individual
to germinate(Table 4, G. Blate pers.obs.). No seeds of variationin germination behavior,losses to seed preda-
thesetaxa germinatedin the presentstudy(afterup to tion and earlyseedlingsurvivalwould shed lighton the
80 d in thefield),eventhoughfewhad been removedby trade-offs involvedin earlyvs late germination.
predators.In contrast,the seeds that did germinate
(Table 4) were mostlyfromtaxa with smaller,softer
seeds,whichalso tendedto have higherpredationrates. Dispersalmode,anti-predator
defensesand
Thus, differencesin germinationrates may tend to
predationrates
reduce differencesin the lifetimeprobabilityof seed
predationbetweenthe taxa with large,physicallywell Howe (1989) predicted that scatter-dispersedseeds
protectedseeds and those with smaller,less protected would tend to germinatein relative isolation from
seeds. thusreducingdensity-dependent
conspecifics, predation

OIKOS 82:3 (1998) 535

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
d)
1.0

0.9

0.8 -u--
~~~~~~~~~~~~~~~~~~~~O
0.7
~ ~ ~ ~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~-Lithocarpus
to ~~~~~~~~~~~~~~~~~~~~~~~~~~
Lansium,
S ---- Coscinium,
g 06 -
Diospyr
71 <--A-, Polyalthia
0.5 -x- $y mygium
5
o -+
Xylopia
0.4 -9-Fordia
X < _ 8 <
0
;, ~~~~~~~~~~~~~~~~~~~Rourea
-0 Friesodelsia 1
0.3 - Sterculia

Artocarpus
0.2

0.1
I

0.0 .-- .. ...--- I I I - I i 1


0 1 2 3 5 7 9 1l 13 17 21 30
Timeinfield(days)

Fig. 3. Survivorshipof seeds in fieldpredationtrialsfor40 species of rain foresttreesat Gunung Palung, West Kalimantan,
Indonesia.Values are proportionsof seeds remainingafter30 d in thefield.Data aggregatedfrom4 replicatetransectsforeach
fieldtrial.See Table 1 forsample sizes and fullspeciesnames. Species listedon figuresare in increasingorderof proportionof
seeds taken by day 30. (a) Trial 1, (b) Trial 2, (c) Trial 3, (d) Trial 4.

risk,and consequentlylesseningselectionpressurefor Conclusion


the evolutionof chemicalor physicaldefenses.Scatter-
The fieldexperiment showedconclusively thatverte-
dispersedseeds should thereforebe poorly protected,
brateseed predators do findand take isolatedseeds
and consequentlydependenton good dispersal (and undernaturalconditions, at thetimeof naturalseed
thus escape fromdensitydependentpredation)to sur- fall.Less thanhalfof theplacedseedsremained after
vive. In contrast,clump-dispersed species(mostlyseeds 30 d. Thus,post-dispersal vertebrateseed predators,
defecatedby primates)would generallygerminatein actingafterdissemination of seedsby animalvectors,
close proximityto conspecifics,resultingin density-de- probablyaccountforsubstantial mortality of isolated
pendentpredationpressure,naturalselectionfordefen- seedsin thisrainforest habitat.The levelsofpost-dis-
sive traits, and thus better developed physical or persalseedpredation we havedocumented forisolated
chemicaldefenses. seedsmaybe sufficient to exerta stronginfluence on
In thisstudy,we did findthatsmall,scatter-dispersed both the population dynamics of trees and natural
seeds that were poorlyprotectedphysically(e.g. all of selection actingon seedtraits.
the Connaraceae, Baccaurea stipulate,and Sterculia The measured predation ratesapplybothto scatter-
dispersedspeciesand to the moreisolatedseeds of
stipulate)had the highestpredationrates.But Artocar-
clump-dispersedspecies.Our resultsdo not fitneatly
pus cf. nitidus,Friesodelsiaspp. and Fissistigmasp., all
withthepatterns predictedbyHowe (1989),suggesting
small clump-dispersedspecies, also lacked significant thatgeneralizations abouttheevolveddefenses ofscat-
physical defensesand sufferedsimilar,high levels of tervs clump-dispersed speciesmaystillbe premature.
mortality. Moreover, both Canarium spp. and Predationrateswere more relatedto the size and
Elaeocarpus stipularis,despite being scatter-dispersed,morphological characteristics
ofseedsthanto theusual
were the most physicallywell-protectedseeds in our dispersalmode (scattervs clumped),forthe 40 seed
trialsand had the lowestpredationrates. taxawe examined.

