Pre-And Post-Germinationdeterminantsofspatial

Pre-and post-germinationdeterminantsofspatial variationinrecruitmentintheperennialherbHelleborus foetidusL.(Ranunculaceae)

JOSE L.GARRIDO,PEDRO J.REYandCARLOS M.HERRERA*

DepartamentodeBiologíaAnimal,BiologíaVegetalyEcología,UniversidaddeJaén,Jaén,E-23071,Spain,and

*EstaciónBiológicadeDoñana,ConsejoSuperiordeInvestigacionesCientíficas,Apartado1056,Sevilla,E-41080,Spain

Summary

1 Recruitmentinplantpopulationsis amultiphase sequentialprocess.Anintegrated viewoftherelationshipsbetweenthestagesandprocessesinvolvedisthereforeneeded, butfewstudieshaveanalysedthedirectandindirecteffectsconcerned.Wepropose a causalmodeloftheserelationshipsandusestructuralequation modelling(SEM)and pathanalysistoinvestigatethedirectandindirecteffectsofpre-andpost-germination processesonrecruitment.

2 Wecollectedinformationonseedproduction,pre-dispersalseedlosses,seedremoval fromtheground, seedlingemergenceandmortality,andseedlingrecruitmentduring thefirstyearafteremergencefrom1 ×1 mplotscentredonreproductiveindividualsin threegeographically distantpopulationsoftheperennialherbHelleborusfoetidus.

3 Theresultssuggestsomecongruencebetweenthethreepopulationsandindicatethat post-germinationprocessesshape thespatial variation inrecruitmentmore than pre-germinationprocesses.Althoughbetween-plotvariationinrecruitmentwasmostly explainedbyseedlingemergenceandmortality,factors suchasseedproductionand removal stillhavesignificanteffectsonrecruitmentthrough various, and frequently contrasting, pathways.

4 Path analysisisapowerfulanalytical toolforrevealingimportantaspectsofthe recruitmentdynamicsofplant populationsandthefactors determining theirspatial variation.Suchaspects,includingindirecteffects,aredifficulttoquantifywithother,more classical,approachestorecruitment.

Key-words: Helleborusfoetidus,pathanalysis,recruitmentdynamics,seedling,structural equation modelling

Introduction

Recruitmentinplantpopulationsisamultiphase pro- cess,withthetotalnumberofrecruitsandtheirspatial variationdependingonpre- andpost-germination processes.Bothmustbeconsideredifwearetounder- standrecruitmentdynamicsfully:seedproduction,seed lossesduetoherbivoryandseeddissemination deter- minehowmanyseedsmovefromfruitstoasitewhere germination mightoccur,whilepost-germination processes(includingemergenceandseedlingsurvival)

Presentaddressandcorrespondence: JoseL.Garrido, DepartmentodeBotanico, FacultaddeBiología, Universi- dad deSantiago deCompostela,Santiago deCompostela,

15782, ACoruña,Spain().

leadtotheeventualestablishmentofnewindividuals. Establishmentsuccess dependsonthedissemination patternandontheconsequences,intermsofdifferen- tial emergenceandsurvival,ofthesuitabilityofthesite reached(JordanoHerrera 1995;SchuppFuentes

1995;Schupp1995).

Anintegrated viewofrecruitmenthastoconsider thesediverseprocessesactingsequentially, andsuch anapproachhasbeenadoptedinseveralstudiesofthe patternsand dynamics of recruitment(Reid 1989; JordanoHerrera 1995;Clark etal.1998; Dalling etal.1998;Harmsetal.2000;ReyAlcántara2000; Wenny2000;Jordano2001).Manystudieshavefound that recruitmentdynamics areverycomplex:indeed, phenomenasuchasseed–seedlingconflictsanduncou- plingandspatialdiscordance betweenlifestagesseem

tobetherulerather than theexception(Houle1992; Jordano Herrera 1995;SchuppFuentes 1995; Schupp1995;ReyAlcántara2000).However,these studieshaverarelyused samplingorexperimental designsthatadequately explorethepossibilitythat aprocessactingonearlylifestages(e.g.onseedsstill attachedtotheplant)mayhaveindirecteffectsonfinal recruitment(though seeJordanoHerrera 1995). Someprocesses,forexampleseedpredationorseedling emergence,mayhavebothdirectandindirecteffectson recruitment,andweneeddesignsandanalysesthatare abletoevaluatetherelativeimportanceofdirectand indirecteffects,aswell asthetotal contributionto recruitment.Structuralequation modelling andpath analysisarepowerfulanalyticaltools,widely used nowadays toinvestigate directand indirect effects onecologicalandevolutionaryprocesses(Pugesek etal.2003,referencestherein), but theyhaverarely beenusedinstudiesofrecruitmentdynamics (but seeJordanoHerrera1995).Weproposehereaninte- grativemodelofthedeterminantsofrecruitment inplantpopulationsandshowtheusefulnessofpath analysisfor disentanglingthe complexnatureof recruitment.

