From Individuals to Groups and Back : The Evolutionary Implications of Group Phenotypic Composition

Review

From Individuals to Groups and Back: The Evolutionary Implications of Group

Phenotypic Composition

Damien R. Farine,

* Pierre-Olivier Montiglio,

* and Orr Spiegel

*

Thereisincreasinginterestinunderstandingtheprocessesthatmaintainphe- notypicvariationingroups,populations,orcommunities.Recentstudieshave investigatedhowthephenotypiccompositionof groupsoraggregations(e.g., its average phenotype or phenotypicvariance) affects ecological and social processes, and how multi-level selection can drive phenotypic covariance among interacting individuals. However, we argue that these questions are rarelystudied together. We presenta unifiedframework toaddressthisgap, and discuss how group phenotypic composition (GPC) can impact on pro- cessesrangingfrom individual fitnesstopopulationdemography.By empha- sising the breadth of topics affected, we hope to motivate more integrated empiricalstudies oftheecologicalandevolutionaryimplicationsofGPC.

GroupPhenotypicComposition:AnEmergentTopic

Manycentralquestionsinevolutionarybiologyrelyonunderstandinghowindividual-leveland group-level selectiveprocesses interact toshape phenotypicvariation and specialisation.

Individualscanaggregateintogroups,andthecompositionofthesegroups,populations,or communities (herein group phenotypic composition or GPC, see Glossary) can affect group-level dynamics and self-organisation. For example, large groups of highly similar individuals can benefit from lower predation pressure by making it difficult for predators totrackindividuals(theconfusioneffect[1,2]).Eventually,thiscandriveselectionoperating onindividualphenotypes[3,4],inthis case byselecting againstrarephenotypesbecause individualswholookdifferentwillbemorelikelytobetakenbypredators(theoddityeffect [1]).Selection arisingfromGPC canleadtoanevolutionaryresponse ofindividualpheno- types. GPC can shape the evolution of behaviours that mediate individual movements between groups, phenotypic plasticity (individuals adjust their phenotype in response to particularGPCs),orphenotypictraitsthat allowindividualsto manipulatethecharacter- isticsoftheir group.Selection fromGPC canalsodirectly modifythecovariancebetween individual and group phenotypes through the removal of particular phenotypes within generations. All these evolutionary responses at the individual level are likely to modify GPCitself.Despiteextensivetheoreticalconsiderations(see[5]),theroleofGPCasanagent ofselection shaping individualfitness andas anemergent property ofthe individualphe- notypeshave rarelybeen consideredtogetherinempirical studies.

Researchacross arange ofdisparatetopicswill benefit fromsimultaneouslydevelopingan understanding of how GPC affects individual fitness and exerts selection on individual

Trends

Membersofanimalgroupsoftenvaryin their phenotypes (e.g., personality, morphology). Many recent studies haveshownhowdifferentaspects(e.

g., phenotypic average, variation, extreme)ofGPCaffectgroup-levelout- comes(e.g.,foragingsuccess,mating system).

Group-level outcomes can shape selection when individual andgroup phenotypesco-vary.Selectionarising fromGPCcandrivechangesintraits that affect aspects of individual-to- groupcovariance,suchasbehaviours thatdetermine groupmembershipor bychangingtheexpressionoftraits(via phenotypicplasticityorindirectgenetic effects).

Aframeworkbasedoninteractingphe- notypetheorycanquantifytheselec- tive consequences andevaluate the evolutionaryimplicationsofGPC.This framework is stilllargely unexplored empiricallybutitisapplicabletomany topicsinevolutionarybiology.

1DepartmentofAnthropology, UniversityofCaliforniaDavis,Davis, CA,USA

2SmithsonianTropicalResearch Institute,Panamá,Repúblicade Panamá

3EdwardGreyInstituteofField Ornithology,UniversityofOxford, Oxford,UK

4DepartmentofEnvironmentalScience andPolicy,UniversityofCalifornia Davis,Davis,CA,USA

Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-2-1kblblbjemzvx5

Erschienen in: Trends in Ecology & Evolution ; 30 (2015), 10. - S. 609-621 https://dx.doi.org/10.1016/j.tree.2015.07.005

phenotypes,andassessinghowindividualphenotypesrespondtoGPC,ultimatelydrivingan evolutionaryresponsetoselectionarisingfromGPC.Forexample,movinganimalgroupscan containbothleadersandfollowers[6,7].Muchcouldbelearntabouttheevolutionofleadership bysimultaneouslyassessinghowconsistentbehaviouraldifferencesaffectgroup-levelout- comes [8] and how group-level outcomes select for particular phenotypes or shape the distributionofphenotypes inthepopulation.Socialselection[4]andsocial heterosis[9]

offercandidateframeworkstostudyselectionarisingfromthesocialcontext,butthesestillneed to be expanded to capture the complexities that can arise from the interactions between individuals(Figure 1). Tostudy the consequencesofgroup composition,we needto draw onevolutionarytheorywhichintegratesquantitativegeneticsandselection.

