Contents lists available atScienceDirect
Behavioural
Processes
j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / b e h a v p r o c
Swarm
intelligence
in
fish?
The
difficulty
in
demonstrating
distributed
and
self-organised
collective
intelligence
in
(some)
animal
groups
Christos
C.
Ioannou
SchoolofBiologicalSciences,UniversityofBristol,Bristol,BS81TQ,UK
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received17July2016
Receivedinrevisedform5October2016 Accepted8October2016 Availableonlinexxx Keywords: Collectiveintelligence Swarmintelligence Collectivecognition Manywrongs Wisdomofcrowds Quorum
a
b
s
t
r
a
c
t
Largergroupsoftenhaveagreaterabilitytosolvecognitivetaskscomparedtosmalleronesorlone individuals.Thisiswellestablishedinsocialinsects,navigatingflocksofbirds,andingroupsofprey collectivelyvigilantforpredators.Researchinsocialinsectshasconvincinglyshownthatimproved cog-nitiveperformancecanarisefromself-organisedlocalinteractionsbetweenindividualsthatintegrates theircontributions,oftenreferredtoasswarmintelligence.Thisemergentcollectiveintelligencehas gainedinpopularityandbeendirectlyappliedtogroupsofotheranimals,includingfish.Despitebeinga likelymechanismatleastpartiallyexplaininggroupperformanceinvertebrates,Iargueherethatother possibleexplanationsarerarelyruledoutinempiricalstudies.Hence,evidenceforself-organised collec-tive(or‘swarm’)intelligenceinfishisnotasstrongasitwouldfirstappear.Theseotherexplanations, the‘pool-of-competence’andthegreatercognitiveabilityofindividualswheninlargergroups,arealso reviewed.Alsodiscussediswhyimprovedgroupperformanceingeneralmaybelessoftenobservedin animalssuchasshoalingfishcomparedtosocialinsects.Thisreviewintendstohighlightthedifficultiesin exploringcollectiveintelligenceinanimalgroups,ideallyleadingtofurtherempiricalworktoilluminate theseissues.
©2016PublishedbyElsevierB.V.
Contents
1. Introduction:themechanismsforimprovedperformanceingroups...00
2. Howcouldfishshoalsachieveswarmintelligence?...00
3. Alternativemechanismsforimprovedcognitiveperformanceinfishshoals...00
3.1. Individual-levelimprovementsincognitiveperformanceingroups...00
3.2. Groupdiversity:the‘pool-of-competence’...00
4. Distinguishingthemechanismsdrivingimprovedcognitiveperformanceingroups...00
4.1. Fishstudies...00
4.2. Othervertebrates ... 00
5. Ecologicalandevolutionaryfactors...00
6. Conclusionandrecommendationsforfuturework...00
Acknowledgments...00
References...00
1. Introduction:themechanismsforimproved performanceingroups
Amajorbenefitanimalsderivefromsocialinteractionsisaccess toinformation(Dalletal.,2005;Danchinetal.,2004).Thissocial informationisrelativelylowcostasitdoesnotrequiredirect
sam-E-mailaddress:[email protected]
plingoftheenvironment(unlikepersonalor‘private’information, whichcanbecostly),andtheformationandmaintenanceofgroups allowsaccesstosocialinformationfrommoreindividualsforlonger periodsoftime.Theflowofinformationcanbeuneven,forexample fromparticularindividualswhohavehadrelevantexperiencesuch asfindingarichfoodpatch,orbemoreegalitarianwhereallgroup membershave anevenprobability ofdetectinganapproaching predatorandallotherindividualscopytheanti-predatoryresponse oftheindividualwhobychancemadethedetection.Withthis
shar-http://dx.doi.org/10.1016/j.beproc.2016.10.005
Fig.1. Waysinwhichindividualsinlargergroupscanmakebetterdecisionsthan
individualsorsmallergroups.Individualsarerepresentedbycircles,andtheflowof
informationbyarrowsfromthesourcetotherecipient.(a)Whenthereis
individual-levelimprovementofcognitiveability,thereisnoflowofinformationbetween
individuals.Instead,beinginagroupreducestheperceptionofpredationrisk,
allow-ingindividualstoallocatemorecognitiveresourcesintoothertasks,orvigilancefor
predatorswhereriskisnot(orless)affectedbypreygroupsize.(b)Informationcan
becentralisedfromallorsomegroupmembers,whereitisprocessedandeitheran
overallgroupdecisionismade,ortheinformationismadeavailabletogroup
mem-berstouse.(c)Inthecaseofleadership,informationflowsfromasingle(orafew)
individual(s)withpertinentknowledgetoothergroupmembers.(d)Onlyinswarm
intelligenceisthereinformationflowandnoobviouskeyindividualthatcentralises
orleads.Notethatforillustration,(a–d)showextremecases.Inthecaseof
house-huntingantsforexample,swarmintelligenceoccursviaself-organisedinteractions
betweenscoutants(asin(d),butwhichareonlyasubsetofthewholecolony),
andonceadecisionismade,therestofthecolonyisled(c)bytheseindividualsto
thenewnestsite(Franksetal.,2003).Similarly,theseprocessesarenotmutually
exclusiveevenatthesametime.Inaforaginggroupofbirds(Morand-Ferronand
Quinn,2011),forexample,individualsmayhaveimprovedforagingduetoareduced
perceptionofrisk(a)andalsobenefitfromthevigilantindividualwhodetectedthe
threat(c).Infishshoals,interactionsoccurdirectlybetweenindividuals(Ioannou
etal.,2011),whileinsocialinsectstheycanbedirect,forexampleresultinginlane
formation(Fourcassiéetal.,2010;Pernaetal.,2012),orindirect,asoccursintrail
formationviathedepositionofpheromones(Moussaidetal.,2009).
ingofinformation,cognitiveperformanceinecologicallyrelevant taskssuchaspredatoravoidanceandforagingcanbeimprovedfor individualsand/orthegroupasawhole.
Thereare diverse ways by which improved intelligence via groupbehaviourcanbeachieved:therecanbestatistical aggre-gationofmultipleopinionsbyacentralisedagent(Galton,1907), arelianceonaknowledgeableormotivatedminorityofthegroup leadingothers(Ioannouetal.,2015),orthedecentraliseddecision makingmostcommonlyseeninsocialinsectcolonies(Bonabeau etal.,1999;Camazineetal.,2001)andslimemoulds(Boisseauetal., 2016;Reidetal.,2016).Althoughtheterms‘swarm’and ‘collec-tive’intelligencehavebeensometimesusedtodescribeallofthese mechanisms(e.g.Krauseetal.,2009),hereitisusefultoexplicitly defineswarmintelligenceasonlyimprovedcognitiveperformance ingroupsthatarisesfromdistributed,self-organiseddecision mak-ing(Bonabeau etal.,1999; Garnieretal., 2007;Kennedy etal., 2001;ReidandLatty,2016).Inthiscase,thecriticalfactoristhat theimprovedperformanceisprimarilyduetoinformationbeing exchanged via repeated local interactions between individuals withoutanysupervisionoftheprocessorcentralisationof infor-mation(Fig.1).Theresultingnetworkofinteractionsthatarises duringswarmintelligenceisthusmuchmorecomplexthanincases ofleadershipandcentralisingofinformation(Fig.1).Itisthis rela-tivecomplexitycomparedtoothermechanisms(Fig.1),resembling aseeminglydisorganisedbutcohesiveswarmofinsects,thatputs the‘swarm’inswarmintelligence(Kennedyetal.,2001).
Underthesedefinitions,statisticalaveragingofmanyguesses cangiveaccuratedecisionsinsometasksduetothecancellingout ofnoisyindividualestimates(oftenreferredtoasthe“wisdomof crowds”;Krauseetal.,2009)andhencecanbeconsidereda mech-anism that improves cognitive performance in groups through (albeitsimple)interactionsbetweenindividuals(e.g.Galton,1907;
Krauseetal.,2009).However,itsrelianceoncentralising informa-tionwhereitisprocessed(Fig.1b)wouldexcludeitfromswarm intelligence.Incontrast,inthe‘many-wrongs’principlethenoisy andinaccurateestimatesofindividualsinwhichdirectiontotravel arecancelledoutthroughaprocessofself-organisation(Codling etal.,2007;Simons,2004).Here,individuals’directionoftravelare influencedbyboththeirprivateinformation(theiropinionor pref-erencetomoveinadesireddirection)andthedirectionoftravelof neighbours(CodlingandBode,2016).Whilethiscanstillbe con-sideredaformofaveraging,itoccursatalocalscaleandthereisa decentralisedexchangeofinformationasindividualsareboth influ-encedby,andinfluence,theirneighbours(Fig.1d).Importantly, improvedperformanceindecisionmakingfromswarmintelligence arisesasanemergentpropertyofindividualcontributionsandthe interactionsbetweenindividuals(i.e.thewholeisgreaterthanthe sumofitsparts).Thismakesperformancethroughswarm intelli-gencemuchlesspredictablethanviaothermechanisms,andoften requirescomputersimulations,aswellasempiricalwork,to under-standfully(Bonabeauetal.,1999).
Tofurtherillustratethedistinctionbetweenswarmintelligence andothermechanismsthatresultinbettercognitiveperformance ofgroups,considercollectivevigilanceforpredators.Here,groups haveagreaterabilitytodetectpredatorsduetopooledvigilance (Elgar,1989;Godinetal.,1988;Magurranetal.,1985;Taraborelli etal.,2012),oftenreferredtoasthe‘many-eyes’effect.Thisisan exampleofgroupintelligencefamiliartoanyonewhohastaken anundergraduatecourseinanimalbehaviour,andisprobablythe earliestexampleofgroupsproviding acognitivebenefitin non-humanvertebrates to be documented (e.g. Miller, 1922).With moreindividualsinagroup,thereisagreaterchanceapredator willbedetected,withthisinformationbeingtransferredtoothers inthegroupviaintentional,activesignals(Seyfarthetal.,1980) orpassivelyfromindividualscopyingfrightorfleeingresponses (TreherneandFoster,1981).Althoughthetransmissionof infor-mationcanoccuratalocalscale,andwithindividualsresponding tothepotentialthreatonlyviaothers thatalsodidnotdirectly respond(Herbert-Readetal.,2015),theflowofinformationcan radiateoutdirectlyfromasinglerespondingindividual(whoacts asatemporaryleader:Fig.1c).Alternatively,therecanbeamore complexexchangeofinformationbetweenindividuals,forexample theremaybefeedbackofinformationbetweenindividualsbefore theydecidewhethertorespond(Fig.1d).Inbothcases,thereisan improvementinpredatoravoidanceinlargergroupsasvigilanceis distributedbetweenmanyeyes,butwhetheritcanbereferredtoas swarmintelligenceaccordingtotheabovedefinitionwilldepend onhowinformationsubsequentlyflowsbetweenindividuals.This islikelytovarybetweenspecies,contexts,andwithgroupsize. Anotherissueisthatthereisaspectrumofinteractionnetworks rangingfromthesimple(Fig.1b,c)tothecomplex(Fig.1d),and itisnotclearatwhat levelofcomplexityanimproveddecision byagroupshouldbeattributedtoswarmintelligence.For exam-ple,ifindividualsrequireacuefromathresholdnumberofother individualsbeforerespondingthemselves,asinquorumdecision making(Sumpter andPratt,2009),but thereis nobidirectional exchangeofinformationbetweenindividuals,shoulditcountas swarmintelligence?