WeuseHelleborusfoetidusL.(Ranunculaceae),an ant-dispersedperennialherb,asacasestudyandcon- ductourstudyinthreegeographicallydistantpopula- tionslocatedindifferentmountainsystemsinsouthern Spain.Weincludedseedlingemergenceandsurvival aspost-germinationdeterminantsofrecruitmentand pre-dispersalseedlosses,numberofseedsarrivingata siteandpost-dispersalseedremovalaspre-germination factors.Byconsideringmostprocessesandstagesfrom matureseed torecruitedplant,weaimtoexplorethe relativeextentstowhichspatialvariationinrecruitment dependsonpreorpost-germinationdeterminants and ondirectandindirecteffects.

Weaddressthefollowingquestions. (i)Whatisthe relativeimportanceofpre- andpost-germinationdeter- minantsforshapingrecruitment?(ii)Howdoprocesses occurringindifferentlifestagesinterrelatetodetermine recruitment? (iii)Istherespatialconcordancebetween seeddeposition,emergenceandrecruitment(i.e.does thespatialvariationinseeddepositiondeterminerecruit- mentpattern,orisitmainlyaffectedbylaterprocesses? (iv)Istheregeographicalcongruenceintherecruitment patternsandin theprocessesdeterminingthem?

Methods

Helleborusfoetidus,hereafter hellebore,isaperennial herbdistributedinwesternandsouth-westernEurope (WernerEbel1994).IntheIberianPeninsulaittyp- icallyoccursinmountainforests,especiallyinclearings andatforestedges.Eachplantconsistsofonetoseveral rametsthat,inthefield,developaterminalinflorescence afterfourtoseven seasonsofvegetativegrowth, althoughplantsbecomereproductivein3–5yearsunder gardenconditions. Flowersareapocarpous,withone

tofivecarpels(mostcommonlytwoorthree)eachcon- tainingeighttofifteenelaiosome-bearingseedsthatare dispersedbyants.Fruitmaturationandseedshedding takeplaceinJuneandearlyJuly.Flowers,andespecially developingfruits,arefrequentlyconsumedbyrodents (mainlyApodemussylvaticus)andlepidopteranlarvae (Noctuaspp.andTrigonophoraflammea,Noctuidae). Detailsofthefloralbiologyandseeddispersalof H.foetidusin theIberianPeninsulacanbe foundin Herrera etal.(2001,2002a,b),Garridoetal.(2002) andFedrianietal.(2004).Mostoftheseedsgerminate after2yearsofdormancy(Herreraetal.2002a;Garrido

2003),andonly5%ofseedsgerminateafter3years. Seedgerminationandseedlingemergenceoccurfrom lateDecembertoearlyMay,withapeakin March (Garrido2003).Populationdynamicsarestronglyvari- ablewithinandbetweenregions,withsomelargeand continuouspopulationsof thousandsof individuals showingpersistenceformanyyears,whileothermore fragmented populationspersistlocallyforonlyafew decades(Garrido2003).Seedlingrecruitmentthus becomesakeyprocessinpopulationdynamics.

Wemonitoredthreepopulationsinthreedifferent mountainsystemsinsouth-easternSpain(JaénProvince): SierradeMágina (Mediterraneanopenscrubland,

1650ma.s.l., 37°44′N, 3°28′W),SierradeCazorla(Pinus nigra forest, 1270 m a.s.l., 37°56′N, 2°52′W), and SierradeSegura(Mediterraneanopenscrubland,1550 ma.s.l.,38°16′N,2°32′W)(seeGarridoetal.2002for furthervegetationandclimatedetails).We useda regular(systematic)samplingdesignconsistingof100

1 ×1 mplots,10 mapartfromeachother,setalong10 paralleltransectsin eachpopulation.However,asseed- lingemergenceneveroccurredmorethan1 m froman adult(reproductive)hellebore(seeGarrido2003),data analysesconsidered onlythe88plotsthat contained adultplants(n =34inMágina,n =17inCazorla,and n =37in Segura).The1 ×1 m plotswerepermanently markedandmonitored forseedlingemergenceand survivalfromJanuary1999toDecember2001.Ineach plotwequantifiedseedproductionbyanyexistingadult hellebore,fruitlossesduetopre-dispersalherbivory, andseed removalfromthegroundbyantdispersersor rodentpredators.Thus,eachsamplingunitprovidedall theinformationrequiredtolinktheprocessesoccurring fromseedproductiontoseedlingrecruitment.