536 OIKOS 82:3 (1998)

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
Whetherthe trendof lower predationrate with in- Charnov, E. J. 1976. Optimal foraging:the marginalvalue
theorem.- Theor. Popul. Biol. 9: 129-136.
creasingseed size holds across a wider range of seed Clark, D. A. and Clark, D. B. 1984. Spacing dynamicsof a
sizes,or forseeds at highdensity,remainsto be tested. tropicalrain foresttree:evaluationof the Janzen-Connell
However, in communitieswhere seed predationrates model. - Am. Nat. 124: 769-788.
are lower on relativelylarge than on small seeds, irre- Coates-Estrada,R. and Estrada,A. 1988. Frugivoryand seed
dispersal in Cymbopetalum baillonii(Annonaceae) at Los
spectiveof seed density,theremay be interesting impli- Tuxtlas,Mexico. - J. Trop. Ecol. 4: 157-172.
cationsforthe recruitment dynamicsof rainforesttrees Connell,J. H. 1971. On the role of naturalenemiesin prevent-
withdifferent the greater
lifehistorytraits.Specifically, ing competitiveexclusionin some marineanimals and in
rain foresttrees.- In: den Boer, P. J. and Gradwell,G. R.
mortalityimposed by post-dispersalseed predatorson (eds), Dynamics of populations. Proceedingsof the Ad-
smallerseeded treespeciesmay tendto compensatefor vanced StudyInstituteon dynamicsof numbersin popula-
the generallyhighfecundity of the smallseeded species. tions. Ooosterbeek, 1970. Centre for Agricultural
As a result,differencesin seed production between Publishingand Documentation,Wageningen,pp. 298-312.
Forget,P.-M. and Milleron,T. 1991. Evidencefor secondary
largeseeded and smallseeded speciesmaybe somewhat dispersalby rodentsin Panama. - Oecologia 87: 596-599.
attenuatedby the time the progenyreach the seedling Forget,P.-M., Munoz, E. and Leigh, E. G., Jr. 1994. Preda-
stage. This size-specific
predationriskwould alter the tion by rodentsand bruchidbeetles on seeds of Scheelea
Palms on Barro Colorado Island, Panama. - Biotropica
parametersof one of the classic life historytradeoffs 26: 420-426.
(betweenthe alternativesof many small and few large Geldenhuys,C. J. 1993. Reproductivebiologyand population
seeds), and make it considerablyeasier for natural structures of Podocarpusfalcatusand P. latifolius in south-
selectionto favorthe evolutionof large seed size. ern Cape forests.- Bot. J. Linn. Soc. 112: 59-74.
Grieg, N. 1993. Predispersalseed predation on five Piper
species in tropicalrainforest.- Oecologia 93: 412-420.
Acknowledgements - We thanktheCenterforResearchand Grubb, P. J. 1977. The maintenanceof species-richnessin
in
Development Biology,IndonesianInstituteof Science plant communities:the importanceof the regeneration
(LIPI) and the Subdirectorate of Nature Conservation niche. - Biol. Rev. 52: 107-145.
(PHPA) oftheMinistry ofForestry, forpermission toconduct Grubb, P. J. 1996. Rainforestdynamics:the need for new
researchat GunungPalung,and fortheircooperation and paradigms.- In: Edwards,D. S., Booth, W. E. and Choy,
assistance.A. G. Blundell,D. G. Gavin,G. D. PaoliandC. 0. S. C. (eds), Tropical rainforestresearch:currentissues.