Improved cognitive ability in larger fish shoals has been known for some time. Evidence suggests it is a taxonomically widespread phenomenon and occurs in a range of ecologically relevantcontexts,includingvigilanceforpredators(Godinetal., 1988; Magurran et al., 1985; Wardet al., 2011), foraging (Day etal.,2001;Pitcheretal.,1982;SmithandWarburton,1992)and avoidanceofpollutants(HallJretal.,1982;McNicoletal.,1996). Whenobservingalarge, denselypackedschooloffish respond-ingdynamicallytoavoidattacksfrompredators(e.g.Handegard etal.,2012;MagurranandPitcher,1987),itishardtobelievethat
suchresponsesarenotaresultofself-organisationandhencean exampleofswarmintelligence.However,thecaseforswarm intel-ligenceinfish,definedinastrictsensewheretheimprovementin cognitiveperformancecomesfromtheintegrationofinformation frommultipleindividualsviaself-organisedlocalinteractions,is maybenotaswellestablishedexperimentallyasitmayfirstappear. Whileself-organisationinfishshoalsisnotindoubt(Hemelrijkand Hildenbrandt,2012;Ioannouetal.,2011;Parrishetal.,2002),there arefewstudiesthatshowbetterperformanceinacognitivetaskin largershoalsandalsodemonstratethatthemechanismisbased onself-organisation.Inthisreview,Iwillargueitisactuallyquite difficulttoruleoutotherexplanationsforimprovedperformance inlargergroups,andsuggestpossibleavenuesforfutureworkto determinemoreclearlythemechanismsunderlyingimproved cog-nitiveperformanceinfishshoals.Part ofthemotivationforthe reviewisatendencyforstudiestousuallyfavouronemechanism overothers,whilebelowIsuggestinsteadthatmultiple mecha-nismsarelikelytobeoperatingsimultaneouslyinmanyrealanimal groups(Morand-FerronandQuinn,2011).Thereviewaimstobring togetherwhatisknownaboutimprovedcognitiveperformancein groupsoffishwithotherrelevantaspectsoffishbehaviourto high-lightoutstandingissuesandencouragefurtherworktosolvethese issuesinthefuture.Whererelevant,Iwilldrawontheliteraturein non-fishspecies,particularlysocialinsectsandbirds,wherealot ofpreviousworkhasbeenpublishedongroupperformancein cog-nitivetasks(althoughoftennotreferredtoascollectiveorswarm intelligence).Similarly,althoughthefocusisonfish,manyofthe issuesIhighlightwillapplytootheranimals.
2. Howcouldfishshoalsachieveswarmintelligence?
As discussed above, the key to determining the underlying mechanismforimprovedcognitioninlargergroupsisto under-standthenetworkofinformationtransferbetweenindividuals.In adiverserangeofcollectives,suchasneuralsystems(Couzin,2009), slimemoulds(Reid and Latty,2016)and social insects (Wilson andHölldobler,2009),thenetworksofhowinformationis trans-ferredanddecisionsultimatelymadearerelativelyclearandeasily mapped,allowingresearcherstodemonstratedistributedand self-organiseddecisionmaking.Infishshoalsandothertypesofgroups likebirdflocks,thecuesthattransferinformationwithingroups areless clearly observed. Althoughgroupbehaviourin animals hasbeenstudied for many decades(Krauseand Ruxton, 2002; WardandWebster,2016),onlyrecentlyhaveadvancesin com-putervisiontrackingofanimalsfromvideoandlightweightGPS unitsallowedhighresolutiondatatobeobtainedfrommultiple individualsinagroupsimultaneously(Attanasietal.,2014; Perez-Escuderoetal.,2014;Pettitetal.,2015).Theseadvancesarenow allowingmodelsforhowfishshoalsandbirdflocksform,moveand makedecisions(Aoki,1982;Gautraisetal.,2008;Hemelrijkand Hildenbrandt,2012)tobetestedwithrealanimals,allowingthe networksofinformationtransfertobedetermined(e.g.Attanasi etal.,2014;Strandburg-Peshkinetal.,2013).
Thereareadiverserangeofwaysinwhichfishcommunicate activelywithsignalsincludingsound(Ladich,2000),bodygesture signals(Godin,1995),andcolourchange(NilssonSköldetal.,2013), particularlyinthecontext ofreproductivebehaviour.However, ourcurrent understandingis thatinthevastmajority of shoal-ingfish,groupformation,maintenanceandinformationtransfer occurthroughpassivecueswhereeachindividualrespondstothe positionand movement ofnearneighbours, the basicprinciple underlyingmodelsofcollectivemotion.Itisgenerallybelievedthat fishprimarilyusetwosensorymodalitiestoachievethis(Ioannou etal.,2011):visionmediates attractionandalignment between individuals(Kowalkoetal.,2013),whilethemechanosensory
lat-eralline,ashorter-rangemodalitythatallowsfishtodetectwater movements,isbelievedtoregulaterepulsionsothatindividuals canavoidgettingtooclose(BurgessandShaw,1981;Faucheretal., 2010),andallowsfishtorespondtorapidchangesinthe move-mentofneighboursasoccursduringastartleresponse(Partridge andPitcher,1980).Importantly,cuesfrommotionarerelatively shortrange,limitingtheextentofglobalcues(especiallyinlarger shoals)thatcouldtransferinformationfromasingleindividualto allothersdirectly.Thus,information transferoccurslocally,and increasesthechancethatinformationflowsbetweenindividualsin complexways(Fig.1d),potentiallyallowingforswarmintelligence toemerge.
Fromanempiricalperspective,thisrelativelysimpleformof informationtransferhasadvantages.Usingautomatedcomputer trackingfromvideo(e.g.Delcourtetal.,2013;Perez-Escuderoetal., 2014), the changes in movement of individuals in response to environmentalandsocialcuescanbequantifiedindetail, allow-ingresearcherstodeterminehow fishbalancethesesourcesof informationindecidingwheretomovenext(Berdahletal.,2013; Strandburg-Peshkin et al., 2013). Further technology to record soundproduction orcolourchangeofindividuals (forexample) doesnotseemtobenecessary.Italsoallowsmanipulationthrough theuseofroboticfish(e.g.Fariaetal.,2010),whichcanbeused toexperimentallymanipulateindividualfishorshoals.For exam-ple,individualrobotfishcanbeusedtoinitiatefrightresponsesto examinewhetherandhowinformationspreadsthroughshoals,or asaconspecificforafocalindividualthatwillmaintainshoal cohe-sionbutnotcontributetoacognitivetask,suchasdetectingfood orapredator.
Adisadvantagehowevertohavingtotrackandquantify individ-ualmovementsisthatthisisdifficultinthefield,whichisespecially truewhenfilmingunderwaterwherevisibilityisrestricted com-paredtoinair.Thusalotofworkongroupperformanceinfishhas beenlimitedtothelaboratorybytestingfishonahigh-contrast backgroundto facilitateobservationand tracking, and the eco-logicalvalidityofsuchstudiesnotconductedinthewildremains unknown(Morand-FerronandQuinn,2011).Whiletrackingoffish movementsispossiblefromsonar,ithasyettobeusedto demon-stratecollectiveintelligenceunderfieldconditions,althoughthe responsesof preyshoalstopredatoryattacksarelikely tobea fruitful area of future research. For example, Handegard et al. (2012)usedhighresolution‘acousticvideo’sonartotrack move-mentinschoolsofjuvenileGulfmenhaden(Brevoortiapatronus) whilebeingattackedbyspottedseatrout(Cynoscionnebulosus). Inthisstudy,weshowedthatthedistanceoverwhich informa-tiontransferredinagroupincreasedwithgroupsize,suggesting thatmoreindividualshaveaccesstosociallyderivedinformation in larger groups, which would help facilitate predator avoid-ance.
Therelativeeasewithwhich aspectsoffishsensorysystems canbemeasured(e.g.Kowalkoetal.,2013;Pitaetal.,2015)gives great potentialfortesting howtheinformation transferdriving swarmintelligenceisinturndeterminedbysensorysystems.This isespeciallytruewithnewmodelsofcollectivemovementand col-lectivedetectionofpredatorsthatmakemorerealisticassumptions aboutthesensorypropertiesofanimals(Lemassonetal.,2013; RountreeandSedberry,2009),anapproachwhichissupportedby olderexperimentalworkinfish(Hunter,1969).ThestudyofPita etal.(2015),forexample,measuredthefield-of-viewandvisual acuityoftwospeciescommonlyusedtostudycollectivebehaviour, zebrafish(Daniorerio)andthegoldenshiner(Notemigonus crysoleu-cas).Thiswasthenusedtomakequantitativepredictionsforthe improvementofcollectivedetectionofpredatorsasafunctionof nearestneighbourdistance,whichshowedthatthepotentialfor improveddetectionwasgreaterinzebrafishcomparedtogolden shiners.
Table1
Summaryofstudiesusingfishthathavedemonstratedanimprovedperformanceinacognitivetaskinlargershoalsandproposedamechanism.Studiesaregiveninthe
ordertheyarediscussedinthemaintext.
