Journalof Ecology (1991), 79, 1013-1030

AND RESPONSE: COMPETITIVE EFFECT AND CORRELATED HIERARCHIES TRAITS THE EARLY STAGES OF COMPETITION

IN

DEBORAH E. GOLDBERG AND KEITH LANDA* Departmentof Biology, University of Michigan,Ann Arbor, MI 48109, U.S.A.

SUMMARY (1) Competitiveabilitycan be compared between species in two ways: effectof differentneighbourspecies on performanceof a single targetspecies or response of differenttarget species to a single neighbourspecies. In a 5-week glasshouse experiment,an additive design was used for all combinationsof seven species as both targetand neighbourspecies to determineif therewere consistenthierarchies in competitiveeffectand/orresponse,what traitsof individualsdeterminedposition in these hierarchies,and whetheror not effectand responsecompetitiveabilitywere related duringthe early stages of competition. (2) Five weeks aftersowing,significant non-linearregressionsof targetbiomass on neighbourdensitywere found for 59% of the forty-nine species combinations and significantlinear regressionson neighbourbiomass were found for 51% of the species combinations.The slopes of these regressionsrepresentper-plantand per-gramcompetitioncoefficients, respectively. in competitiveeffectper plant.Differences (3) Neighbourspecies differedstrongly in effectper gram,response per plant, and response per gram were much weaker. Nevertheless,consistentcompetitivehierarchieswere found for both effectand response on both a per-plantand per-grambasis. (4) Differenttraitsdeterminedposition in the effectand response hierarchies. Neighbour species with larger seed mass and larger maximumpotentialmass had strongerper-plantcompetitiveeffects,whilstneighbourspecies withhighermaximum relativegrowthrates had strongerper-gramcompetitiveeffects.The reverseof this latterpatternwas seen forcompetitiveresponse: targetspecies withlower maximum relativegrowthrates were betterresponse competitors.Mean effectand response competitiveabilityof the seven species were uncorrelatedwitheach other. (5) These differencesin traitsassociated with strongeffectand strongresponse competitiveability emphasize the importanceof distinguishing between them in experimentalstudies,at least duringthe early stages of competition. INTRODUCTION Individual competitiveability can be compared between species in two distinct ways: in theircompetitiveeffector abilityto suppressotherindividualsand in their competitiveresponseor abilityto avoid being suppressed(Jacquard 1968; Goldberg * Presentaddress: Departmentof Biology, Indiana University,Bloomington,IN 47405, U.S.A.

1013

This content downloaded from 104.153.225.98 on Tue, 11 Nov 2014 10:56:39 AM All use subject to JSTOR Terms and Conditions