Inthis paper weproposethat aframeworkbasedonthe theoryofinteracting phenotypes [4,5,10]canintegratethekeymechanismsdetermininghowGPCaffectsselectionactingon individualphenotypesandthesubsequentevolutionaryimplicationsofGPC.Weusethis

5Authorshaveequalcontribution, orderedalphabetically

*Correspondence:

damien.farine@zoo.ox.ac.uk (Farine,D.R.),makitsimple@gmail.com (Montiglio,P-O.),

orr.spiegel@mail.huji.ac.il(Spiegel,O.)

Phenotypic variaon:

(A) (E) Evoluonary response:

Selecve consequences:

(D) Group composion currencies:

(B)

Group-level outcomes:

Foraging Predator avoid.

Extreme phenotype:

Variance:

Mean

phenotype: vs

Size: Colouraon: Behaviour: Avoidance

Random Assortave ψzy: Group affects individual phenotype

ψzy: Individual phenotype affects group Net selecon

depends on COV_zy

vs vs vs

+

+

??

vs

Phenotype (z_i) Fitness (Wi)

βyz≠ 0 βz≠ 0 βy≠ 0

vs

(i) Natural selecon

(ii) Group-level selecon

(iii) Social selecon Mang syst.

(C)

Figure1.TheEvolutionaryImplicationsofGroupPhenotypicComposition(GPC)inaNutshell.(A)Individuals showremarkablephenotypicvariationintheirmorphology,behaviour,andlifehistory.(B)Hence,groups(orpopulations/

communities)canvaryintheirGPC(e.g.,theirmeanphenotypeorwithingroupvariation).(C)GPCaffectsgroup-level outcomes(e.g.,thetotalamountoffoodagroupacquires),thus(D)impactingonindividualfitness(Box1).Beyondthe consequencesofindividualphenotypesonfitness(naturalselection),GPCcanfavourallmembers(group-levelselection) orfavourparticularphenotypesoverothers(socialselection).Blue(dashed)andblack(solid)linesrepresenttwogroupswith differentGPCs.(E)GPCcandrivedifferentevolutionaryresponses.Theseincludetraitsaffectingcovariationbetween individualphenotypesandtheirGPC(e.g.,decisionstojoinorleaveparticulargroups),theevolutionofindividualphenotypic plasticityinresponsetoGPC(e.g.,individualschangetheirphenotypestomatchthegroup),ortheevolutionofindividual contributionstoGPC(e.g.,individualschangethebehaviourofgroupmembers).Theseevolutionaryresponsescanthen affectthedistributionofphenotypesinsubsequentgenerations.

framework to draw general predictions about how GPC affects the evolution of individual phenotypesandecologicalprocesses,andhowthefeedbackbetweentheevolutionofindividual phenotypesandgroup-leveloutcomesaltersGPC.Wethenshowhowthisframeworkcanbe appliedtoanarrayoffieldsbydetailingempiricalstudiesthathaveinvestigatedtheroleofGPC ineithershapingindividualfitnessorgeneratingemergentpropertiesofanimal groups.We hopethatoutliningthesimilarities amongtopics willhelp tocreate anintegratedresearch agendatobeginaddressingtheeffectsofGPConindividualfitness(selectiveconsequences) and the response of individuals to the selective consequences of GPC (evolutionary implications)inunison.

WhenShouldGPCMatter:AFramework

UnderstandingtheevolutionaryimplicationsofGPCrequiresquantifyingitseffectonindividual fitness,itsnetselectiveforceonindividualphenotypes,anditsevolutionaryimplicationsacross generations(Box1).Thesethreecomponentsareusuallystudiedinisolation,buthereweoutline howtheycanbeintegrated.

VariationinIndividualFitnessResultingfromGPC

Manystudieshave investigatedhowphenotypicvariationamongindividuals isassociated with variation in individual fitness, and how GPC affects the relationship between the phenotype of an individual and their fitness. GPC can be a function of the phenotypes ofitsmembers(suchasthefrequencyoffast-exploringindividuals,orthephenotypeofthe least-exploratoryindividual)[11]oranemergentpropertythatisnotattributabletoanysingle individual, suchas themating system(see section2.2 in[12]). GPCcan affect individual fitnessbyinfluencingtheoverallperformanceofthe grouponcollectivetasks,affecting all themembersofanygivengroupequally,orbyaffectingtherelativeperformanceofdifferent phenotypeswithingroups.Forinstance, agroup withmoreaggressive individualscan be moresuccessfulatforaging,butaggressiveindividualscanhaveahigherfitnessthannon- aggressiveindividualsbecausetheycanmonopolisealargershareofthetotalresources.It isinterestingtonotethat the selectionacting onindividualphenotypeswithingroupscan conflictwiththeselectionactingontheindividualphenotypesamonggroups[13].Studying fitnessinthecontextofphenotypesandhowGPCmodulatesfitness,bothwithinandacross species,providesgoodinsightsontheoverallpotentialselectiveconsequencesofGPCfor theindividuals withinthem.