Refs. Species Proposedmechanism
SmithandWarburton(1992) Blue-greenchromis(Chromisviridis) Individual-levelimprovement
Sumpteretal.(2008) Three-spinedsticklebacks(Gasterosteusaculeatus) Swarmintelligence(quorumdecisionmaking)
Wardetal.(2011) Mosquitofish(Gambusiaholbrooki) Swarmintelligence
Pitcheretal.(1982) Goldfish(Carassiusauratus)andminnows(Phoxinusphoxinus) Notpool-of-competence
Berdahletal.(2013) Goldenshiners(Notemigonuscrysoleucas) Swarmintelligence
Bisazzaetal.(2014) Guppies(Poeciliareticulata) Pool-of-competence
Wangetal.(2015) Zebrafish(Daniorerio) Pool-of-competence
3. Alternativemechanismsforimprovedcognitive performanceinfishshoals
3.1. Individual-levelimprovementsincognitiveperformancein groups
Althoughnumerousstudieshavedemonstratedimproved per-formanceinlargerfishshoals,thereisevidencefromstudiesoffish andotheranimalsthatothermechanismscanprovide(albeit non-mutuallyexclusive)alternativestoswarmintelligence(Table1). Firstly,amajordriverfortheevolutionofgrouplivingisa reduc-tioninpredationrisk(KrauseandRuxton,2002;WardandWebster, 2016).Thiscanoccurbecauseofattackabatement(riskisdiluted betweenindividualswhenpredatorscanonlyconsumealimited numberof preyandtherateofattackis lessthan proportional togroupsize,e.g.Santosetal.,2016),theconfusionofpredators whenmultiplepreyarewithinthevisualfield(Ioannouetal.,2009; Lemassonetal.,2016),andcollectivevigilanceforpredators(Elgar, 1989;Godinetal.,1988).Theseeffectshavebeenshowntoreduce theperceptionofriskingroups(e.g.MagurranandPitcher,1983), allowingindividuals tobelessvigilant forpredators andspend moreoftheircognitiveresourcesonotheractivitiessuchas forag-ing(Goldenbergetal.,2014;GriffinandGuez,2015;Morgan,1988). Thus,individualsthemselvescanaffordtodevotemoreoftheir limitedcognitiveresources(Dukas,2002)toothertasksandmay showgreatercognitiveperformanceingroups,withoutany infor-mationexchangetakingplacebetweenindividuals.Ifthecognitive taskisrelatedtoforaging,thiseffectcanalsooccurasaresultof perceivedcompetitionforfood,whichgenerallyincreasesinlarger groups(GrandandDill,1999;Johnsson,2003).Thisindividual-level explanationforimprovedperformanceingroupshastendedtobe themechanismfavouredinolderstudiesshowingimproved cogni-tiveabilityinfishshoals.Forexample,SmithandWarburton(1992) foundblue-green chromis (Chromisviridis)in largergroups fed morequicklyonconcentratedswarmsofprey,whichtheyargued wasduetoareducedconfusioneffectasthefishcouldreducetheir anti-predatoryvigilanceinlargergroups.Thiswassupportedby furtherbehaviouralobservations:asthefishfed,feedingbecame lessefficientandshoalcohesionincreased,suggestinganincrease inperceivedriskrelativetotheneedtofeedasthefishbecame satiated.
Aninterlinked potentialproblemin experiments iswhether the preferences of individuals change as group size increases, which results in changes to motivation at the individual level ratherthaninformationbeingthekeyfactor.Sumpteretal.(2008), usingthree-spinedsticklebacks(Gasterosteusaculeatus),attributed theimprovedabilityofgroupstodiscriminatebetweenwhichof two differentreplicafish tofollow toa quorum decision mak-ingmechanism,awelldocumentedwayinwhichbeesandants achieveswarmintelligence(SumpterandPratt,2009).However, thisassumesthatindividualsaloneorinsmallgroupsareas dis-criminatoryregarding groupmembers’ phenotypes as those in largergroups.Inmoderatelysizedshoals,thepresenceofa
phe-notypically odd individual can increase predation on both the oddand non-oddgroupmembers(the “oddity”effect;Landeau andTerborgh,1986).Thissuggeststhatmembersoflargergroups shouldbemorediscerningofthephenotypictraitsofwhothey shoalwith, whilesingleorpairs offishmaybenefitmorefrom followinganyindividual(andhenceincreasingtheirgroupsize) comparedtothecostofshoalingwithaphenotypicallydifferent individual.Conversely,ifsmallergroupsareexpectedtobemore discerning,thenthiswouldbeexpectedtodampentheapparent improvementofcognitiveabilityinlargergroups.Intwostudies discussedinmoredetailbelow,singlefishratherthanpairswould beexpectedtopreferthelargeroftwogroups(Bisazzaetal.,2014), andindividualsinsmallergroupswouldbelikelytobemorerisk averse,andbemorelikelytoavoidamodelpredator(Wardetal., 2011).Infact,thereisalsorecentevidencefromantsthat individ-ualsinsmallercoloniescompensatefortheirsmallernumbersby workingharder,aneffectwhichshoulddecreasetheeffectofgroup sizeoncognitiveability(CroninandStumpe,2014).
3.2. Groupdiversity:the‘pool-of-competence’
Inter-individualvariationhasbeenshowntobeimportantin collectiveanimal behaviour, fromvariation in overt traits such assex,body sizeandage,tolessconspicuousvariationintraits suchashunger(Nakayamaetal.,2012)andrisk-takingtendency (IoannouandDall,2016).Thesetraitswilloftencontributeto vari-ationbetweenindividuals inknowledge (McCombet al.,2001), motivation(McDonaldetal.,2016)andcognitiveability(Trompf andBrown,2014).Socialinteractionsthemselvescaneven estab-lishormagnifydifferencesbetweenindividuals(Bergmüllerand Taborsky,2010;Randsetal.,2003).Moreknowledgeable, moti-vatedorcognitivelyableindividualsaremorelikelytoinfluence othergroupmembers,andarelesslikelytobeinfluencedbythem (Calovietal.,2015;Couzinetal.,2005).Thegreaterthevariation betweenindividualsinthesetraits,themorelikelygroupdecisions aremadebyaminorityandarefollowedbytherestofthegroup (GriffinandGuez,2015).Largergroupsarestatisticallymorelikely tocontainknowledgeable,motivatedandableindividuals,which canthusexplainimprovedcognitiveperformanceinlargergroups. Thisisknownasthe‘pool-of-competence’effect(Morand-Ferron andQuinn, 2011)and isa leadingexplanationforwhy ratesof problem-solvingimprovewithgroupsizeinflocksofbirds(Liker andBókony,2009;Morand-FerronandQuinn,2011).Itisalsoa formofleadership(King,2010),whichistaxonomicallywidespread includinginfish,andhasbeenshown tocorrelatewithholding pertinentinformation(Ioannouetal.,2015;Reebs,2000)and moti-vation(Harcourtetal.,2009).ThiseffectmayexplainwhyGodin andMorgan (1985),usingbandedkillifish(Fundulusdiaphanus), foundastrongnegativerelationshipbetweengroupsizeandthe variabilityinthedistancegroupsrespondedtoamodelpredator. Largergroupsaremorelikelytocontainarepresentativesample ofthepopulationandhencebelessvariablethansmallergroups, whichisasamplesizeeffect.
4. Distinguishingthemechanismsdrivingimproved cognitiveperformanceingroups
4.1. Fishstudies
A handful of recent empirical studies have gonesome way towarddistinguishingthemechanismsthatresultingreater cogni-tiveperformanceingroups,highlightingthedifferentapproaches, andalsothedifficultyindoingso.Wardetal.(2011)demonstrated thatlargershoalsofmosquitofish(Gambusiaholbrooki)werefaster andmore accurateinavoiding amodel predatorinone armof aYmaze.In alater study,Bottinellietal. (2013)developedan individual-basedmodeloffishshoalingintheYmazethatrecreated boththeimprovedaccuracyoflargergroupsandthemovement characteristicsofthefishduringdecisionmaking.Thatamodel withindividualbehavioursrepresentativeofcollectivemotionin realfishcouldexplaintheseexperimentalresultsisgoodsupport foraswarmintelligencemechanism.Asthedecisionmaking con-textwasoneofpredatoravoidance,itisalsolesslikelythatthe improvedperformancecouldalternativelybeexplainedby individ-ualshavingareducedperceptionofriskasinthiscaseavoidance ofthepredatorwouldbereducedinlargergroups.Itisconceivable thatgroupingreducesriskmoreforparticulartypesofpredator (CresswellandQuinn,2010;NeillandCullen,1974),allowingfishto belessvigilant(forexample)foraerialpredatorsandmorevigilant forsit-and-waitpredatorssuchassimulatedbythemodelpredator intheWardetal.study.Thishasyettobetested,however.
Wardetal.’sstudyisalsoararecaseusingfishwhereitwas testedwhethertheimproveddecisionmakinginlargershoalswas becausetheseweremorelikelytocontainbetterperforming indi-viduals.Fishweretestedalonerepeatedly9timestodetermine whethersomeindividualshadabetterabilitytoavoidthemodel predator.Theobserveddistributionofthenumberoftimes individ-ualsavoidedthepredatordidnotdiffercomparedtothatexpected fromarandomdistribution,suggestingthattherewasno consis-tentdifferencesbetweenindividuals.However,therelianceona non-statisticallysignificantresultraisesanissuethatisrelevantto experimentalstudiesofgroupdecisionmakingmoregenerally:it isrelativelystraightforwardtoshowthatgroupdecisionsare dis-proportionatelyinfluencedbyparticularindividuals,forexample theboldest(apositiveresult,P<0.05),butmuchmoredifficultto demonstratethatanegativeresult(P>0.05)isduetoegalitarian groupdecisionmakingratherthanalackofstatisticaltestpower. InthecaseofWardetal.’sstudy,alargersampleoffishmayhave pushedthePvaluebelow0.05andalteredtheconclusionsdrawn inthestudy.ThisisamajordrawbacktousingPvaluesinstatistical inference.
Independentlyofthisstatisticalissue,itisalsonotclearwhether thecognitiveperformanceofanindividualwhentestedaloneis representativeofitsperformanceinagroup,whichisrelatedto thepointaboveregardingreducedperceptionofriskby individ-ualsingroups.Isolatingindividualsfromgroupsinanimalsthat have evolved to be social, including many species of fish, can causestress(Hennessyetal.,2009;ReidandLatty,2016),as evi-denced by changes in stress hormones (Galhardo and Oliveira, 2014)andmetabolicrate(Nadleretal.,2016).Relativedifferences betweenindividualsmaychangewhentestedaloneorinasocial context,aswellasaveragelevelsofperformance.Amoreuseful approach andpotential challengeforfuture studies isto quan-tifytherelativecontributionsofindividualdifferencesandswarm intelligencetocollectiveperformance.Thisisparticularly impor-tantasitislikelythatdifferentmechanismsoperatesimultaneously inrealanimalgroups(Morand-FerronandQuinn,2011).Thiswould involvequantifyingindividualperformanceofgroupmembersbut ina waythat isrepresentativewhen individuals interactfreely ingroups,andideallymanipulatinggroupmembershipandgroup
sizeinacontrolledwaytodeterminetherelativecontributionsof eacheffect.AclassicstudybyPitcheretal.(1982)wentsomeway towardthiskindofdesign.Theyshowedthatlargershoalsfound ahiddenfoodsourcemorequickly,andalsodemonstratedthere wasnoaveragedifferencebetweenfocalindividualsintheir for-agingperformancewhentheyweretestedacrosstrialsindifferent shoalsizes(notethat“solitaryfishwerenotincludedascontrols becausepilotwork indicatedtheybehaved aberrantly”: Pitcher etal.,1982).Althoughevidenceagainstthepool-of-competencein foragingfishshoals,inprinciplethismechanismdoesnot necessar-ilyrelyonvariationbetweenindividualsbeingconsistent.Instead, theremaybepopulation-levelvariationinperformancebutthis changesbetweenindividualsovertime,forexampleduetohunger alteringmotivation.Suchaneffectwouldbemuchmoredifficult totiedownandquantifyexperimentallycomparedtoconsistent individualvariationinperformancethatcanbequantifiedthrough repeatedtestingofknownindividuals.