Seedproduction wasmeasuredbydirectlycounting thenumber ofcarpelsreleasingmature seeds.Pre- dispersallosseswere obtainedbydirectlycountingthe totalnumberofflowersconsumedbyrodentsandcat- erpillars.Toestimateseedremovalontheground, two openPetridisheswereplacedineachplot,eachcon- taining20freshseeds(withelaiosomeattached)and somelittertosimulatenatural conditions. Aprevious studysuggestedthathelleboreseedsbecomemuchless attractive toand/ordifficulttodetectbyantsand rodents after 24 hours (Fedriani etal.2004)and the numberofseedsremainingwasthereforecountedafter

36 hours.

Fig. 2 Structuralequation model derived from the pooled dataforallthreepopulations(n =88).Modelandsymbolsas inFig.1.

Fig.1Structuralequationmodelfor eachpopulation(Cazorla, n =17;Mágina, n =34;Segura, n =37)of theinfluenceof differentprocessesonrecruitment (recruitedseedlings)ofH. foetidus:seedproduction(estimatedasthenumberofcarpels releasingseeds),pre-dispersalherbivory(pre-dispersallosses), seedremovalontheground (post-dispersal removal), emer- gence density(emergedseedlings),andseedlingmortality probability(mortality).Standardizedpathcoefficientssignifi- cantlydifferentfrom0 aremarkedwithasterisks(*P0.05,

**P0.01,***P0.0001).Continuouslinesindicatepositive effectsanddashedlinesnegativeeffects,withthe magnitudeof thecoefficientdetermining theirwidth.

DATA ANALYSES

Thevariablesconsideredintheanalyseswere: the numberofcarpels releasingseedsineachplotin1999(a relativeestimate ofseedproduction,seeabove);pre- dispersallosses(proportionofcarpelseatenbyrodents andcaterpillarsin1999);seedremovalontheground (cumulative proportionofseedsdisappearingfrom Petridishesineachplot,averagedfor1999and2000); seedlingemergence(number emergingineachplotin

2001,i.e.theyearwhenmost1999-producedseedsger- minate;Garrido2003);seedlingmortality (cumulative proportionfor1999 and2000); andrecruitment (cumulative number ofseedlingsafterfirstyearfor

1999 and2000).

Causalrelationshipsbetweenthesevariableswere analysed withastructuralequation model(SEM), usingtheSEPATHModuleofSTATISTICA(Statsoft Inc.2000,Tulsa,USA)andlog-transformeddata.As recruitmentisamultiphase sequential process, we builtamodel(Figs1and2)that explicitlyassumes this sequential structure,rather than considering

alternative non-sequentialmodels(forexample,with directpathsfromseedproductiontoestablishedseedlings). Thehypotheses of causalrelationshipsbetween processesweretestedbyagoodness-of-fittestbetween theobservedcovariancematrixandthatexpectedifthe modelwascorrect.Because,evenaftertransformation, somevariablesdidnotshowa normaldistribution,the discrepancyfunctionusedtofitthemodelwasGLS (generalizedleastsquares),whichispreferredtoothers (e.g.maximumlikelihood)incasesoflackofmulti-

variatenormality (TomerPugesek2003).

OneSEMwasconstructedforeachpopulationand subsequently compared withtheotherpopulations through multigroupcomparison(see Bentler1989; Bishop Schemske1998;Grace 2003)totestthe hypothesisoftotalequalityofpathcoefficientsinthe threepopulations.

Results

Althoughclosetosignificance,thebetween-population comparisonofSEMsdidnotrejectthenullhypothesis oftotalequalityofthepathcoefficients(χ2=52.09,d.f.

=38,P=0.07).Therewasthereforenoreasontolook forbetween-populationdifferencesinpathcoefficients (BishopSchemske1998)andwehavetoassume thatcausalrelationshipsbetweenstagesandprocesses determining therecruitment patternsweresimilarin thethreepopulations(Fig.1).Pooling samplesfrom differentlocalities,which improvedthestatistical powerofthemodel,wasthusstatistically reasonable. Thegoodness-of-fittestsforthemergeddatasetindi- catedthatthemodeladequatelydescribedtherelation- ships between variables (χ2 =7.35,d.f.=8,P=0.5; JoreskogGFI=0.972;McDonaldNoncentralityIndex

=0.95–1.000),thus corroboratingthe underlying hypothesisofcausalrelationshipsbetweenthe processes involvedinrecruitment.