Webbcontributed assistance and suggestions in thefield.We Kluwer,Dordrecht,pp. 215-233.
thankP. J.Grubb,V. L. Sorkand C. 0. Webbforcomments Hammond,D. S. 1995. Post-dispersalseed and seedlingmor-
on the manuscript, and J. Dykesforstatistical advice. M. talityof tropicaldry foresttreesaftershiftingagriculture,
Ashtonprovidedhelpfulsuggestions on aspectsof research Chiapas, Mexico. - J. Trop. Ecol. 11: 295-313.
design.Thisresearch was supported byNSF DEB-9520889 to Harper,J. L. 1977. Populationbiologyof plants. - Academic
D. R. Peart;GrantNo. 9.249,Program in Scienceand Tech- Press,London.
nologyCooperation, HumanCapacityDevelopment, Bureau Harper, J. L., Lovell, P. H. and Moore, K. G. 1970. The
forGlobalPrograms, FieldSupport andResearch, USAID, to shapes and sizes of seeds. - Annu. Rev. Ecol. Sys. 1:
D. R. Peartand M. Leighton; andfinancial support fromthe 327-356.
TropicalResourcesInstitute of Yale University. Howe, H. F. 1989. Scatter-and clump-dispersaland seedling
demography:hypothesisand implications.- Oecologia 79:
417-426.
Howe, H. F. 1993. Aspectsof variationin a neotropicalseed
dispersalsystem.- Vegetatio107-108: 149-162.
References Howe, H. F., Schupp, E. W. and Westley,L. C. 1985. Early
Adler, G. H. 1995. Fruit and seed exploitationby Central consequencesof seed dispersalfor a neotropicaltree (Vi-
American spiny rats, Proechimyssemispinosus.- Stud. rola surinamensis)in Panama: effectsof herbivoryand
Neotrop. Fauna Environ.30: 237-244. canopy closure. - J. Trop. Ecol. 6: 259-280.
Augspurger,C. K. 1981. Reproductivesynchrony of a tropical Janzen, D. H. 1969. Seed-eatersversus seed size, number,
shrub: experimentalstudies on effectsof pollinatorsand toxicityand dispersal.- Evolution23: 1-27.
seed predators on Hybanthusprunifolius(Violaceae). - Janzen,D. H. 1970. Herbivoresand thenumberof treespecies
Ecology 62: 775-788. in tropicalforests.- Am. Nat. 104: 501-528.
Augspurger,C. K. 1983. Seed dispersalof the tropicaltree, Janzen,D. H. 1971. Seed predationby animals. - Annu. Rev.
Platypodiumelegans,and the escape of its seedlingsfrom Ecol. Sys. 2: 465-492.
fungalpathogens.- J. Ecol. 71: 759-771. Leighton,M. 1982. Fruit resourcesand patternsof feeding,
Becker,P. and Wong, M. 1985. Seed dispersal,seed predation, spacingand groupingamong sympatricBornean hornbills
and juvenilemortalityof Aglaia sp. (Meliaceae) in lowland (Bucerotidae). - Ph.D. Disseration,Univ. of California,
dipterocarprainforest.- Biotropica 17: 230-237. Davis, CA.
Blundell, A. G. 1996. A species list of mammals for the Leighton, M. 1990. Seed dispersal syndromesof Bornean
GunungPalung National Park,West. Kalimantan,Indone- fruits:the dependenceof plantson birdsand mammalsfor
sia. - Trop. Biodiv. 3: 251-259. seed dispersaland forestregeneration.- Report to Center
Boman,J. S. and Casper, B. B. 1995. Differentialpostdispersal for Research and Developmentin Biology and the Subdi-
seed predationin disturbedand intacttemperateforest.- rectorateof Nature Conservation,Ministryof Forestry,
Am. Midl. Nat. 134: 107-116. Indonesia.
Bustamante,R. O., Grez, A. A., Simonetti,J. A., Vdsquez, R. Metcalfe,D. J. and Grubb, P. J. 1995. Seed mass and light
A. and Walkowiak, A. M. 1993. Antagonisticeffectsof requirementsfor regenerationof Southeast Asian rain
frugivoreson seeds of Cryptocaryaalba (Mol.) Looser forest.- Can. J. Bot. 73: 817-826.