Anotherrecentstudythathasshedlightonhowswarm intel-ligence can occurin fish shoals used a different experimental approach.Berdahletal.(2013)gave1–256goldenshinersa gradi-enttrackingtaskbyprojectingmovingpatchesofdarkareasonto thebottomofalargetestarena,wheredarknessispreferredby thesefish.Herewefoundthatlargergroupswerebetterableto trackthedarkerpatches,measuredasthedarknesslevel atthe positionsofthefishaveragedovertimeandfish,withasignificant improvementinperformanceevenbetweenthelargestgroupsizes (128versus256fish).Atthesegroupsizes,wewouldexpectthat individualfishwouldnotchangetheirperceptionofcompetition orrisk,eitherbecausetheycannotperceivebeinginalargegroup oralargergrouptwicethesize,orbecausethereislittleeffectof groupsizeoncompetitionorpredationriskatthisscaleandhence noreasontomodifybehaviour(Rieucauetal.,2014).Althoughthe pool-of-competenceeffectcouldnotberuledout,ourstudydidlink observedbehavioursofindividualstothemechanismunderlying theimprovedcognitiveperformance,whichmettherequirements ofswarmintelligence.Fromtrajectoriesofindividualfish,Berdahl etal.observedthatfishdidnotacceleratetowarddarkerareas, butinsteadtheymovedtowardareasmoredenselypackedwith shoalmates.Asdarkerareastendedtohavemorefishasindividuals respondedtothelevelofdarknessintheirlocalvicinitybyslowing down(whichalsodecreasedinter-individualspacing),thesedarker areasbecamemoreattractivebecauseoftheotherfishbeingmore likelytooccupytheseareas.Thesebehaviourswerethen incorpo-ratedintoanindividualbasedmodelofshoalingandtherewasa goodmatchbetweenmodelderivedandexperimentalresults, sup-portingtheproposedmechanismforhowswarmintelligencearose intheseexperimentsvialocalinteractionsandself-organisation.
However,itisyettobeshownwhetherthemechanismfound tounderliethisswarmintelligencegeneralisesmorewidely.One featureofourdesignwastouseshallowwaterof8cm,whichis realisticforthesefish(Halletal.,1979)andoftenusedinstudiesof collectivebehaviourtofacilitatecomputertracking.However,this wouldhavelimitedindividuals’abilitiestoseethelevelof dark-nessfurtheraway.Thusthefish’suseofsocialinformation(moving towardareaswithotherfish)comparedtopersonalinformation (movingdirectlytowarddarkpatches)maybegreaterthanthat usuallyobserved.Inotherwords,individuallevelcognitiveability tomovetodarkareasmayhavebeenlimitedbytheexperimental design.Moreexperimentsareneededtotestwhethervaryingwater depthwouldchangethegroupsizeeffectonperformanceandthe weightingofpersonalandsocialinformation,althoughthismay introduce furtherconfoundingeffects asperceptionofrisk may changeindeeperwater(e.g.Harcourtetal.,2009).
Incontrasttothisuseoflargegroupsizes(atleastrelativeto laboratorystudies),Bisazzaetal.(2014)comparednumerical acu-ity(theabilitytodistinguishbetweennumbers)betweensingle
fishandpairsofguppies(Poeciliareticulata).Inbothan ecologi-callyrelevanttaskofdistinguishingbetweentwoshoalsofdifferent sizesand anabstract taskofdistinguishingbetween two stim-uliwithadifferentnumber ofblack dots(whilecontrollingfor stimulussurfacearea,etc.),pairsoffishoutperformed individu-alstestedalone.Moreover,theaverageperformanceofthepairs wascomparedtosimulatedpairsmadeupofrandomlyselected individuals tested alone.Average performance of thereal pairs matchedcloselythebestperformingindividualinthesimulated pairs,andoutperformedtheaverageperformanceofthetwo indi-vidualsinthesimulatedpairs.Thisprovidesconvincingevidence forapool-of-competenceeffect,wherethebetterindividualledthe groupdecisioninthetrialswithtwofish.Asimilareffectwasalso foundbyWangetal.(2015),wherethespeedandaccuracyofpairs ofzebrafishinacolourdiscriminationtaskwasnotsignificantly differenttotheindividualthattendedtomakeslowerbutmore accuratedecisionswhentestedalone.
4.2. Othervertebrates
Todistinguishthe‘pool-of-competence’fromimproved indi-viduallevelperformanceinlargergroups,tworecentstudiesof birdflockshaveuseddifferentapproaches.LikerandBókony(2009) demonstratedthatlargergroupsofhousesparrows(Passer domes-ticus)werefastertoinnovateinanovelforagingtaskandopened moreofthesefeeders.Byexaminingtheeffectofgroupsizeonother behaviours,theyarguedthattherewasnoevidencethat individual-levelperformancewouldbebetterinlargergroups,forexample duetoreducedanti-predatoryvigilance.Thisinvolvedshowingno effectofgroupsizeontheproportionofindividualsattemptingto solvethetask,therateofattemptsperindividual,latenciestofirst approachthetaskorindividualscanrates.Inaseparate neopho-biatest,therewasnodifferencebetweensmallandlargegroupsin thelatencytoapproachanovelobject.Thestudyconcludedthat themostlikelyexplanationforbetterperformanceingroupswas thepool-of-competenceeffect,althoughtheydidnotshowdirect evidenceforthismechanism.Again,arelianceonnon-significant statisticalteststoruleoutcompetinghypothesesisnotwithout problems,asdiscussedabove.
Ina later study,Morand-FerronandQuinn (2011)used nat-urallyoccurringflocksofgreat(Parusmajor)andblue(Cyanistes caeruleus)tits,againstudyingtheirabilitytoinnovateinopening novelfeedingdevices.Theauthorsusedmultiplelinesofevidence todeterminewhetherthepool-of-competenceorareduced per-ceptionofriskcouldexplaintheobservedimprovementinproblem solvingwithgroupsize(measuredasagreaterproportionof feed-ersopened).AsinthestudybyLikerandBókony(2009),latency toapproachthefeederswasnotrelatedtogroupsize,suggesting birdswerenotmoreneophobicinsmallergroups.Feederswere alsoplacednearandfarfromvegetationtomanipulateperceived predationrisk,andalthoughtheproportionoffeedersopenedwas greaterclosertocover,therelationshipbetweengroupsizeand performancewasnotmoresteepfurtherfromcover,suggestinga reductioninriskinlargergroupswasnotdrivingthegroupsize effect(seeGrand andDill,1999).Thestrongestandmostdirect evidenceforthepool-of-competencecamefrombeingableto indi-viduallyidentifybirds,asflockscontainingatleastoneindividual previouslyobservedtohavesolvedthetaskincreasedgroup per-formance.Examiningonlyindividualsthatsolvedthetaskatleast once,theauthorsalsodemonstratedconsistentindividualvariation intheprobabilitytosolvethetaskpervisit,butalsoanimproved abilityperindividualwiththenumberofcompanionspresent, sup-portingareducedperceptionofriskinlargergroupscontributing tosolvingability.Thus,therewasevidenceforbothmechanismsto beoperatinginthissystem.
Themechanismforbetterperformanceinlargergroupswasalso inferredbytheshapeoftherelationshipbetweengroupsizeand performanceonthetask.Thepool-of-competencemechanismwas arguedtobesupportedbasedonthelinearrelationshipobserved betweengroupsizeandperformance,whileifdrivenbya relax-ationofcognitiveresourcesfromanti-predatoryvigilance,there wouldbeasaturatingeffectasgroupsizeincreased(asobserved instudiesofvigilanceinbirdflocks,e.g.Fernandez-Juricicetal., 2007).However,theoreticalworkhasarguedthatthesetwo mech-anismscannot bedistinguishedbased onlyontheshape ofthe observedrelationshipbetweenperformanceandgroupsize(Griffin andGuez,2015).Therelationshipbetweengroupsizeand perfor-mance,ifcausedby apool-of-competence effect,islikely tobe determinedbythetypeofproblemindividualsaretryingtosolve andtheextenttowhichinformationistransferredbetween indi-viduals.Ifacertainproportionofthegroupneedstobeinformed, motivatedand/orable,largergroupswillnotshowanincreased levelofperformancefromthepool-of-competence.Ifinsteada par-ticularnumberofindividualsisrequired(e.g.Couzinetal.,2005), thenonaverage,groupperformanceshouldincreasewithgroup size,aslargergroupsaremorelikelytocontaintherequirednumber ofindividuals.Howandtowhatextentinformationistransferred throughthegroupfromthosethatdirectlyrespondtoastimulus willbeamajordeterminingfactor(Attanasietal.,2014;Handegard etal.,2012;Herbert-Readetal.,2015).Inallcases,fromapurely sta-tisticalprocess,therewillbelessvariabilitybetweenlargergroups compared tobetweensmallergroups(asfoundin thestudyof vigilancebyGodinandMorgan,1985,discussedabove).
5. Ecologicalandevolutionaryfactors
Amajordifferencebetweenshoalsoffish(andmostother col-lectivesofvertebrates)andacolonyofsocialinsectsisthedegree ofrelatednessbetweenindividuals.Recognitionandpreferential shoaling with kin and familiar individuals is well documented infish(Frommenetal.,2007;KamelandGrosberg,2013;Ward etal.,2007),andsomespeciesformstable,smallgroupsofrelated individuals (most notably numerous species ofcichlid and reef fish:Bsharyetal.,2002).Cooperatingwithkinprovidesinclusive fitnessbenefits,andrepeatedinteractionswiththesame individ-ualfavoursreciprocalaltruism(FletcherandZwick,2006;Hesse et al., 2015), both of which would favouroverall group cogni-tiveperformanceratherthanbehaviourstoexploitothersinthe group.Mostgregariousspeciesoffishhowevershowfission-fusion dynamicswhereindividualsfrequentlyexchangemembershipof groups(Ioannouetal.,2011),sothatgroupsareprimarilymadeof unrelated,unfamiliarindividualswhichincreasesconflictsof inter-estandfavours selfishrather thancooperativebehaviour.Thus, behavioursthatdrivegroupintelligenceatthelevelofthe indi-vidualmustprovide shorttermbenefits aswellasresulting in improvedaveragedcognitiveperformanceoverthegroup.Thisis instarkcontrasttosocial insectcolonies,where collective cog-nitionoftenoccurswithinsterileworkercastesthatrelyonthe colony’sreproductivesuccessfortheirinclusivefitness(Couzin, 2009;Seeley,2010,2009),tothepointwherethecolonycanbe con-sidereda“super-organism”(WilsonandHölldobler,2009).Thus, weshouldexpectselectionpressureongroupintelligencetobe muchstrongerinsocialinsectscomparedtofishshoals,andhence bemoredevelopedandsophisticated.