Theanalysisofthepooled data isshowninFig.2 andTable 1.Theseresultsindicatemanysignificant directandindirecteffectsonrecruitment,andcomplex relationshipsbetween the processes involved.Totaleffects werelowerforpre-than post-germination processes, although alltheprocessesconsideredhadasignificant

Table1 Directandtotaleffects ondependent variablesinthestructuralequations model.Significantdirecteffects areshownin boldtypeandmarkedwithasterisks(*P0.05,**P0.01,***P0.0001).Totaleffectscannotbeevaluatedassignificantor notastheyincludeallthepathways,boththosestatisticallysignificantandthosenot.Themodelexplained39%ofthevariance inrecruitment.r2indicatestheproportionofvariationinthedependentvariableexplainedbythemodelineachcase.DE=direct effects,TE=totaleffects

Dependentvariables

SeedproductionPost-dispersal removal EmergenceMortalityRecruitment

Independentvariables / DE / TE / DE / TE / DE / TE / DE / TE / DE / TE
Pre-dispersal losses / −0.21* / −0.21 / – / −0.04 / – / −0.04 / – / −0.01 / – / −0.02
Seedproduction / – / – / 0.18 / 0.18 / 0.23* / 0.18 / – / 0.07 / – / 0.11
Post-dispersal removal / – / – / – / – / −0.23* / −0.23 / – / −0.08 / – / −0.14
Emergence / – / – / – / – / – / – / 0.37*** / 0.37 / 0.66*** / 0.59
Mortality / – / – / – / – / – / – / – / – / −0.19*** / −0.19
r2 / 0.05 / 0.04 / 0.09 / 0.14 / 0.39

effectonrecruitment.Therewasa directpositiveeffect ofemergenceandaweakerdirectnegativeeffectof mortalityonrecruitment.Emergencewasequallyaffected byseedproductionandbyseedremoval,although in oppositedirections.Seedlingmortalitywasdirectlyand positivelyaffectedbyemergence(i.e.densitydependent) andthereforeindirectlybyfactorsaffectingemergence (seedremoval(–)andseedproduction (+)).Seed removalontheground indirectlyaffectedrecruitment bothnegatively,viaitseffectonemergence,andposi- tively,viathepositiveeffectofemergenceonmortality. Thissuggeststhatthenegativeeffectofseedremoval onrecruitmentmaybebuffered byreducingdensity- dependent mortality (indirecteffect=0.02).Theneg- ativeindirecteffectofseedproduction onrecruitment viamortality (indirecteffect=−0.02)wasmorethan compensatedforbyapositiveeffectviaemergence (indirecteffect=0.15)sothatthetotalbalanceofits effectswaspositive(seeTable1).Pre-dispersal losses affectedrecruitmentlessthanseedremoval(totaleffects of–0.02and–0.14,respectively).Finally,seedproduc- tiondidnothaveasignificanteffectonseedremovalon theground.

Discussion

GEOGRAPHICAL VARIATION IN RECRUITMENT PATTERNING

Explorationofgeographical variation inrecruitment patterninghasrarelybeenattempted forasingle species(though seeRooney etal.2000).Althoughthe between-population comparisonofthefittedSEMs fromthe differentlocalitiessuggestedthattherewas geographicalcongruenceinthedeterminants ofthe spatialvariationofrecruitment,these resultsshouldbe interpretedcautiously. Thenumber ofsamplingunits perpopulationwaslowforoptimal SEMcomparison (Bentler1989;BishopSchemske1998)andthetest was,infact,closetosignificance.Inaddition, Fig.1 showsthat path coefficientsinMágina wererather differenttothoseattheothertwolocalities,atleastin

termsoftheeffectofemergenceonmortality andthe effectofpre-dispersallossesonseedproduction.On theother hand, theapparentcongruence mayreflect therelativelysmall geographicalrangestudied(all threelocalitiesareinsouthern Spain):thishelleboreis foundthroughoutWesternEurope andgreaterdiffer- encesindeterminantsofrecruitment mayariseon widergeographical scales.