(Lauraceae): consequenceson seedlingrecruitment. - Acta Mittelbach,G. G. and Gross, K. L. 1984. Experimentalstud-
Oecol. 14: 739-745. ies of seed predationin old-fields.- Oecologia 65: 7-13.
Chapman, C. A. 1989. Primate seed dispersal: the fate of Ng, F. S. P. 1978. Strategiesof establishmentin Malayan
dispersedseeds. - Biotropica21: 148-154. foresttrees.- In: Tomlinson.,P. B. and Zimmerman,M.

OIKOS 82:3 (1998) 537

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions
H. (eds),Tropicaltreesas livingsystems.
Cambridge Univ. Sork,V.L. 1987.Effects of predation and lighton seedling
Press,pp. 129-162. establishmentin Gustaviasuperba. - Ecology 68: 1341-
Osunkoya,0. 0. 1994.Postdispersal survivorshipof north 1350.
Queenslandrainforest seedsand fruits: of forest, Swaine,M. D. andWhitmore,
effects of
T. C. 1988.On thedefinition
habitatand species.- Aust.J.Ecol. 19: 52-64. ecologicalspeciesgroupsin tropical - Vegeta-
rainforests.
Payne,J.,Francis,C. M. andPhillipps,K. 1985.A fieldguide tio 5: 81-86.
to the mammalsof Borneo.- The Sabah Societywith SYSTAT Inc. 1992.SYSTAT fortheMacintosh, Version5.2.
WWF Malaysia,Kota Kinabalu,Sabah. - SYSTAT, Inc.,Evanston, IL.
Podolsky,R. H. and Price,M. V. 1990. Patch use by Terborgh, J.,Losos,E., Riley,M. P. and Bolanos-Riley,M.
Dipodomys deserti(Rodentia:Heteromyidae): profitability, 1993.Predationby vertebrates and invertebrateson the
preference,anddepletion dynamics.- Oecologia83:83-90. seedsoffivecanopytreespeciesofan Amazonian -
forest.
Putz,F. E. and Appanah,S. 1987.Buriedseeds,newlydis- Vegetatio 107-108: 375-386.
persedseeds,and the dynamicsof a lowlandforestin Thomas,K. E. 1995.Regeneration
Malaysia.- Biotropica19: 326-333. of twoprimate-dispersed
Schupp,E. W. 1988a.Factorsaffecting seed
post-dispersal rain forest trees, Alangiumjavanicum and Baccaurea
survivalin a tropicalforest.- Oecologia76: 525-530. bracteata.- Thesis, Reed College, Portland,OR.
Schupp,E. W. 1988b.Seedandearlyseedling predationinthe Whelan,C. J.,Willson,M. F., Tuma,C. A. and Souza,P. I.
forestunderstory gaps.- Oikos51: 71-78.
and in treefall 1991.Spatialand temporal patterns of postdispersal
seed
Schupp,E. W. 1992.The Janzen-Connell modelfortropical predation. - Can. J.Bot. 69: 428-436.
treediversity:population and theimportance Whitmore,
implications T. C. 1984.Tropicalrainforests oftheFar East.-
of spatialscale.- Am.Nat. 140:526-530. OxfordUniv.Press,Oxford.
Schupp,E. W. and Frost,E.J.1989.Differential predationof Wilson,D. E. andJanzen, D. H. 1972.Predation on Scheelea
Welfiageorgiiseedsin treefallgaps and theforestunder- palmseedsby Bruchidbeetles:seeddensity and distance
story.- Biotropica 21: 200-203. fromthe parentpalm. - Ecology 53: 954-959.

538 OIKOS 82:3 (1998)

This content downloaded from 128.103.149.52 on Wed, 24 Apr 2013 14:01:00 PM


All use subject to JSTOR Terms and Conditions

Вам также может понравиться