1014

Competitiveeffectand response

& Werner 1983; Goldberg & Fleetwood 1987). Operationally,these correspondto positionsin rankingsof the reductionin some componentof individualfitnessof a single target species grown with differentneighbourspecies (effectcomparisons between neighbours)and positions in rankingsof reductionin fitnessof different targetspecies in the presence of the same neighbourspecies (response comparisons betweentargets).Positionin both effectand responsehierarchiesmay influencethe long-termpopulation dynamicoutcome of competition(Goldberg 1990). However, because it is verydifficult to quantifypopulation-levelcompetitiveabilityfor longlived organismssuch as most terrestrialplants,most experimentalstudiesfocus on individual-level competitiveabilityand thesenecessarilyentailan explicitcomparison of different neighboursand/ordifferent targets. Despite the recognitionof the distinctionbetween comparisonsof competitive effectand response (Aarssen 1983), discussionsof the natureof competitiveability and traitsthatdeterminecompetitiveabilityrarelytake thisdistinctioninto account (Goldberg 1990). Similarly,many of the studies that compare 'competitiveability' amongspeciesexplicitycompareonlyeffect(e.g. Turkington& Harper 1979; Fonteyn & Mahall 1981; Eissenstat& Caldwell 1988; Gaudet & Keddy 1988) or onlyresponse (e.g. Gross 1984; Keddy 1989), even when the designwas such thatboth could have been compared (e.g. Turkington& Harper 1979; Fonteyn& Mahall 1981; but see Wilson & Keddy 1986; Miller & Werner 1987; Peart 1989; Gurevitchet al. 1990). Consequently,a numberof theimportantcontroversies concerningcompetitiveability in plants (Thompson 1987; Tilman 1988; Thompson & Grime 1988) may simplybe semantic- resultingfromconfusionof thetwotypesof competitiveability(Goldberg 1990). In thispaper, we compare competitiveeffectand response between a group of species to address several importantquestions concerningcompetitiveability: (i) How differentare species in competitiveeffectand in competitiveresponse? Goldberg& Werner(1983) hypothesizedthatper-unitsize competitiveeffectsshould be equivalent, at least withingrowthforms,because of the overall similarityin resources required by plants, but that competitiveresponse would differbetween in theirabilityto toleratelow resourceavailabilitydue species because of differences to the presence of neighbours. If true, this would greatlysimplifyanalyses of competitiveinteractionsin plant communitiesbecause only response to competition fromall neighbourswould need to be compared between species. (ii) If effectsor responsesare not completelyequivalenton eithera per-individual or per-grambasis, how consistentare hierarchiesof effector response competitive abilitybetween species? If consistenthierarchiesoccur, similartraitsshould determine eithereffector response competitiveabilityregardlessof the particularpair of competingspecies. (iii) If consistenthierarchiesoccur,whattraitsdeterminepositionin the hierarchy? A number of traits have been hypothesizedor demonstratedto be related to competitiveability(Grime 1988; Tilman 1988; Keddy 1989). However, few attempts have been made to separate expectedor actual relationshipsbetweenthe effectand the response componentsof competitiveability(see Goldberg 1990). (iv) Are the two componentsof competitiveabilitypositivelyor negativelycorrelated or uncorrelated?That is, do similartraitsdetermineeffectand responseor, if different betweentraitsthatconfercompetitiveeffectand traits,are theretrade-offs competitiveresponsesuch thatthereis a negativecorrelationbetweenthe two types of competitiveability?If negativecorrelationsoccur, it becomes importantto consider how the two interactto determinedominanceand persistencein naturalplant

This content downloaded from 104.153.225.98 on Tue, 11 Nov 2014 10:56:39 AM All use subject to JSTOR Terms and Conditions

D. E.

GOLDBERG AND

K.

1015

LANDA

communitiesand if thereare conditionsunderwhicheffector responsecompetitive abilityis more important(Goldberg 1990). The answersto all these questionsprobablydepend on time scale. For example, several studieshave shownthatcompetitivehierarchieschange over timewithinthe same environment(Connolly,Wayne & Murray1990; Menchaca & Connolly1990), and thereforewhichtraitsdeterminecompetitiveabilitymustdepend on factorssuch as relative sizes or stages of the life cycle of the competingplants. Because our focus is on individualcompetitiveabilityand not the eventual population-dynamic (5-week) experimentbut use outcome of competition,we chose to use a short-term a large numberof species combinations.Therefore,the resultsthroughoutapply only to the initialstages of competition. competitive An additiveexperimentaldesignwas used to measure the short-term species, effectsand responsesbetween all pairwisecombinationsof seven different combinations.In an additivedesign,a 'target'species is held includingintraspecific at a constantdensity,whilstdensityof a 'neighbour' species is increased. If the competition, targetspecies is at a densitylow enough that there is no intraspecific the slope of a regressionof mean individualperformanceof the targetspecies on This coefficient can be densityor biomass of neighboursis a competitioncoefficient. interpretedas the per-individualor per-grameffectof the neighbouron the target species or as the response of the target species to one individualor 1 g of the neighbourspecies (Goldberg & Werner 1983). Comparison of differentneighbour species on the same targetspecies gives comparisonsof effectcompetitiveability, targetspecies grownwiththe same neighbourspecies whilstcomparisonof different gives comparisonsof responsecompetitiveability(Fig. 1). Bettereffectcompetitors have steeper negative slopes, whilst better response competitorshave shallower negativeslopes (Fig. 1). Facilitationis indicatedby positiveslopes. METHODS Species and traits The experimentalspecies were all herbaceousplantscommonlyfoundin old fields or pasturesof various ages (Table 1). Seed mass for each species was obtained by (a ) 0

0

E

(b)

~~N3>N2 >N,

T >T

>T3

3

F-

Neighbourdensity FIG. 1. Examples of comparisonof (a) competitiveeffectamong neighbourspecies and (b)

competitiveresponseamongtargetspecies. For comparisonsof competitiveeffect,neighbour species (N) with steeper slopes are bettercompetitors(larger per-individualor per-gram effect).For comparisonsof competitveresponse, targetspecies (T) with shallowerslopes are better competitors(smaller change in performancefor a given change in neighbour densityor biomass).