AN INTEGRATED SEQUENTIAL VIEW OF

RECRUITMENT PATTERNING

Studies of seedling recruitmentcan be approached fromthreemajorperspectives:(i)thedemographicper- spective,focusingontransitionprobabilitiesbetween stages,allowingtheidentificationofcriticallifestages and processes for recruitment(JordanoHerrera

1995;Clarketal.1999;ReyAlcántara2000;Traveset etal.2003);(ii)therecruitmentlimitation perspective, focusedonclarifyingwhether plant populationsand communities areseed-ormicrosite-limited (Eriksson

Ehrlén1992;Clarketal.1999;Turnbull etal.2000; Zobeletal.2000);or(iii)therecruitmentpatterning perspective,whichaddressesthe factorsshapingspatial variationinrecruitment(Houle1992,1994;Jordano& Herrera 1995;WadaRibbens 1997).Data fromthe presentstudyfallintothethirdcategoryandresults should beinterpretedintheseterms. From thisper- spective,ourstudyshowsthatpost-germinationdeter- minants(emergenceandsurvival)hada strongereffect onpatterningofrecruitmentthan pre-germination determinantssuchasthetotalnumberofseedsreach- ingtheground ortheirremoval byrodents andants (seetotaleffectsin Table1).Weoptedforamodelthat realisticallydescribedthemultiphasesequentialnature oftherecruitmentprocess,butalternative modelsare possibleandmightleadtodifferentconclusions.

Post-germinationdeterminantsofrecruitmentpattern

Themostimportantdeterminantofspatialvariationin recruitmentwasthenumberofemergedseedlings,and

thuspre-germination processes(includinggermina- tion)becomeimportantthroughindirecteffects.Hence, thesmallproportionofvariationintheemergencepat- tern(9%)explainedbyseedproductionandremovalon thegroundsuggeststhatgermination andprocesses affectingtheseed bank(notexploredhere)mayhavea strong influenceonrecruitmentpatterning.Inhelle- bore,availabledatasuggestthattheseprocessesentail long-term seedlossesmainlyduetotheactivityof rodentsandpathogensduringthe prolongedseed dormancy ofthisspecies (Garrido2003).

Seedlingmortality wasthesecondmostimportant determinantofrecruitment(seetotaleffectinTable1), andwas directlyinfluencedbyemergencedensity. Studiesshowingdensity-dependentseedlingmortality arefrequentintheliterature(Janzen1970;Augspurger

1983;Watkinson1987;Willsetal.1997;Hanley1998; Kitajima Fenner 2000),and thishasfrequently beenrelatedtopathogens suchasdamping-off fungi (Augspurger Kelly1984;Kitajima Fenner 2000) andbacteria(Pedersenetal.1999).Inthepresentstudy, symptomsofinfectionwere scarce,withdesiccation duringthesummerdroughtbeingthemajorcauseof mortality (Garrido2003).Thus,theobserveddensity dependencemaybeduetocompetitionforwater.The finalconsequenceofthisdensity-dependentmortality was thatthebestsitesforemergenceweretheworstfor survival.

Pre-germinationdeterminantsofrecruitment

Allpre-germinationprocesseshadstatisticallysignifi- cant pathwaystowardsemergence, mortalityand recruitment,indicatingacertainpredictivepowerover thesepatterns.Contrarytootherstudies(seeKitajima

Fenner2000;Jordano2001,andreferencestherein), thepatternofrecruitmentdepended onlyslightlyon thetotalnumberofseedsdepositedinsitu.Infact,there wasnosignificantconcordancebetweenthespatial variation innumberofcarpelsreleasingseedsandthe numberofrecruitedseedlings(Spearman correlation, rs=0.07,P=0.46,n =88),indicating that processes actingafterseeddepositionsubstantiallymodifiedthe

patternofrecruitmentand suggestingthat recruit- mentisstronglysitedependent.Thepotentialofpost- disseminationprocesses,particularlyofseedling-related processes,tomodifythepatternofrecruitment,which hasmainlybeendemonstratedforwoodyplants(e.g. JordanoHerrera 1995;García 1998;Castro 2000; ReyAlcántara2000;Jordano2001),isnotsurprising asseedlingsaremorevulnerabletotheabioticenviron- mentthan seeds(Kitajima Fenner 2000andrefer- encestherein).