NetSelectionArisingfromGPC

Inadditiontorequiringphenotypicvariation infitness,netselectionstronglydependsonthe covariancebetweenthephenotypeofanindividualandthephenotypesofthoseitassociates with[4,14].Forexample,ifindividualphenotypesexperienceallpossibletypesofGPCs(i.e., GPCiseffectivelyrandom)thennonetselectionwillarisefromGPC(althoughothermecha- nisms,suchasgeneticdrift,maystillshapeGPC).Akeyinsightisthatarangeofbehavioural traitsandecologicalprocessescandeterminethecovariancepatterninGPC,andhencedrive netselection.Forexample,limiteddispersalcanresultinindividualswithsimilarphenotypes forminggroupsmoreoftenthanrandom.Thus,behaviourscanthereforemodulatethestrength ofselectionarisingfromGPConothertraits[4].Forinstance,aggressivemalewaterstriders (Aquariusremigis)respondtonegativefitnesseffectsofaggressivenesswithintheirgroupby movingbetweengroups[15].Similarly,manyspeciesrespondtocompetitionunderdifferent ecologicalconditionsbyswitchingbetweensingle-speciesgroupsandmixed-speciesgroups [16].

EvolutionaryResponsetoGPC

Extensiveworkonthemechanismsofinheritance,phenotypicplasticity,andindirectgenetic effectshave detailedthe evolutionary response of phenotypes to selection forces (i.e., via

Glossary

Baldwineffect:evolutionin changingenvironmentsfacilitatedby increasedplasticityacross generations.Forexample,individuals thatcanadaptbylearningnew behaviourshavemoreoffspring, resultinginahigheraveragelearning capacityinsubsequentgenerations.

EvolutionaryimplicationsofGPC:

theeffectofGPConthechangesin thedistributionofphenotypes.Such achangecanarisethroughchanges inthegenotypespresentorthrough indirecteffects.

Group-leveloutcome:anyaspect ofagroupofindividualsthataffects thefitnessofindividualmembers.For example,themeanamountoffood thatindividualsinagroupconsume, thegroup'smatingsystem,orthe group'smigrationroute.

Group-levelselection:selection arisingfromdifferencesinthemean fitnessamonggroupsofindividuals.

Groupphenotypiccomposition (GPC):anydescriptorofthetypesof phenotypesfoundwithinagroup.

Examplesincludetheaveragebody sizeofindividuals,thevariationin malecolour,ortheaggressiveness levelofthemostaggressiveindividual inthegroup.Inthisframework,

‘groups’canrepresentmanyaspects ofthesocialenvironmentsof individuals,includingbreedingunits, socialnetworks,neighbourhoods, populations,andcommunities.

Indirecteffects:theeffectofan individualoragroupofindividualson theexpressionofthephenotypeofa givenorganism.Forexample,small individualscangrowfasteriftheyare inagroupcontaininglarger individuals.

Netselectionforce:thenet differencebetweentheaverage phenotypicvalueinthepopulation beforeselectionandtheaverage phenotypicvalueinthepopulation afterselection.

Nicheconstruction:theabilityfor anorganismtomodifyits environmenttofavouritsown phenotype.Forexampleparticular individualscanadaptivelycontrol groupmembershiporalterthe phenotypeofothergroupmembers.

ParameterCOVzy:covariance betweenGPCandindividual phenotypes,ortheextentthat particularphenotypesareconsistently foundeithertogetherorapart.For example,highcovarianceinbody

‘indirecteffects’[5,14]ortheBaldwineffect[17]).Thesetheoreticaltoolsareapplicableto studyingtheevolutionaryimplicationsofGPC(Box1).Individualscanrespondtotheeffectof GPCbyalteringthephenotypiccompositionofthegroup(forexamplebycontrollingaccessto thegroup)and/orbychangingtheir ownphenotype(phenotypicplasticityinresponsetothe

sizeimpliesthatgroupsare composedofeitheralllargeorall smallindividuals.

Parameterby:therelationship betweenthefitnessofanindividual andGPC.

Parameterb^yz:theinteraction betweentheeffectofGPCandthe phenotypeofanindividualontheir fitness.Thisistheextenttowhich theselectiongradientassociatedwith thephenotypeofanindividual changeswithGPC.

Parameterbz:therelationship betweenthephenotypeofan individualandtheirfitness.