Relatedness within groups and membership stability could explain why there are clearly specialised behaviours that have evolvedin social insects that facilitatecollectivedecision mak-ing,forexampletandemrunninginants(Shafferetal.,2013)and thewaggledanceinbees(BiesmeijerandSeeley,2005).Similarly, intentionalvocalsignalsareusedtotransferinformationwithin
groupsofmammals(Bousquetetal.,2011;Seyfarthetal.,1980) andbirds(Belletal.,2009),althoughthesegroupsalsodifferfrom typicalshoalsoffishastheytendtobemadeupofrelated individ-ualsandhavestablemembershipovertime.Thereislittleevidence currentlyofequivalentactivecommunicationinfish,whereinstead themechanismsforinformation transferthat drivegroup deci-sionsarebelievedtobethesameasthoseforgroupformationand maintenance,i.e.basedonpositionsandmovementsofnear neigh-boursinagroup.Inotherwords,noobvioussignalsseemtohave evolvedspecificallyforinformationtransferduringgroupdecision making.Thisisincontrasttotheuseofsignalsbyfishto trans-ferinformation in non-groupdecision makingcontextssuchas reproduction(Slabbekoornetal.,2012),aggression(O’Connoretal., 1999)andcooperation(Grutter,2004).Itissurprisingthatacoustic signalsinparticulardonotseemtobeusedbyshoalingfishduring groupdecisionssincethereducedvisibilityinwatercomparedto air(especiallyunderturbidconditions)andgreatertransmission ofsoundinwaterwouldmakethisanidealmodalitywithwhich toactivelytransferinformation(Ladich,2000).Ofcourse,thelack ofsignalsmayreflectthehighefficiencyofchangesinmotionto transferinformationwithinshoals,ortheremaybeecological con-straintssuchaseavesdroppingbypredators(Magnhagen,1991). Exploringtheevolutionofactivesignalsacrosstaxaunder differ-entecologicalandsocialconditionsshouldshedlightonwhysuch signalshaveevolvedinsomesystemsandnotothers.
Anotherinterestingareaforfutureresearchistorelate individ-ualandcollectivelevelsofcognitiveability(Couzin,2009).Fishcan showawiderangeofadvancedcognitiveabilities,andthereare moreaspectsofsocialbehaviourinfishthanjusttheoptimisationof collectivedecisionmaking(Brownetal.,2006;Bsharyetal.,2002). Ontheonehand,socialbehaviourssuchasreciprocalcooperation, deceptionandformingcoalitionsselectforincreasedcognitive abil-ityinindividuals(thesocialbrainhypothesis:Dunbar,1998).All elsebeingequal,averagegroupsizeofaspecieswouldpositively correlatewithcognitiveabilityofindividuals.Ontheotherhand, ahighdegreeofgroup-levelperformancemayrelaxselectionon individuallevelcognitiveability(whichiscostly)andwouldthen correlatenegativelywithaveragegroupsize.Arecentcomparative analysisofwaspssuggeststhatinvestmentinbrainstructuresdoes infactdecreaseinsocialcomparedtosolitaryspecies,supporting a“distributedcognitionhypothesis”ratherthanthesocialbrain (O’Donnelletal.,2015).Comparativestudiesoffisharefairly com-mon(Seehausenetal.,1999;Tsuboietal.,2014)anditwouldbe fascinatingtoseewhetherthecomplexityofsocialbehaviours(e.g. deception)andrelianceongroupintelligencecouldexplain differ-encesintherelationshipbetweenaveragegroupsizeandindividual cognitiveability.
It isalso worthgiving someconsideration towhetherthere areecologicalorevolutionaryconsequencesof improved cogni-tion in groups being based ondifferent mechanisms.Although littleresearchhasaddressedthisquestion,itisclearfromother researchoncomplexsystemsthatdisproportionalinfluenceof par-ticularunitsinanetwork,whethertheybeakeystonespeciesinan ecosystem(SoléandMontoya,2001),‘super-spreaders’inadisease network(Adelmanetal.,2015;Wongetal.,2016)orknowledgeable individualsinagroup(McCombetal.,2001),makesthenetwork structuresusceptibleaslossofinfluentialunitshasalargeeffect onthenetworkasawhole.Thus,groupdecisionsbasedonthe dis-tributedswarmintelligence(Fig.1d)maybeparticularlyrobustto thelossofindividualsfromthegroupduetodiseaseorpredation astheydonotrelyonparticularindividualsactingasleadersor ashubsfortheaggregationofinformation.Inasimilarmanner, distributedswarmintelligenceinanimalsmaybemorerobustto someindividualsinfluencingtheoutcomeofthedecisiontofavour theirownself-interest,whichmaybeespeciallytrueinmanyfish
specieswheregroupsaremadeuplargelyofunrelatedindividuals thatdonotformstablebonds,asdiscussedabove.
Apotentialdisadvantage,however,toswarmintelligence com-paredtodecisionsbeingmadebyaminorityofindividualsisits speed.Althoughnotdemonstratedexplicitly,itwouldbeexpected thattherepeatedinteractionsbetweenindividualsthattakeplace duringdistributedgroupdecisionmakingisslowerthanamore lin-earspreadofinformationfromasingleindividualactingasaleader. Decisionswhere speedis a criticalfactor, forexample whether individualsshouldrespondtoapotentialpredator,mayshowa simplernetworkofinter-individualinteractionstominimisethe numberofconnectionsandmaximiseresponsespeed(Fig.1c). Indi-rectevidenceforthiscomesfromants,wherethequorumthreshold requiredforacolonytomakeadecisiondecreasesinharsh condi-tions(Cronin,2016;Franksetal.,2003),whichspeedsupdecision makingandshiftsitfrommoreofacollectivetoanindividual-level decision(Franksetal.,2003).Itshouldalsoberecognisedthat sim-plyhavingdistributed,self-organisedinteractionsingroupsdoes notalwaysresultinswarmintelligencewhenfacingcognitivetasks. Thenetworkofinteractionbetweenindividualsrequires calibra-tion(Kennedyetal.,2001),whichinanimalgroupshasoccurred throughaprocessofnaturalselectionovermanygenerationsand resultedinrelativelycognitivelysimpleindividualsbeingableto solvedifficultproblemswhenactingcollectively(Berdahletal., 2013;Reidetal.,2016;Sasakietal.,2013).However, experimen-talmanipulationsshowswarmintelligenceinanimalgroupscan belimited,forexamplewithgroupsofantsorcaterpillarsfailing toexploitbetterfoodsourcesasaresultofanestablishedforaging trailtoaninferiorfoodsource(Beckersetal.,1993;Dussutouretal., 2007).Thissuggeststhatsuchsystemsmayberelatively inflex-ibleinevolutionarynovelcontexts,withpotentialinnovationby individualsbeingsuppressed.
6. Conclusionandrecommendationsforfuturework
Theexamplesdiscussedabovedemonstratethediversityof con-texts in which groupdecision making occurs.As withdecision makinginindividuals,decisionscanbesingleevents(e.g.finding afoodsource(Pitcheretal.,1982)oravoidingapredator(Ward etal.,2011))orcontinuous,beingmadeofmanysequential deci-sions,suchastrackinganodourgradient(Grünbaum,1998).Some tasksaresuitableforasingleindividualtoperform,whichisthen copiedbyneighboursandspreadsthroughthegroup(Godinand Morgan,1985;LikerandBókony,2009),whileothersrelyonthe spatialspreadofindividualsinthegrouptoimproveaverage cog-nitiveperformance (Berdahlet al.,2013;Pitaet al.,2015).This diversityislikelytobereflectedinvariationinthemechanisms drivinggroup-levelintelligence;differentrequirementsofthetask andtheextentofvariationbetweenindividualsinrelevanttraits suchasmotivationorabilitywillfavourdifferentmechanisms.It isalsoworthreiteratingthatthedifferentmechanismsdiscussed herearenotmutuallyexclusive,andinrealworldanimalgroups it is probablycommonfor a combination ofmechanisms tobe operating simultaneouslytogenerateimprovedperformance in groups (Morand-Ferronand Quinn,2011).Thisisin contrastto thetendencyofmoststudiestofavouraparticularmechanismto explaintheirobservedresults(Table1).Theissuesdiscussedinthis reviewhighlightthedifficultlyinexperimentallydemonstrating self-organisedswarmintelligenceinanimals,andwillhopefully helpresearchersdesignnewexperimentsthatcanshedlighton thisphenomenon(Table2).
Incontrasttomanytopicsinanimalbehaviour,includingother aspects of social behaviour (Elgar, 2015), studies of collective behaviouroftenseekthecommonalitiesintrendsand underly-ingmechanismsthat applyacrossseeminglydisparatesystems,
Table2
Summaryofpotentialexperimentalapproachestohelpdistinguishswarmintelligencefromothermechanismsthatalsoleadtoimprovedcognitiveperformanceingroups.
Theexamplereferencesdonotthemselvesnecessarilyshowimprovedperformanceinlargergroups,noraretheyexclusivetostudiesonfish.