Theeffectofremovalofseedsontheground high- lightstheecologicalsignificanceoftheant–seedinter- action. Removal ofseedsfromtheground wasvery highandrapid,averaging76%in36hoursand,asants removehellebore seedsfasterthan rodents (Fedriani etal.2004),wasprobablymore closelyrelated to

secondarydispersalthantoconsumption.Seedremoval hadadirectnegativeeffectonemergenceandanega- tivetotaleffectonrecruitment.Whilstit isnotsurpris- ingthatahighseedremovalleadsto aninsitureduction inemergence,efficientseeddispersal byantsshould havepromotedhighrecruitmentinpatchesawayfrom reproductivehellebores.Ourinitialsystematicsam- plingdidnot, however, detect emergenceorrecruit- mentinanyofthe1 ×1 mplotslackingreproductive hellebores(≥63plotswithoutreproductiveindividuals ineachlocality).Surprisingly therefore wefound no evidenceinfavourofefficientdispersalbyants,andthe overalleffectofseedremovalonbothemergenceand recruitmentwasnegative.Veryfew studieshave demonstratedatotalpositiveeffectofantdispersalon plant fitnessorpopulationrecruitmentand explora- tionslikethosereported here,togetherwithexperi- mentaltestsoftheapparentcausal relationshipsdetected bySEManalyses(Bentler1989; BishopSchemske

1998),areneededtodetermine whethersuchbenefits actuallyexist.

THE USEFULNESS OF PATH ANALYSIS IN RECRUITMEN DYNAMIC STUDIES

Pathanalysisallowedustotreatthedifferentprocesses involvedinrecruitmentdynamicsinanintegrative sequential way.Weweretherefore abletoelucidate complexrelationships betweenprocessessuchasthe differentpathways bywhichseedproductionaffected seedlingrecruitment(positive,viathedirecteffect ofemergenceonrecruitment,ornegative,duetothe dependenceofmortality ondensityofemergedseed- lings).Somerelationshipswere evenmorecomplex: although thetotaleffectofpost-dispersal removalon recruitmentwasnegative,thepositiveresultofreduc- ingdensity-dependent mortality represented acom- pensationpathway(Crawley2000),reducingthesizeof theoveralleffect.Suchtangledrelationships maybe impossibletoelucidateusingmoreclassicalapproaches totheanalysisofrecruitment.

Itisalsoimportanttonotethatthecausalrelation- shipsdemonstratedbypathanalysisneedexperimental validation.Someoftheeffectsshowninourstudyhave alreadyfoundsomeexperimental support.Forexam- ple,experimentalevidencefromourthreepopulations suggests that ant exclusion increases recruitmentin situ, butthattheincreaseisinsignificant(Herreraetal.

2002a),supportingtheobservationthat dispersal, if any,isover veryshortdistances(lessthan1m).Further- more,although pre-dispersal lossesduetoherbivory affectedseedproduction,theyhadaslightnegativeeffect onrecruitment. Herbivore- andpollinator-exclusion experimentsonhelleborehavecorroboratedthatfloral herbivores decrease insituseedlingrecruitmentin thepresenceofeffectivepollination(Herrera etal.

2002a).

Inconclusion,pathanalysishasrevealedimportant aspectsof earlyplant recruitmentdynamics and the

factors determining itsspatial pattern.Though these insightsneedexperimental verification,ourresults highlightthefactthatforafull comprehension of recruitmentprocesses,weneedtouseanalytical tools thatconsiderallstagesandprocesses(andtheirdirect and indirect causalrelationships)inanintegrated way.Wesuggestthataresearchprogram combining monitoringofnaturalrecruitment,pathanalysisand experimentaltestingwouldgreatlyimproveourunder- standingofnaturalrecruitmentprocesses.

Acknowledgements

Theauthors thank JulioM.Alcántaraforproviding usefulcomments.We alsobenefitedfromthegenerous fieldassistanceofEsterOrive.Wethank theJunta Rectora ofSierradeCazorla, SegurayLasVillasand Sierra Mágina NaturalParks forproviding working facilities.Thisstudy wassupportedbygrants PB96-

0856, fromtheDirecciónGeneraldeEnseñanza Superior eInvestigación, Ministerio deEducación y Cultura, and BOS2000-1122-C03-01/03, from the Ministerio deCienciay Tecnología,Spain.

References

Augspurger, C.K.(1983)Offspringrecruitmentaroundtrop- icaltrees:changesincohort distancewithtime.Oikos,40,

189–196.

Augspurger,C.K.Kelly,C.K.(1984)Pathogenmortalityof tropicaltree seedlings:experimentalstudiesoftheeffectsof dispersaldistance,seedling density,andlightconditions. Oecologia,61,211–217.

Bentler, P.M. (1989)EQS Structural Equations Program

Manual.BMDPStatisticalSoftware,LosAngeles.

Bishop,J.G.Schemske,D.W.(1998)Variation inflowering phenologyandits consequencesforlupinescolonizing Mount St.Helens.Ecology,79,534–546.