ParameterCyz:theeffectofan individualwithaparticularphenotype onGPC.Forinstance,anaggressive individualcanincreasetheoverall numberofaggressiveinteractionsin agroup.

ParameterC^zy:theindirecteffectof GPContheexpressionofthe phenotypeofanindividual.For example,agroupcanconstrainthe maximummovementspeedofits members.

Phenotypictrait:anyaspectofan individualorganismthatcanbe measured.Examplesoftraitsinclude behaviourssuchasaggressiveness, explorationorpatternsofspaceuse, andmorphology(e.g.,bodysizeand colouration).

Selectiongradient:therelationship betweenthefitnessofanindividual andtheirphenotype,orthe relationshipbetweentheirfitnessand theGPCtheyexperience.

SelectiveconsequencesofGPC:

theextenttowhichGPCchangesthe distributionofindividualphenotypes withingroupsowingtodifferencesin fitnesswithinagivengeneration.This canhappenthroughdirectremovalof particularphenotypesortheirfailure toreproduce.

Socialheterosis:anincreasein fitnessofaphenotype(orgenotype) arisingfromitsinteractionwith differentphenotypes(orgenotypes).

Individualswithaparticularallelesor traitvaluehavehigherfitnessinthe presenceofindividualscarryinga differentalleleorexpressinga differenttraitvalue.

Socialselection:differencesin fitnessamongindividualsarisingfrom differencesintheirsocial

environment.Socialselectiondiffers fromnaturalselection,whicharises fromdifferencesinphenotypeamong individuals.Notethatsocialselection arisingfromGPCisusuallyestimated Box1.AFrameworktoStudyGPC

UnderstandingtheselectiveconsequencesandevolutionaryimplicationsofGPCrequiresconsideringinconcert(i)the fitnessconsequencesofindividualandgroupphenotypes,(ii)howselectionaffectsindividualandgroupphenotypic variationwithinageneration,and(iii)howphenotypesaretransmittedacrossgenerations.First,itisimportanttoconsider howgroupcompositioncanaffecttherelationshipbetweenthephenotypeofanindividualandtheirfitness.Thiscanbe donebypartitioningtheeffectsofindividualtraits,groupcompositionandtheirinteractiononfitnessthrougharegression (i.e.,asocialselectionapproach)oracontextualanalysisapproach[3,14]using:

wizibzþy_jbyþziy_jbyz (EquationI)

wherewiisthefitnessoftheindividual(relativetothepopulation),zithephenotypeoftheindividuali,b^ztheselection gradientassociatedwiththisphenotype(i.e.,howanincreaseinonephenotypicunittranslatestoafitnesschange),and yjisthephenotypiccompositionofgroupj(forexampleitsaveragezvalueoranyotherGPCcharacteristic).byisthesocial orgroupselectiongradient(sensu[4]),describinghowchangesingroupcompositionaffectthefitnessofitsmembers.

NotethatsocialselectiongradientsareusuallyestimatedusingtheGPCexcludingthefocalindividual,whereasgroup selectiongradientsareestimatedusingtheoverallGPCofthegroup[3,4].b^yzdescribestheextenttowhichtheselection gradientassociatedwiththephenotypeofanindividualchangeswithGPC.

Second,oneneedstoassesswhethertheselectiongradientsassociatedwithgroupcomposition(byandbyz)truly translatetofitnessdifferencesinthepopulation.Thisisachievedbyquantifyingthecovariancebetweentheindividual phenotypeandGPC:

Sz¼VARzbzþCOVzyby (EquationII)

whereSzisthenetselectiondifferentialactingonindividualphenotypesz,VARzisthephenotypicvariationinzvaluesin thepopulation,andCOVzyisthecovarianceobservedbetweenthephenotypeofindividualsandthecompositionoftheir group[4].

Third,topredicttheresponsetoselection,oneneedstotakeintoaccountthatthegroupcansimultaneouslyaffect,and contributeto,theexpressionofthephenotypesofitsmembers(i.e.,indirectgeneticeffects,[5,18]).Thephenotypeofan individual,zi,isthesumofitsbreedingvalue(theadditiveeffectsofitsgenes),ai,andtheeffectsoftheenvironmentthe individualexperiences, ei [70,71].However, ziisalsodeterminedbythegroup composition.TheparameterCzy

determinesthestrengthanddirectionoftheeffectofyjonthephenotypeoftheindividualzi(see[5]).Similarly,group compositionyj,canalsohaveabreedingvalue,andisaffectedbythephenotypeofitsmembers(throughCyz).Notethat thebreedingvalueofthegroup(aj)isafunctionofthebreedingvaluesofitsmembers.

zi¼aiþeiþCzyyj (EquationIIIA)

y_j¼ajþCyz aiþeiþCzyy_j

(EquationIIIB)