Method Examplereference(s)
Describemechanism(s)forhowlocalinteractionsscaletothegroupdecisionand improvedperformance
Berdahletal.(2013)
Mapnetworkoflocalinteractionswithinthegroupduringgroupdecisionmaking Strandburg-Peshkinetal.(2013)
Quantifyextentofvariationbetweenindividualsintaskperformancewhentested
alone
Wangetal.(2015),Wardetal.(2011)
Trackperformanceofall/focalindividualsrepeatedlytestedingroups Bisazzaetal.(2014),Morand-FerronandQuinn
(2011),Pitcheretal.(1982)
Measureotherbehavioursthatindicateanon-swarmintelligencemechanism,e.g.
reducedvigilanceinlargergroups
LikerandBókony(2009),Magurranand
Pitcher(1983)
Manipulatemotivation(hunger,predationrisk)aswellasgroupsizetoseparate
individual-levelfromgroup-leveleffects
GrandandDill(1999),Morand-Ferronand
Quinn(2011),Morgan(1988)
Testwhetherthevarianceinperformancebetweengroupsataparticulargroupsizeis
affectedbygroupsize
GodinandMorgan(1985)
Demonstrateimprovedcognitiveperformanceovergroupsizestoolargeforindividual
fishtobeabletoperceivethesedifferencesinshoalsize
Berdahletal.(2013)
Quantifytheshapeoftherelationshipbetweengroupsizeandperformance Morand-FerronandQuinn(2011)
whetherthesebeanantcolonyorafishshoal(e.g.Argandaetal., 2012),andevenextendingtosystemsbeyondgroupsofnon-human animalssuchaspedestriancrowds(Moussaidetal.,2009)and neu-ralsystems(Couzin,2009;Passinoetal.,2008).Thisofteninvolves assumingindividuals follow relatively simple behavioural rules thatarestilladequatetoexplaintheobservedcollectivebehaviour. Whileamajorstrengthofthisfieldasitallowsbroadlyapplicable principlestobeidentified(Sumpter,2006),otheraspectsofan ani-mal’sbehaviourmaythrowaspannerintheworkswhenmodels arebeingtestedagainstempiricaldata(Gordon,2007).Thisis par-ticularlytruewhenthecomplexityofindividualsisoversimplified. Thereislittledoubtthatswarmintelligenceoccursinsocialinsects suchasbeesandants,andthereisalargebodyofexperimentaland theoreticalevidencedemonstratingthedetailedmechanismsfor howswarmintelligenceoccurs(e.g.BiesmeijerandSeeley,2005; Bonabeauetal.,1999; Campoetal.,2011; Shafferet al.,2013). Inthecaseofswarmintelligenceinfishshoals,thereis substan-tialexperimentalevidenceforalternativemechanismsforbetter performanceingroups,namelyindividual-levelcognitiveeffects ofbeingin a groupand variation betweenindividualsthat can resultinsomeleadingothers(Table1).Thus,evenwhen experi-mentaldatafitamodelofbehaviourbasedonaswarmintelligence mechanism,other,verydifferent,mechanismsmayprovideequally goodfits(Sumpteretal.,2012).It isimportanttonotethatthe successofdemonstratingswarmintelligenceinsocialinsectshas comefromdocumentingthebehavioursandprocessesthat under-liegroupdecision making,ratherthanrulingoutotherways in whichimprovedperformanceingroupscanarise.Incontrast,the behaviouralobservationsforhowswarmintelligencecouldoccur inothergroupsofanimalsarerarelydemonstrated(butseeBerdahl etal.,2013).
Fishhavebeenapopulartaxonomicgroupwithwhichtostudy social behaviour.Many speciesare smallenough tokeep suffi-cientlylargenumbersofindividualsandgroupsunderlaboratory conditions.Thisallowsforthetestingoflargegroupsizes,andthe testingofanadequatenumberofindependentgroupstobeableto makeinferencesaboutthesourcepopulationasawhole(apoint recentlymadebyVogeletal.(2015)regardingslimemoulds). More-over,duetothefission-fusionsocialstructure ofmanyofthese species,theyarenaturallyfoundatarangeofgroupsizes(Ioannou etal.,2011), andgroupmembershipcanbemanipulatedeasily withminimalstressoraggression.Thustheyallowtherigourand controloftesting inthelaboratorylikeants(e.g.Collignonand Detrain,2014;Shafferetal.,2013),whileshowingsimilar inter-individualcommunicationandpatternsofmovementtothoseof birdswherestudiesof collectivebehaviourare oftenlimitedto fieldstudies(Attanasietal.,2014;Pettitetal.,2015).Forthese
rea-sonsandtheabilitytotrackindividualsingroups,fishholdgreat potentialtocontributetounderstandingthemechanismsthatcan resultingroupintelligence.However,likemostgroupsofanimals, informationisnottheonlyreasontheyformandmaintaingroups (experimentallyshownbyMilleretal.,2013),andindividualsare morecognitivelycomplexthanmodelsassume(Bsharyetal.,2002). Althoughassumingsimplicityofindividualagentswhenmodelling collectivebehaviourallows modelstobegeneric and tractable, researchersshouldstriveforthebroadestpossibleunderstanding ofcognitionandbehaviourtorecognisethelimitationsofthese simplifyingassumptions.Thechallengeforempiricistsistodesign experimentsandanalysesthatconsidertheseeffectsandquantify thecontributionofswarmintelligencetoimprovedcognitive per-formanceingroupsrelativetootherpotentialmechanisms,andto identifythebehaviour(s)attheleveloftheindividualthatresultin swarmintelligence.
Acknowledgments
IthankStephenPratt,SimonGarnier,ChrisReid,GuyTheraulaz, JamesHerbert-Readandananonymousreviewerforcomments. ThisworkwassupportedbyaNERCIndependentResearch Fellow-ship(NE/K009370/1).
References
Adelman,J.S.,Moyers,S.C.,Farine,D.R.,Hawley,D.M.,2015.Feederusepredicts
bothacquisitionandtransmissionofacontagiouspathogeninaNorth
Americansongbird.Proc.R.Soc.BBiol.Sci.282,20151429.
Aoki,I.,1982.Asimulationstudyontheschoolingmechanisminfish.Bull.Jpn.Soc.
Sci.Fish.48,1081–1088,http://dx.doi.org/10.2331/suisan.48.1081.
Arganda,S.,Pérez-Escudero,A.,dePolavieja,G.G.,2012.Acommonrulefor
decisionmakinginanimalcollectivesacrossspecies.Proc.Natl.Acad.Sci.109,
20508–20513,http://dx.doi.org/10.1073/pnas.1210664109.
Attanasi,A.,Cavagna,A.,DelCastello,L.,Giardina,I.,Grigera,T.S.,Jelic,A.,Melillo,
S.,Parisi,L.,Pohl,O.,Shen,E.,Viale,M.,2014.Informationtransferand
behaviouralinertiainstarlingflocks.Nat.Phys.10,691–696,http://dx.doi.org/
10.1038/nphys3035.
Beckers,R.,Deneubourg,J.L.,Goss,S.,1993.Modulationoftraillayinginthe
antLasiusniger(Hymenoptera:Formicidae)anditsroleinthecollective
selectionofafoodsource.J.InsectBehav.6,751–759,http://dx.doi.org/10.
1007/BF01201674.
Bell,M.B.V.,Radford,A.N.,Rose,R.,Wade,H.M.,Ridley,A.R.,2009.Thevalueof
constantsurveillanceinariskyenvironment.Proc.R.Soc.Lond.BBiol.Sci.276,
2997–3005.
Berdahl,A.,Torney,C.J.,Ioannou,C.C.,Faria,J.J.,Couzin,I.D.,2013.Emergent
sensingofcomplexenvironmentsbymobileanimalgroups.Science339,
574–576,http://dx.doi.org/10.1126/science.1225883.
Bergmüller,R.,Taborsky,M.,2010.Animalpersonalityduetosocialniche
specialisation.TrendsEcol.Evol.25,504–511,http://dx.doi.org/10.1016/j.tree.
Biesmeijer,J.C.,Seeley,T.D.,2005.Theuseofwaggledanceinformationbyhoney
beesthroughouttheirforagingcareers.Behav.Ecol.Sociobiol.59,133–142,
http://dx.doi.org/10.1007/s00265-005-0019-6.
Bisazza,A.,Butterworth,B.,Piffer,L.,Bahrami,B.,Petrazzini,M.E.M.,Agrillo,C.,
2014.Collectiveenhancementofnumericalacuitybymeritocraticleadership
infish.Sci.Rep.4,4560,http://dx.doi.org/10.1038/srep04560.
Boisseau,R.P.,Vogel,D.,Dussutour,A.,2016.Habituationinnon-neuralorganisms:
evidencefromslimemoulds.Proc.R.Soc.BBiol.Sci.283.
Bonabeau,E.,Dorigo,M.,Theraulaz,G.,1999.SwarmIntelligence:FromNaturalto
ArtificialSystems.OxfordUniversityPress,Oxford.
Bottinelli,A.,Perna,A.,Ward,A.,Sumpter,D.,2013.Howdofishusethemovement
ofotherfishtomakedecisions?ProceedingsoftheEuropeanConferenceon
ComplexSystems2012,591–606,Springer.
Bousquet,C.A.H.,Sumpter,D.J.T.,Manser,M.B.,2011.Movingcalls:avocal
mechanismunderlyingquorumdecisionsincohesivegroups.Proc.R.Soc.B
Biol.Sci.278,1482–1488,http://dx.doi.org/10.1098/rspb.2010.1739.
Brown,C.,Laland,K.N.,Krause,J.,2006.FishCognitionandBehavior.Blackwell,
Oxford.
Bshary,R.,Wickler,W.,Fricke,H.,2002.Fishcognition:aprimate’seyeview.Anim.
Cogn.5,1–13,http://dx.doi.org/10.1007/s10071-001-0116-5.
Burgess,J.W.,Shaw,E.,1981.Effectsofacoustico-lateralisdenervationina
facultativeschoolingfish:anearest-neighbormatrixanalysis.Behav.Neural
Biol.33,488.
Calovi,D.S.,Lopez,U.,Schuhmacher,P.,Chaté,H.,Sire,C.,Theraulaz,G.,2015.
Collectiveresponsetoperturbationsinadata-drivenfishschoolmodel.J.R.
Soc.Interface12.
Camazine,S.,Deneubourg,J.-L.,Franks,N.R.,Sneyd,J.,Theraulaz,G.,Bonabeau,E.,
2001.Self-OrganizationinBiologicalSystems.PrincetonUniversityPress,
Princeton.
Campo,A.,Garnier,S.,Dédriche,O.,Zekkri,M.,Dorigo,M.,2011.Self-Organized
discriminationofresources.PLoSOne6,e19888.
Codling,E.A.,Bode,N.W.F.,2016.Balancingdirectandindirectsourcesof
navigationalinformationinaleaderlessmodelofcollectiveanimalmovement.
J.Theor.Biol.394,32–42,http://dx.doi.org/10.1016/j.jtbi.2016.01.008.
Codling,E.A.,Pitchford,J.W.,Simpson,S.D.,2007.Groupnavigationandthe
many-wrongsprincipleinmodelsofanimalmovement.Ecology88,
1864–1870,http://dx.doi.org/10.1890/06-0854.1.
Collignon,B.,Detrain,C.,2014.Accuracyofleadershipandcontroloftheaudience
inthepavementantTetramoriumcaespitum.Anim.Behav.92,159–165,http://
dx.doi.org/10.1016/j.anbehav.2014.03.026.