Castro, J.(2000)Dinámicadelaregeneracióndelospinares autóctonosdePinosilvestre (PinussylvestrisL.var. nevadensisChrist)deSierraNevadaySierradeBaza.PhD thesis.Universidad deGranada,Granada.

Clark,J.S.,Beckage,B.,Camill,P.,Cleveland,B.,Hillerislambers, J.,Lichter,J.etal.(1999)Interpretingrecruitmentlimita- tioninforests.AmericanJournalofBotany,86,1–16.

Clark,J.S.,Macklin,E.Wood,L.(1998)Stagesandspatial scalesofrecruitmentlimitation insouthern Appalachian forests.Ecological Monographs,68,213–235.

Crawley, M.J. (2000)Seedpredatorsand plant population dynamics.Seeds:theEcologyofRegenerationinPlantCom- munities (ed.M.Fenner),pp.167–182.CABInternational, Wallingford.

Dalling,J.W.,Hubell,S.P.Silvera,K.(1998)Seeddispersal, seedlingestablishmentandgappartitioning among tropicalpioneertrees.JournalofEcology,86,674–689.

Eriksson,O. &Ehrlén,J.(1992)Seedandmicrositelimitation ofrecruitmentin plantpopulations. Oecologia,91,360–364. Fedriani,J.M.,Rey,P.J.,Garrido,J.L.,Guitián,J.,Herrera,C.M., Medrano,M.etal.(2004)Geographicalvariation inthe potentialofmiceto constrainanant-seeddispersalmutualism.

Oikos,105,181–191.

García,D.(1998)RegeneraciónNaturalDelEnebro(Juniperus communis L.)enáreasdealtamontaña mediterránea: conectandolaecologíareproductivaconelreclutamiento poblacional.PhDthesis.UniversidaddeGranada,Granada. Garrido,J.L. (2003)Semillas yplántulasdeHelleborus

foetidus(Ranunculaceae):variacióngeográfica, ecologíay evolución.PhDthesis.Universidad deJaén,Jaén.

Garrido,J.L., Rey,P.J., Cerdá, X.Herrera, C.M. (2002) Geographicalvariation indiaspore traits of an ant dis- persedplant(Helleborus foetidus):areantcommunity composition anddiasporetraitscorrelated?Journal of Ecology,90,446–455.

Grace, J.B.(2003) Comparinggroups usingstructural equations.Structural EquationModeling:Applicationsin EcologicalandEvolutionaryBiology(edsB.H. Pugesek, A.Tomer A.vonEye).Cambridge University Press, Cambridge.

Hanley,M.E.(1998) Seedlingherbivory,community com- position andplantlifehistorytraits. PerspectivesinPlant Ecology,Evolutionand Systematics, 1,191–205.

Harms, K.E., Wright, S.J.,Calderón,O.,Hernández,A.& Herre,E.A.(2000)Pervasivedensity-dependent recruit- mentenhancesseedlingdiversityinatropicalforest.Nature,

404,493–495.

Herrera, C.M., Cerdá, X.,García, M.B., Guitián, J., Medrano,M.,Rey,P.J.etal.(2002b)Floral integration, phenotypiccovariancestructureandpollinatorvariationin bumblebee-pollinatedHelleborusfoetidus. Journalof EvolutionaryBiology,15,108–121.

Herrera,C.M.,Medrano,M., Rey,P.J.,Sánchez-Lafuente,A.M., García, M.B. Manzaneda,A.J. (2002a)Non-additive fitnesseffectsofpollinatorsandherbivoressuggestamech- anismforadaptivecorrelatedevolutionofmutualism-and antagonism-relatedplanttraits.ProceedingsoftheNatural AcademyofSciences,99,16823–16828.

Herrera, C.M., Sánchez-Lafuente,A.M., Medrano, M., Guitián, J.,Cerdá, X.Rey,P.J.(2001)Geographical variationinautonomousself-pollinationlevelsunrelatedto pollinatorserviceinHelleborusfoetidus(Ranunculaceae). AmericanJournalofBotany,88,1025–1032.

Houle,G.(1992)Spatialrelationshipbetweenseedandseed- lingabundanceandmortalityinadeciduousforestofnorth- easternNorthAmerica.JournalofEcology,80,99–108.

Houle,G.(1994)Spatiotemporalpatterns inthecomponents ofregenerationoffoursympatrictreespecies:Acerrubrum, A.saccharum,BetulaalleghaniensisandFagusgrandifolia. AustralianJournalofEcology,82,39–53.