This content downloaded from 104.153.225.98 on Tue, 11 Nov 2014 10:56:39 AM All use subject to JSTOR Terms and Conditions

Competitiveeffectand response

1016

1. Traitsof species used in the experiment.Life formsare perennial(P) or annual (A). Growthformsare erect grasses (EG), erect dicots (ED) or low-growingdicots (LD). Seed mass is the mean froma minimumof five seed lots of 20-50 seeds each. Median days to seeds per species. Maximumfinalmass and emergencewas calculatedfroma minimumof fifty percentageallocationof total biomass to shoots are means (n = 7) fromplantsgrownwithno neighboursand harvestedat 5 weeks aftersowing.RGRmax. was calculatedas log (maximum finalmass/seedmass)/time,where time= days since sowingor days since median emergence date. TABLE

Species Lolium perenneL. Trifolium pratenseL. T. repensL. Rumex crispusL. Chenopodium album L. Amaranthus retrofiexusL. Phleum pratenseL.

Life form

Seed Growth mass form (mg)

RGRmax. (mgmg-' day-')

Max. finalmass (mg)

sowing

emergence

Median days to emerge

Allocation to shoot (%)

P

EG

2-90

130

0-123

0-105

5.4

56-3

P P

LD LD

1-74 1-68

36 39

0-099 0-097

0-084 0-085

5.5 4-4

68-8 74.5

P

LD

1-28

24

0-137

0-077

16-2

77.5

A

ED

0-64

21

0-155

0 100

12-2

90-2

A

ED

0 56

30

0 148

0.114

7.8

91 9

P

EG

0-35

27

0 165

0 125

81

66.7

seeds each. Emergencetime weighinga minimumof fiveseed lots of twentyto fifty seeds of each was calculated as the median days to emergenceof a minimumof fifty maximum rate potential relative Maximum growth potential species. (RGRmax.), totalmass (root + shoot), and percentageallocationto shootswere determinedfrom the targetplants grown with no neighbours(see below). RGRmax.was calculated as log (final mass/seed mass)/days. Both days between sowing and harvest and betweenmedianemergencetimeand harvestwere used in the denominator.SowingRGRmax.givesa measureof relativegrowthrate over the entireexperimentalperiod allows separationof the influenceof time to emergence while emergence-RGRmax. and of post-emergencegrowthrate on competitiveability. Densitygradientsand harvesting The experimentswere carried out in a glasshouse at the Matthaei Botanical Gardens of the Universityof Michigan.Containers(flats),25.4 cm x 25 4 cm in area soil mix of equal parts peat, sand, x 63 cm deep, were filledwith a nutrient-rich compostedsoil and perlite.Each flatinitiallyhad nine targetindividualsand 0, 8, 16, 32, 64, 128 or 356 neighbourindividuals.This resultedin seven flatsforeach of the species combinations,fora totalof 343 flats.The nine targetindividualsin forty-nine array. Neighbour individuals each flat were arranged in a regular three-by-three were also planted in regulararraysin the low-densityflatsbut were broadcastover the soil surfacefordensities> 16 flat-'. For both targetand neighbourspecies, the actual numberof seeds planted was based on expected percentagegermination.If more thanone seed of the targetspecies germinatedat a givenlocationin each flat,

This content downloaded from 104.153.225.98 on Tue, 11 Nov 2014 10:56:39 AM All use subject to JSTOR Terms and Conditions

D. E.

GOLDBERG AND

K.