Couzin,I.D.,Krause,J.,Franks,N.R.,Levin,S.A.,2005.Effectiveleadershipand
decision-makinginanimalgroupsonthemove.Nature433,513–516.
Couzin,I.D.,2009.Collectivecognitioninanimalgroups.TrendsCogn.Sci.13,
36–43.
Cresswell,W.,Quinn,J.L.,2010.Attackfrequency,attacksuccessandchoiceofprey
groupsizefortwopredatorswithcontrastinghuntingstrategies.Anim.Behav.
80,643–648.
Cronin,A.L.,Stumpe,M.C.,2014.Antsworkharderduringconsensus
decision-makinginsmallgroups.J.R.Soc.Interface11.
Cronin,A.L.,2016.Groupsizeadvantagestodecisionmakingareenvironmentally
contingentinhouse-huntingMyrmecinaants.Anim.Behav.118,171–179,
http://dx.doi.org/10.1016/j.anbehav.2016.06.010.
Dall,S.R.X.,Giraldeau,L.A.,Olsson,O.,McNamara,J.M.,Stephens,D.W.,2005.
Informationanditsusebyanimalsinevolutionaryecology.TrendsEcol.Evol.
20,187–193.
Danchin,E.,Giraldeau,L.A.,Valone,T.J.,Wagner,R.H.,2004.Publicinformation:
fromnosyneighborstoculturalevolution.Science305,487.
Day,R.L.,MacDonald,T.,Brown,C.,Laland,K.N.,Reader,S.M.,2001.Interactions
betweenshoalsizeandconformityinguppysocialforaging.Anim.Behav.62,
917–925,http://dx.doi.org/10.1006/anbe.2001.1820.
Delcourt,J.,Denoël,M.,Ylieff,M.,Poncin,P.,2013.Videomultitrackingoffish
behaviour:asynthesisandfutureperspectives.FishFish.14,186–204,http://
dx.doi.org/10.1111/j.1467-2979.2012.00462.x.
Dukas,R.,2002.Behaviouralandecologicalconsequencesoflimitedattention.
Philos.Trans.Biol.Sci.357,1539–1547.
Dunbar,R.I.,1998.Thesocialbrainhypothesis.Brain9,178–190.
Dussutour,A.,Simpson,S.J.,Despland,E.,Colasurdo,N.,2007.Whenthegroup
deniesindividualnutritionalwisdom.Anim.Behav.74,931–939,http://dx.doi.
org/10.1016/j.anbehav.2006.12.022.
Elgar,M.A.,1989.Predatorvigilanceandgroupsizeinmammalsandbirds:a
criticalreviewoftheempiricalevidence.Biol.Rev.64,13–33.
Elgar,M.A.,2015.Integratinginsightsacrossdiversetaxa:challengesfor
understandingsocialevolution.Front.Ecol.Evol.3,1–8,http://dx.doi.org/10.
3389/fevo.2015.00124.
Faria,J.,Dyer,J.,Clément,R.,Couzin,I.,Holt,N.,Ward,A.,Waters,D.,Krause,J.,
2010.Anovelmethodforinvestigatingthecollectivebehaviouroffish:
introducingRobofish.Behav.Ecol.Sociobiol.64,1211–1218,http://dx.doi.org/
10.1007/s00265-010-0988-y.
Faucher,K.,Parmentier,E.,Becco,C.,Vandewalle,N.,Vandewalle,P.,2010.Fish
lateralsystemisrequiredforaccuratecontrolofshoalingbehaviour.Anim.
Behav.79,679–687,http://dx.doi.org/10.1016/j.anbehav.2009.12.020.
Fernandez-Juricic,E.,Beauchamp,G.,Bastain,B.,2007.Group-sizeand
distance-to-neighboureffectsonfeedingandvigilanceinbrown-headed
cowbirds.Anim.Behav.73,771–778,http://dx.doi.org/10.1016/j.anbehav.
2006.09.014.
Fletcher,J.A.,Zwick,M.,2006.Unifyingthetheoriesofinclusivefitnessand
reciprocalaltruism.Am.Nat.168,252–262.
Fourcassié,V.,Dussutour,A.,Deneubourg,J.-L.,2010.Anttrafficrules.J.Exp.Biol.
213,2357LP–2363LP.
Franks,N.R.,Dornhaus,A.,Fitzsimmons,J.P.,Stevens,M.,2003.Speedversus
accuracyincollectivedecisionmaking.Proc.R.Soc.Lond.Ser.BBiol.Sci.270,
2457–2463,http://dx.doi.org/10.1098/rspb.2003.2527.
Frommen,J.G.,Mehlis,M.,Brendler,C.,Bakker,T.C.M.,2007.Shoalingdecisionsin
three-spinedsticklebacks(Gasterosteusaculeatus)—familiarity,kinshipand
inbreeding.Behav.Ecol.Sociobiol.61,533–539,http://dx.doi.org/10.1007/
s00265-006-0281-2.
Galhardo,L.,Oliveira,R.F.,2014.Theeffectsofsocialisolationonsteroidhormone
levelsaremodulatedbyprevioussocialstatusandcontextinacichlidfish.
Horm.Behav.65,1–5,http://dx.doi.org/10.1016/j.yhbeh.2013.10.010.
Galton,F.,1907.Voxpopuli(thewisdomofcrowds).Nature75,450–451.
Garnier,S.,Gautrais,J.,Theraulaz,G.,2007.Thebiologicalprinciplesofswarm
intelligence.SwarmIntell.1,3–31,
http://dx.doi.org/10.1007/s11721-007-0004-y.
Gautrais,J.,Jost,C.,Theraulaz,G.,2008.Keybehaviouralfactorsinaself-organised
fishschoolmodel.Ann.Zool.Fennici45,415–428.
Godin,J.G.J.,Morgan,M.J.,1985.Predatoravoidanceandschoolsizeina
cyprinodontidfish,thebandedkillifish(FundulusdiaphanusLesueur).Behav.
Ecol.Sociobiol.16,105–110.
Godin,J.G.J.,Classon,L.J.,Abrahams,M.V.,1988.Groupvigilanceandshoalsizeina
smallcharacinfish.Behaviour104,29–40,http://dx.doi.org/10.1163/
156853988X00584.
Godin,J.-G.J.,1995.Predationriskandalternativematingtacticsinmale
Trinidadianguppies(Poeciliareticulata).Oecologia103,224–229,http://dx.doi.
org/10.1007/BF00329084.
Goldenberg,S.U.,Borcherding,J.,Heynen,M.,2014.Balancingtheresponseto
predation—theeffectsofshoalsize,predationriskandhabituationon
behaviourofjuvenileperch.Behav.Ecol.Sociobiol.68,989–998,http://dx.doi.
org/10.1007/s00265-014-1711-1.
Gordon,D.M.,2007.Controlwithouthierarchy.Nature446,143.
Grünbaum,D.,1998.Schoolingasastrategyfortaxisinanoisyenvironment.Evol.
Ecol.12,503–522.
Grand,T.C.,Dill,L.M.,1999.Theeffectofgroupsizeontheforagingbehaviourof
juvenilecohosalmon:reductionofpredationriskorincreasedcompetition?
Anim.Behav.58,443–451,http://dx.doi.org/10.1006/anbe.1999.1174.
Griffin,A.S.,Guez,D.,2015.Innovativeproblemsolvinginnonhumananimals:the
effectsofgroupsizerevisited.Behav.Ecol.26,722–734,http://dx.doi.org/10.
1093/beheco/aru238.
Grutter,A.S.,2004.Cleanerfishusetactiledancingbehaviorasapreconflict
managementstrategy.Curr.Biol.14,1080–1083.
HallJr.,L.W.,Burton,D.T.,Margrey,S.L.,Graves,W.C.,1982.Acomparisonofthe
avoidanceresponsesofindividualandschoolingjuvenileAtlanticmenhaden,
BrevoortiatyrannussubjectedtosimultaneouschlorineandTconditions.J.
Toxicol.Environ.HealthPartACurr.10(6),1017–1026.
Hall,D.J.,Werner,E.E.,Gilliam,J.F.,Mittelbach,G.G.,Howard,D.,Doner,C.G.,
Dickerman,J.A.,Stewart,A.J.,1979.Dielforagingbehaviorandpreyselectionin
thegoldenshiner(Notemigonuscrysoleucas).J.Fish.BoardCan.36,1029–1039.
Handegard,N.O.,Boswell,K.M.,Ioannou,C.C.,Leblanc,S.P.,Tjøstheim,D.B.,Couzin,
I.D.,2012.Thedynamicsofcoordinatedgrouphuntingandcollective
informationtransferamongschoolingprey.Curr.Biol.22,1213–1217,http://
dx.doi.org/10.1016/j.cub.2012.04.050.
Harcourt,J.L.,Ang,T.Z.,Sweetman,G.,Johnstone,R.A.,Manica,A.,2009.Social
feedbackandtheemergenceofleadersandfollowers.Curr.Biol.19,248–252,
http://dx.doi.org/10.1016/j.cub.2008.12.051.
Hemelrijk,C.K.,Hildenbrandt,H.,2012.Schoolsoffishandflocksofbirds:their
shapeandinternalstructurebyself-organization.InterfaceFocus2,726–737,
http://dx.doi.org/10.1098/rsfs.2012.0025.
Hennessy,M.B.,Kaiser,S.,Sachser,N.,2009.Socialbufferingofthestressresponse:
diversity,mechanisms,andfunctions.Front.Neuroendocrinol.30,470–482,
http://dx.doi.org/10.1016/j.yfrne.2009.06.001.
Herbert-Read,J.E.,Buhl,J.,Hu,F.,Ward,A.J.W.,Sumpter,D.J.T.,2015.Initiationand
spreadofescapewaveswithinanimalgroups.R.Soc.OpenSci.2,140355,
http://dx.doi.org/10.1098/rsos.140355.
Hesse,S.,Anaya-Rojas,J.M.,Frommen,J.G.,Thünken,T.,2015.Kinshipreinforces
cooperativepredatorinspectioninacichlidfish.J.Evol.Biol.28,2088–2096,
http://dx.doi.org/10.1111/jeb.12736.
Hunter,J.R.,1969.Communicationofvelocitychangesinjackmackerel(Trachurus
symmetricus)schools.Anim.Behav.17,507–514.
Ioannou,C.C.,Dall,S.R.X.,2016.Individualsthatareconsistentinrisk-taking
benefitduringcollectiveforaging.Sci.Rep.6,33991,http://dx.doi.org/10.1038/
srep33991.
Ioannou,C.C.,Morrell,L.J.,Ruxton,G.D.,Krause,J.,2009.Theeffectofpreydensity
onpredators:conspicuousnessandattacksuccessaresensitivetospatialscale.