Janzen, D.H. (1970)Herbivores and thenumber of tree speciesintropical forests.AmericanNaturalist,104,501–

528.

Jordano,P. (2001)Conectandolaecologíadelareproducción conelreclutamiento poblacionaldeplantasleñosasMedi- terráneas.EcosistemasMediterráneos(edsR.Zamora F.I.Pugnaire),pp.183–211.TextosUniversitarios,Granada CSIC/ AEET.

Jordano,P.Herrera, C.M.(1995)Shufflingtheoffspring: uncouplingandspatialdiscordanceofmultiplestages in vertebrate seeddispersal.Écoscience,2,230–237.

Kitajima, K. Fenner, M. (2000)Ecology of seedling regeneration.Seeds(ed.M.Fenner), pp.331–359.CABI Publishing,Wallingford.

Pedersen,E.A.,Reddy,M.S.Chakravarty,P.(1999)Effect ofthreespecies ofbacteriaondamping-off, rootrot development, andectomycorrhizalcolonizationoflodge- polepineandwhitespruceseedlings.Journal ofForest Pathology,29,123–134.

Pugesek, B.H., Tomer, A.vonEye,A.(2003)Structural EquationModeling:ApplicationsinEcologicalandEvolu- tionaryBiology.Cambridge UniversityPress,Cambridge. Reid,N.(1989)Dispersal ofmistletoesbyhoneyeaters and flowerpeckers: componentsof seeddispersal quality.

Ecology,70,137–145.

Rey,P.J.Alcántara,J.M.(2000)Recruitmentdynamicsofa fleshy-fruitedplant(Oleaeuropaea):connectingpatterns of seed dispersaltoseedlingestablishment.JournalofEcology,

88,622–633.

© 2004 British EcologicalSociety, Journal ofEcology,

93,60–66

Rooney,T.P.,Mccormick,R.J.,Solheim,S.L.Waller,D.M. (2000)Regionalvariation inrecruitmentofhemlockseed- lingsandsaplingsintheUpperGreatLakes,USA.Ecological Applications,10,1119–1132.

SASInstitute (1999)SAS/STAT Software. SASInstitute, Cary,NorthCarolina.

Schupp, E.W.(1995)Seed-seedlingconflicts,habitatchoice, andpatternsofplant recruitment.AmericanJournalof Botany,82,399–409.

Schupp,E.W.Fuentes, M.(1995)Spatialpatternsofseed dispersalandtheunificationofplantpopulationecology. Écoscience,2,267–275.

Tomer,A.Pugesek,B.H.(2003)Guidelinesfortheimple- mentationandpublicationofstructuralequation models. Structural EquationModeling:ApplicationsinEcological andEvolutionaryBiology(edsB.H.Pugesek,A.TomerA.vonEye).Cambridge University Press,Cambridge.

Traveset,A.,Gulias,J.,Riera,N.Mus,M.(2003)Transition probabilitiesfrompollination toestablishmentinarare dioecious shrub species(Rhamnus ludovici-salvatoris)in twohabitats.JournalofEcology,91,427–437.

Turnbull,L.A.,Crawley,M.J. Rees,M.(2000)Areplant populations seed-limited?Areviewofseedsowingexperi- ments.Oikos,88,225.

Wada,N.Ribbens,E.(1997)Japanesemaple(Acerpalma- tumvar.matsumurae,Aceraceae)recruitment patterns: seeds,seedlings,andsaplingsin relationtoconspecificadult neighbours. AmericanJournalofBotany,89,1294–1300.

Watkinson,A.R.(1987)Plantpopulation dynamics.Plant Ecology(ed.M.J.Fenner),pp.359–400.Blackwell Science, Oxford.

Wenny,D.G.(2000)Seeddispersal,seedpredation,andseed- lingrecruitmentofaneotropicalmontane tree.Ecological Monographs,70,331–351.

Werner, K. Ebel, F. (1994)Life-history of theGenus

HelleborusL.(Ranunculaceae).Flora,189,97–130.

Wills,C.,Condit, R., Foster, R.B. Hubbell, S.P.(1997) Strongdensity-anddiversity-related effectshelpinmain- tainingtreespecies diversityinaneotropical forest. Proceedingsof theNational Academy of Sciencesof the USA, 94,1252–1257.

Zobel,M.,Otsus,M.,Liira,J.,Moora,M.Möls,T.(2000) Issmall-scalespeciesrichnesslimitedbyseedavailabilityor micrositeavailability?Ecology,81,3274–3282.