Ifthegroupphenotypeisafunctionofallmembersofthegroup,ratherthanjustone,thenaiwillbethesum,averageof varianceorindividuals’breedingvalues.NotealsothepotentialfeedbackbetweenGPCandindividualphenotypesvia phenotypicplasticity(Czy).Thus,theresponsetoselectionwouldbe:

Dzi¼ GzþCzyGzy

bzþ GzyþCzyGz

by (EquationIVA)

DYj¼ GyþCyzGzy

byþ GzyþCyzGz

bz (EquationIVB)

whereGzistheadditivegeneticvariance(thevarianceinbreedingvalues,ai)associatedwithgroupmembers’phenotype z,Gy,thegeneticvarianceinGPC,andGzyisthegeneticcovariancebetweenindividuals’phenotypeszandgroup compositionY.Iffeedbackisallowed,theevolutionaryresponseofziandYjwillincludeanadditionalterm{1/(1Czy* Cyz)}²(see[5]forderivationandalso[10]).

Integratingthesethreeapproachesintoasinglesetofequationsisnotdevelopedhere(see[10]),buttheseenableusto describethechangeinindividualphenotypes,Dzi,andchangesingroupcomposition,DYj,followingselectionfromone generationtothenext.Thecomplexityofthisequationarisesfromthemultiplecontributionsofindirecteffects(Cyz,Czy) totheresponsetoselection(i.e.,throughtheireffectonthevarianceinindividualphenotype,theircontributionto individualbreedingvalues,andtheirimpact onthecovariancebetween thephenotype ofanindividualandthe compositionoftheirgroup[10].Thisintegratedapproachalsooutlinesthatgroupcompositionhasselectiveconse- quencethroughalimitednumberofinteractingkeyparameters(Gz,Gzy,Gy,Czy,Cyz,bz,by,andbyz).

phenotypiccompositionoftheirgroup).Thestrengthandshapeofsuchindividualresponseswill largelydeterminewhetherselectioncandriveanevolutionaryresponsetoGPC[5,18].Interest- ingly,phenotypescandifferinhowmuchtheycanordocontributetothegroup'soutcome(e.g., bymanipulatingtheGPC)orinhowmuchtheycanordoadjusttheirphenotypetotheirGPC.For example,alargerbodysizeorhigheraggressivenesscouldenableanindividualtoalteritssocial environmenttoagreaterextent.Suchindividualsmightalsoexhibitalowerplasticitylevelthan smallerorlessaggressiveindividuals.

ApplyinganIntegratedFrameworktoStudytheEvolutionofGPC

CombiningtheeffectofGPConindividualfitnesswithananalysisofitsselectiveconsequences andevolutionaryimplicationsallowsustoidentifyfivemainwaysinwhichgroupcompositioncan matterfortheevolutionofindividualphenotypesorforthemaintenanceofphenotypicvariation (Box 1). First, GPC can simply alter the fitness consequences associated with individual phenotypes(i.e., via directselection,by,orbyselection mediatedby thepropertiesoftheir group,byz,Box1).Second,individualphenotypescanco-varywithgroupcomposition(high COV_zy)andthusmodulatethenetselectionforceexertedonindividualphenotypes.Third, GPCcanaffecttheexpression,andthereforetheheritability,ofindividualphenotypes.Hence, GPCcanamplifyorweakentheabilityofthepopulationtorespondtoselectivepressures(high orlowCzy).Fourth,ifGPCmodulatesselection,andindividualphenotypes canrespondto selectivepressures,thiscandriveafeedbackbetweenthem.Asynergybetweentheeffectofthe grouponindividualphenotypes(Czy)andthecontributionofindividualphenotypestothegroup composition(Cyz) cansubstantiallyincreaseordecreasethepaceofevolutionarychangein responsetoselectionarisingfromGPC.Fifth,GPCisalsoanemergentpropertyofgenotypes andtheirpatternsofexpression.Hence,GPCorgroup-leveloutcomescouldalsobeassociated withgeneticvariation,respondtoselection,andbeanadaptation.Theexactfunctionlinking geneticvariation(orvariationinbreedingvalues)attheindividualleveltoGPCatthegrouplevel, f(Gz, Gy), will determine how much genetic variation in group composition is available for selectiontoshapegroup-levelprocesses.Further,interactingphenotypetheorytypicallycon- siders only the group mean, although indirect genetic effects can make the group mean phenotypevery differentthan asimplemean ofthe meanofthe breedingvalues. In cases wheregroup varianceor extremesareimportant, linkingGPCto individual breeding values remainsunclear.Determiningtheamountofgeneticvariationingroupmeanphenotype,inthe phenotypicvariation,orinthephenotypicrangestillrequirespropermathematicalformalisation andisamajorgapinthisarea.