LANDA

1017

the seedlings were thinnedto one, leaving the earliest germinatingseedling. No attemptwas made to countor thinthe numberof germinating neighbourindividuals. Locations of targetseeds were marked with toothpicksso that targetindividuals could be distinguishedfrom neighbourindividuals,especially in the intraspecific competitionflats.Flats were wateredonce or twicedaily as needed and no fertilizer was added at any time. The flatswere harvestedafter5 weeks for above-groundbiomass of neighbours, and above- and below-groundbiomass of targetindividuals.Neighbourindividuals were countedas theywere harvested.All plantswere dried at 65 ?C forat least 48 h and weighed. Analysis To compare targetspecies withdifferentmaximumsizes, mean mass per target plant in each flatwas divided by the mean maximummass per plant of that target species (mass in the absence of neighbours;n = 7, see Table 1). In addition,because the values at 0 neighbourdensityforall seven neighbourspecies for a given target species were actuallyreplicatesof the same treatment,these were averaged and the mean used in all the regressionsfor that target species. These two procedures from1 resultedin regressioninterceptsthat were only rarelysignificantly different for (3/49 neighbourdensity,6/49 for neighbourbiomass) and thereforeallowed comparisonsof slopes (competitioncoefficients) thatwere independentof differences in intercepts(performancein the absence of competition). The proportionof maximumtargetmass was regressedagainstneighbourdensity and againstneighbourabove-groundmass, yieldingper-plantand per-gramcompetitioncoefficients, respectively.It is importantto note thattargetplantswere never also includedas neighbours,so thatthereis no potentialforstatisticalconfounding caused by using the same plants as both targetsand neighbours(cf. Mitchell-Olds 1987). Both untransformed regressionsand regressionsusing the reciprocalof proportion of maximumtarget mass were calculated to determinethe form of the relationshipbetween targetsand neighbours.In the transformed regressions,positive slopes indicatenegative(competitive)effects.The reciprocaltransformation is based on a simple formof the reciprocal-yieldequation for intraspecificcompetition(Shinozaki & Kira 1956) and has been used extensivelyin this or modified formto quantifyboth intraspecificand interspecificdensitydependence (Firbank & Watkinson1990). Slopes of the regressionsamong the seven neighbourspecies for each target species (competitiveeffect)and among the seven targetspecies for each neighbour species (competitiveresponse) were comparedwithanalysesof covariance. Pairwise analyses of covariance were then used to test whethereach pair of neighbouror targetspecies differedin competitiveeffector response,respectively.Because there were twenty-one possible pairwisecomparisonsforeach targetor neighbourspecies, an alpha of 0-05/21= 0-0024 was used as a minimumsignificancelevel in the pairwisecomparisonof species. Consistencyof hierarchiesof competitiveeffectsand responses were tested by Kendall's coefficientof concordance (W). Relationships among species traitsand between species traitsand mean effectsor responseswere tested with Pearson correlationcoefficients.The BMDP (Dixon 1983) and SYSTAT (Wilkinson1989) statisticalpackages were used for all analyses.

This content downloaded from 104.153.225.98 on Tue, 11 Nov 2014 10:56:39 AM All use subject to JSTOR Terms and Conditions

1018

Competitiveeffectand response RESULTS Species traits

Of the quantitativetraitsmeasured for each species (Table 1), only seed mass and maximumplant mass were significantly correlatedwith each other (rs = 0-84, P < 0.05). Not surprisingly, species withlargerseeds had largermaximummass at harvest. However, this significantrelationshipdisappeared if Lolium, with much largerseeds and maximumplant mass than any of the otherspecies (Table 1), was excluded (r4= 0 68, P > 0 10). Maximumplantmass was thensignificantly negatively correlatedwith time to emergence (r4= -0-86, P < 0.05), indicatingthat species that took longer to emerge had smaller maximummass at harvest.Furthermore, withoutLolium, seed mass was negativelycorrelatedwithRGRmaX. (fromsowing: r4= -0-96, P< 001; fromemergence:r4= -0 87, P< 0 05). The observationthat species with larger seeds have lower relative growthrates in the absence of any competitionhas been foundin several other studies (e.g. Gross 1984). Percentageallocation to shoot was significantly higherin annuals than perennials (F1,5= 13-13, P1

N

C

+ onN

0

N > 00 ibn m 'IC

~~~~~~~*

^

=

N in 'C m

int

N Ane

O;