Theintegrated framework(Box 1)provides insightsinto theevolutionaryimpactofGPC by predictingwhenitwillmatteranditsconsequencesforindividualphenotypes.GPCcanonly matterwhenitintroducesdifferences,orequalisesexistingdifferences,infitnessacrossdifferent phenotypes(byandbyzdifferfromzero). Assumingthisisthecase,GPCwillhavestronger selectiveconsequenceswhenthephenotypesofthemembersofagrouporcommunityare non-random(thecovarianceCOVzyishigh).Animportantpredictionhereisthatnetselectionhas thepotentialtobestrongerinsmallergroups(because,bydefinition,COVzydecreaseswith groupsize).Similarly,relatednessamonggroupmembersleadstoahighercovariance(COVzy), andthereforegroupswithhighrelatednessshouldexperiencestrongerselectionarisingfrom GPC.Highercovariancecanalsoarisefromhabitatpreferences,segregation,orlocaladaptation coupledwithlimiteddispersal.HenceweexpectGPCtobeassociatedwithstrongerselectionin thesesituations.NotethatsomeaspectsofGPC(e.g.,theaveragephenotypeofagroup)can co-varywiththephenotypeofitsmembersto agreater extentthanotheraspects(e.g.,the variance,orthehighestphenotypicvalue).

In addition to predicting when GPC should have important selective consequences, this frameworkallowsustopredicttheparticularevolutionaryimplicationsofselectionarisingfrom

byexcludingthecontributionofthe focalindividualtoGPC.

GPC(Figure1).Whenindividualphenotypeisfixed(non-labile),selectionresultingfromGPC shouldresultintheevolutionofbehavioursthatwillaffectthecovariancebetweenthepheno- typesofmembersofthegrouporcommunity (COV_zy),suchas decisionrulesforjoining or leaving,habitatselection,ormanipulationofgroupcomposition byindividuals(e.g.,evicting particularmembers fromthegroups). Ontheotherhand, whentraitsarelabileandcanbe modulated in response to the environment (e.g., individuals can express varying levels of aggressivenessindifferentcontexts),thenweexpectthatselectionassociatedwithGPCwill affecttheevolutionofindividualphenotypicplasticityinresponsetotheirsocialenvironment(Czy) and/ortheevolutionoftheeffectthatindividualshaveonthecharacteristicsoftheirgroup(Cyz, whichcaneventuallyleadto nicheconstruction[19,20]).Thesetraits(CzyandCyz)could evolve[21]todifferamongindividuals[22]andco-varyaccordingtootheraspectsofindividual phenotypes[23].

EmpiricalEvidencefortheSelectiveConsequencesofGPC

Inthefollowingsection weshowhowoutstanding questions inecologyandevolutionfrom apparentlydisconnectedtopicsallrelyonunderstandingtheselectiveconsequencesofGPC.

This frameworkapplies across topics relevant to the ecologyofindividuals ingroups (i–iii), emergentpropertiesofgroups(iv–vi),andcommunityorpopulationdynamics(vii).Wearguethat theseempiricalexamplescanbestudiedbyquantifyingthekeyparametersoutlinedabove(see alsoBox1).

(i)Foraging

Animals can vary in howthey search for and/or handlefood. Variations in foraging skills within groups can create variation in fitness among groups (by) by simply introducing differencesin the amount of food different typesof groups acquire: for example, shoals offishwithbolderormore-exploratoryindividualsfound andapproachednovelfoodfaster [24,25].Similarly,coloniesofsocialspiderswithmore-aggressiveindividualshave ahigher foragingsuccessbecauseaggressiveindividualsapproachpreyfaster [26].Individualscan alsovaryintheabilityorpropensitytoinnovate[27]andsociallylearn[28],henceGPCmight affecttheabilityforgroupstoacquireandexploitnovelforagingopportunities(by).GPCcan alsoaffectthe relativefitnessofmembers withingroups (byz):for example,foodisshared moreequally inmixed-personalityshoals ofguppiescomparedtoshoals withonlyshy or boldindividuals[24].

(ii)Nichepartitioning

GPCcanalsomodulatethebenefits ofgroup-livingbyvaryingcompetition(CzyorCyz)[29].

Manyspeciesparticipateinmixed-speciesforaginggroups,inwhichoneproposedbenefitis thatresourcecompetitionisreducedvianichepartitioning.Reducedcompetitioncanfacilitate largerflocks,resultinginsecondarybenefitssuchasrapidinformationtransfer(byorCOVzy)[30– 33]andbetterproblemsolving[34].Withinspecies,differentindividualsmightalsospecialiseon specificresourcesorvaryintherangeoffooditemstheyeat[35],whichaffectscompetitionfor foodresources[36].Whencompetitionwithingroupsintroducesdifferentialaccesstoresour- ces,(i.e., strongsocial selectioncomponentthrough byz andCOVzy),weexpectgroups to containindividualswithofagreaterrangeoffoodpreferencesorspecialisationsthanwouldbe expectedbychance[29,37].

(iii)Predator–prey(orhost–parasite)interactions

GPCcanaffectphenotype-dependentpredationriskviaconfusionandoddityeffects[1].Thus, GPCcanaffectthefitnessofallgroupmembersequallybycollectivelyreducingtheirpredation risk (confusioneffect, by) or favour particular phenotypes (oddity effect, byz). For example, predationriskforpreyofparticularbodysizesvariesdependingontheratioofsmalltolarge individuals[2],andoddfisharedisproportionatelylikelytoleavetheirshoal(thusaffectingCOV_zy)

[38].Thedefensiveabilityofagivengroupalsovarieswithgroupcomposition:inhoneybee Apis mellifera colonies, the proportion of defensive individuals affects the recruitment of workersagainstintruders,the defensiveresponse ofthe colony,andthe fitness ofallthe colonymembers[39].Diseasedynamicsandparasiteprevalence canalsodependonthe presence of a few highly social or promiscuous individuals in groups acting as super- spreaders[40–42].Forexample,‘TyphoidMary’,ortheflightattendantspreadingthesevere acute respiratory syndrome (SARS) virus between Asia and America, disproportionately contributed to major disease outbreaks. The consequences of group composition for disease dynamicsis usually thought todecrease the fitness of all the group members in asimilarway(by).However,GPCcouldalsoaffectdiseasedynamicsthatdifferentiallyimpact onparticular phenotypes(byz).High levelsof kinship could favourdisease transmission if kinshippromotesphysicalcontactamongkinmembers[43],andnetselectionforcecanbe stronger due to higher phenotypic covariance among group members (high COVzy). By contrast,groupswithhighergeneticvariation(Gz)canhavelowerdiseaseprevalenceifthis reducescontactor transmissionrates.

(iv)AltruismandEusociality

GPCisacentralmechanismintheevolutionofcooperation,altruism, andeusociality[44].

Cooperation amongcolonymembers is more commonincolonies withhigh relatedness, either because it is beneficial for individuals tocooperate with kin(increase their Czy) or because related groups with cooperative individuals outcompete other groups (by and COVzy) [45]). Beyondthese well-established effects, overallcolony fitness candepend on the personality composition of its members. In North American ants Temnothorax long- isinosus, colonies with greater variance in worker aggressiveness and exploration have increased per capita productivity via greater specialisation (by) [46,47]. In the fire ant (Solenopsisinvicta),athresholdfrequencyof‘tolerantworkers’determinestheacceptance ofmultiple queensinto the colony [48].This could potentiallyimpactonthe reproductive outputofthecolony.Similarly,inhoneybees,workersvaryintheirresponsivenesstoqueen pheromone(i.e.,Czy),andless-responsiveworkersaremorelikelytoactivatetheirovaries and competewith the queen over reproduction [49].Hence, worker responsiveness can affecttheextentofcompetitionoverreproductionwithinthecolony(byz).GPCcanevolvein responseto differentenvironments: insocial spiders, whereall individuals reproduceand havenocastespecialisation,colonieswithmore-docileindividualsfairbetterincommunities withless competitors[50](Box 2). GPC can alsobecomeadaptedto specificecological conditions:whenperturbed, colonies ofsocial spidersrecreatethe specific optimalcom- positionoftheir nativeenvironment[51].

(v)CollectiveDecision-MakingandCollectiveMovement

Phenotypiccompositioncanmodulatetheemergentpropertiesofanimalgroups,potentially leadingtopersistentgroup-leveldifferencesinfitness.Complexcollectivebehavioursoften arise fromsimpleinteractionrules, suchasgroups pooling informationtomake accurate decisions[52,53].Havingamixofinformedanduninformedindividualscanpromotebetter decision-makingbecausethelattersupportconsensusviaamajority[54].Informedindivid- uals can also lead groups to resources [55]. In homing pigeons, experienced individuals contributemoretotherouteofthegroupbacktothecoop[56],whileoldfemalematriarch killerwhales(Orcinusorca)leadthegrouptoresourceswhenfoodabundanceislow[57].In greattits(Parusmajor),phenotypicvarianceingroupsisimportantbecauseslow-exploring individualsmaintain groupcohesion, while fast-exploring individuals samplenew patches, thus‘leading’thegroup[8].Thus, whoisinmovinggroupscanaffecthowwellthegroup makesdecisions(throughCyzorby),andpotentiallythefitnessofgroupmembers(z_i).Group compositioncan even have carry-over effectsonbehaviour ifnew individuals acquirethe behavioursofgroupmembersviaconformistorculturaltransmission(Czy)[58].Forexample,

juvenilewhoopingcranes(Grusqmericana)learnthemigrationroutesfromolderindividuals [59],whichcouldleadtoconsistentandmaintaineddifferencesinmigrationcostsbetween groups.

(vi)MatingSystem

Classicalworkontheevolutionofalternativematingtacticsvianegativefrequency-dependent selectionhasextensivelyconsideredtheselectiveconsequencesofthefrequenciesofdifferent phenotypesinapopulationoragroup(byz)(e.g.,evolutionarystablestrategies[60]).However, otheraspectsofGPC,suchasthepresenceorabsenceofparticularphenotypes,canaffect matingsystems.Forexample,thepresenceofaggressiveindividualsingroupsofwaterstriders hasawell-knowneffectontheoperationalsexratioandmatingactivitywithinmatingpools(e.g., [61]).Thepresenceofaggressiveindividualsdrivesotherindividualsoutofthelocalareaand decreasestheintensity ofscramble competitionforfemales(aggressive individualsexhibita strongCyz).Thisreducestheoverallmatingratesofthegroup(by),butunderthesenewsocial conditionsmore-aggressivemalescangetabiggershareofthemating(byz)(whenmeasuredat theindividuallevel[13]).Similarly,GPCislikelytohavestrongimpactongroupreproductive successinfamilygroups,cooperativelybreeders,andmonogamousspecieswithbi-parental care[62,63].

(vii)PopulationandCommunityDynamics

ThelinkbetweenGPCandpopulationleveloutcomesisattheheartofclassicalhypotheses onpopulationregulation,wherepopulationcompositioninaggressivenessorterritorialitycan bothdriveand bedriven bypopulation density (throughbyz andpotentiallyan interaction betweenCzyandCyz)[64].Wearenotawareofanystudyempiricallylinkingpopulationor community composition in behavioural traits to their dynamics, and these dynamics to fluctuationsinselectionactingonbehaviouraltraits.Nevertheless,thedistributionofpartic- ular morphological [65], life-history [66], or behavioural traits [67] can affect the overall population growth of all members (by) or favour particular phenotypes relative to others

Box2.EmpiricalExampleI:GPCinSocialSpiders

Todemonstratehowourframeworkcanbeappliedtoanempiricalstudysystem,considerthecaseofsocialspiders.

Thesehavebeenstudiedextensivelyinthecontextoftheselectiveimportanceofgroupcomposition.Initialstudiesin threespeciesshowedhowcoloniesaremadeupofdocileandaggressiveindividuals(Anelosimusstudiosus[26,72,73];

Stegodyphussarasinorum[74];Stegodyphusdumicola[75]).Studiesrepeatedlyquantifiedindividualbehaviourand trackedtheoverallperformanceofgroupcompositionusingobservationalandexperimentalapproaches.Thesestudies havetackledtheclassicalquestionofhowindividualspecialisationandconsistentbehaviouralvariationcanevolve throughmulti-levelselectionandsocialinteractions.Docileandaggressiveindividualsspecialiseondifferenttasks,such ascolonymaintenance,defence,orforaging[72,76,77].

Manyaspectsofthephenotypiccompositionofacolony(yj)affectitsdynamics.Thefrequencyofaggressiveordocile individualsdeterminestheforagingsuccessordefensiveabilitiesofthecolony(i.e.,by[26]).Inaddition,thephenotypeof individualsattheextremeofthephenotypicdistributionofthecolony[74],andthepersonalityoffounderindividuals[78], haveimportanteffectsontheoverallcolonybehaviour,performance,andthefitnessofindividuals(i.e.,b^y).Theseeffects ofcolonycompositiononperformanceareespeciallylikelytoleadtoselectionpressureswhenthecontributionofan individualtocolony-levelperformancedependsontheirphenotype(e.g.,ahighervalueofzisassociatedwithagreater absolutevalueofCyz).Forexample,boldindividualsdisproportionatelyaffectthepreycapturerateofacolony[79].

Moreover,colonycompositioncanhaveanimpactontheinteractionsbetweenthecolonyanditscommunity[50].

Overall,throughthesedifferentmechanismsgroupcompositionaffectsthefitnessofallcolonymembersintwoways.

First,mixedgroupcompositionincreasesthefitnessofallcolonymembersequally,withbothdocileandaggressive individualsbenefitingfromhavingadiversegroup(byisstrong[72]).Second,particulargroupcompositionswillinteract withthesurroundingcommunitytofavourgreatermassgainineitherdocileoraggressiveindividuals(byzisstrong[80]).

Eventually,theselectiveconsequencesofgroupcompositioncouldleadtotheevolutionofadaptationsenablingcolonies toexpressparticulargroupcompositions,andtoregulateitonceperturbed[51].However,howcolonyperformance selectsforparticularphenotypes(Sz)andhowindividualsrespondtoselectivepressure(i.e.,bymodifyingtraitsaffecting COVzy,Cyz,Czy)largelyremaintobeinvestigated.