Am.Nat.173,499–506.
Ioannou,C.C.,Couzin,I.D.,James,R.,Croft,D.P.,Krause,J.,2011.Socialorganisation
andinformationtransferinschoolingfish.In:Brown,C.,Laland,K.,Krause,J.
(Eds.),FishCognitionandBehavior.Wiley-Blackwell,NewYork,pp.217–239,
http://dx.doi.org/10.1002/9781444342536.ch10.
Ioannou,C.C.,Singh,M.,Couzin,I.D.,2015.Potentialleaderstradeoffgoal-oriented
andsocially-orientedbehaviorinmobileanimalgroups.Am.Nat.186,
Johnsson,J.I.,2003.Groupsizeinfluencesforagingeffortindependentofpredation
risk:anexperimentalstudyonrainbowtrout.J.FishBiol.63,863–870.
Kamel,S.J.,Grosberg,R.K.,2013.Kinshipandtheevolutionofsocialbehavioursin
thesea.Biol.Lett.9,20130454.
Kennedy,J.,Eberhart,R.C.,Shi,Y.,2001.SwarmIntelligence.Springer,NewYork.
King,A.J.,2010.Followme!I’maleaderifyoudo;I’mafailedinitiatorifyoudon’t?
Behav.Processes84,671–674.
Kowalko,J.,Rohner,N.,Rompani,S.,Peterson,B.,Linden,T.,Yoshizawa,M.,Kay,E.,
Weber,J.,Hoekstra,H.,Jeffery,W.,Borowsky,R.,Tabin,C.,2013.Lossof
schoolingbehaviorincavefishthroughsight-dependentand
sight-independentmechanisms.Curr.Biol.
Krause,J.,Ruxton,G.D.,2002.LivinginGroups.OxfordUniv.Press,Oxford.
Krause,J.,Ruxton,G.D.,Krause,S.,2009.Swarmintelligenceinanimalsand
humans.TrendsEcol.Evol.25,28–34.
Ladich,F.,2000.Acousticcommunicationandtheevolutionofhearinginfishes.
Philos.Trans.R.Soc.Lond.BBiol.Sci.355,1285–1288.
Landeau,L.,Terborgh,J.,1986.Oddityandtheconfusioneffectinpredation.Anim.
Behav.34,1372–1380,http://dx.doi.org/10.1016/s0003-3472(86)80208-1.
Lemasson,B.H.,Anderson,J.J.,Goodwin,R.A.,2013.Motion-guidedattention
promotesadaptivecommunicationsduringsocialnavigation.Proc.R.Soc.B
Biol.Sci.280,20122003,http://dx.doi.org/10.1098/rspb.2012.2003.
Lemasson,B.H.,Tanner,C.J.,Dimperio,E.,2016.Asensory-driventrade-off
betweencoordinatedmotioninsocialpreyandapredator’svisualconfusion.
PLoSComput.Biol.12,e1004708.
Liker,A.,Bókony,V.,2009.Largergroupsaremoresuccessfulininnovative
problemsolvinginhousesparrows.Proc.Natl.Acad.Sci.106,7893–7898.
Magnhagen,C.,1991.Predationriskasacostofreproduction.TrendsEcol.Evol.6,
183–186,http://dx.doi.org/10.1016/0169-5347(91)90210-O.
Magurran,A.E.,Pitcher,T.J.,1983.Foraging,timidityandshoalsizeinminnowsand
goldfish.Behav.Ecol.Sociobiol.12,147–152.
Magurran,A.E.,Pitcher,T.J.,1987.Provenance,shoalsizeandthesociobiologyof
predator-evasionbehaviourinminnowshoals.Proc.R.Soc.Lond.Ser.BBiol.
Sci.229,439–465.
Magurran,A.E.,Oulton,W.J.,Pitcher,T.J.,1985.Vigilantbehaviourandshoalsizein
minnows.Z.Tierpsychol.67,167–178,http://dx.doi.org/10.1111/j.1439-0310.
1985.tb01386.x.
McComb,K.,Moss,C.,Durant,S.M.,Baker,L.,Sayialel,S.,2001.Matriarchsas
repositoriesofsocialknowledgeinAfricanelephants.Science292,491–494.
McDonald,N.D.,Rands,S.A.,Hill,F.,Elder,C.,Ioannou,C.C.,2016.Consensusand
experiencetrumpleadership,suppressingindividualpersonalityduringsocial
foraging.Sci.Adv.2,e1600892,http://dx.doi.org/10.1126/sciadv.1600892.
McNicol,R.,Scherer,E.,Gee,J.,1996.Shoalingenhancescadmiumavoidanceby
lakewhitefish,Coregonusclupeaformis.Environ.Biol.Fishes47,311–319,
http://dx.doi.org/10.1007/BF00000503.
Miller,N.,Garnier,S.,Hartnett,A.T.,Couzin,I.D.,2013.Bothinformationandsocial
cohesiondeterminecollectivedecisionsinanimalgroups.Proc.Natl.Acad.Sci.
U.S.A.110,5263–5268,http://dx.doi.org/10.1073/pnas.1217513110.
Miller,R.C.,1922.Thesignificanceofthegregarioushabit.Ecology3,122–126.
Morand-Ferron,J.,Quinn,J.L.,2011.Largergroupsofpasserinesaremoreefficient
problemsolversinthewild.Proc.Natl.Acad.Sci.,http://dx.doi.org/10.1073/
pnas.1111560108.
Morgan,M.J.,1988.Theinfluenceofhunger,shoalsizeandpredatorpresenceon
foraginginbluntnoseminnows.Anim.Behav.36,1317–1322,http://dx.doi.
org/10.1016/S0003-3472(88)80200-8.
Moussaid,M.,Garnier,S.,Theraulaz,G.,Helbing,D.,2009.Collectiveinformation
processingandpatternformationinswarmsflocks,andcrowds.Top.Cogn.Sci.
1,469–497.
Nadler,L.E.,Killen,S.S.,McClure,E.C.,Munday,P.L.,McCormick,M.I.,2016.
Shoalingreducesmetabolicrateinagregariouscoralreeffishspecies.J.Exp.
Biol.219,2802LP–2805LP.
Nakayama,S.,Johnstone,R.A.,Manica,A.,2012.Temperamentandhungerinteract
todeterminetheemergenceofleadersinpairsofforagingfish.PLoSOne7,
e43747,http://dx.doi.org/10.1371/journal.pone.0043747.
Neill,S.R.J.,Cullen,J.M.,1974.Experimentsonwhetherschoolingbytheirprey
affectsthehuntingbehaviourofcephalopodsandfishpredators.J.Zool.L.172,
569.
NilssonSköld,H.,Aspengren,S.,Wallin,M.,2013.Rapidcolorchangeinfishand
amphibians-function,regulation,andemergingapplications.PigmentCell
MelanomaRes.26,29–38,http://dx.doi.org/10.1111/pcmr.12040.
O’Connor,K.I.,Metcalfe,N.B.,Taylor,A.C.,1999.Doesdarkeningsignalsubmission
interritorialcontestsbetweenjuvenileAtlanticsalmon,Salmosalar?Anim.
Behav.58,1269–1276,http://dx.doi.org/10.1006/anbe.1999.1260.
O’Donnell,S.,Bulova,S.J.,DeLeon,S.,Khodak,P.,Miller,S.,Sulger,E.,2015.
Distributedcognitionandsocialbrains:reductionsinmushroombody
investmentaccompaniedtheoriginsofsocialityinwasps(Hymenoptera:
Vespidae).Proc.R.Soc.BBiol.Sci.282,20150791,http://dx.doi.org/10.1098/
rspb.2015.0791.
Parrish,J.K.,Viscido,S.V.,Grunbaum,D.,2002.Self-organizedfishschools:an
examinationofemergentproperties.Biol.Bull.202,296–305.
Partridge,B.L.,Pitcher,T.J.,1980.Thesensorybasisoffishschools:relativerolesof
laterallineandvision.J.Comp.Physiol.A135,315–325.
Passino,K.,Seeley,T.,Visscher,P.K.,2008.Swarmcognitioninhoneybees.Behav.
Ecol.Sociobiol.62,401–414,http://dx.doi.org/10.1007/s00265-007-0468-1.
Perez-Escudero,A.,Vicente-Page,J.,Hinz,R.C.,Arganda,S.,dePolavieja,G.G.,2014.
idTracker:trackingindividualsinagroupbyautomaticidentificationof
unmarkedanimals.Nat.Methods11,743–748,http://dx.doi.org/10.1038/
nmeth.2994.
Perna,A.,Granovskiy,B.,Garnier,S.,Nicolis,S.C.,Labédan,M.,Theraulaz,G.,
Fourcassié,V.,Sumpter,D.J.T.,2012.Individualrulesfortrailpatternformation
inargentineants(Linepithemahumile).PLoSComput.Biol.8,e1002592.
Pettit,B.,Zsuzsa,A.,Vicsek,T.,Biro,D.,2015.Speeddeterminesleadershipand
leadershipdetermineslearningduringpigeonflocking.Curr.Biol.25,1–6,
http://dx.doi.org/10.1016/j.cub.2015.10.044.
Pita,D.,Moore,B.A.,Tyrrell,L.P.,Fernández-Juricic,E.,2015.Visionintwocyprinid
fish:implicationsforcollectivebehavior.PeerJ.3,e1113,http://dx.doi.org/10.
7717/peerj.1113.
Pitcher,T.J.,Magurran,A.E.,Winfield,I.J.,1982.Fishinlargershoalsfindfood
faster.Behav.Ecol.Sociobiol.10,149–151.
Rands,S.A.,Cowlishaw,G.,Pettifor,R.A.,Rowcliffe,J.M.,Johnstone,R.A.,2003.
Spontaneousemergenceofleadersandfollowersinforagingpairs.Nature423,
432–434,http://dx.doi.org/10.1038/nature01630.
Reebs,S.G.,2000.Canaminorityofinformedleadersdeterminetheforaging
movementsofafishshoal?Anim.Behav.59,403–409.
Reid,C.R.,Latty,T.,2016.Collectivebehaviourandswarmintelligenceinslime
moulds.FEMSMicrobiol.Rev.,http://dx.doi.org/10.1093/femsre/fuw033.
Reid,C.R.,MacDonald,H.,Mann,R.P.,Marshall,J.A.R.,Latty,T.,Garnier,S.,2016.
Decision-makingwithoutabrain:howanamoeboidorganismsolvesthe
two-armedbandit.J.R.Soc.Interface13.
Rieucau,G.,Fernö,A.,Ioannou,C.C.,Handegard,N.O.,2014.Towardsofafirmer
explanationoflargeshoalformation,maintenanceandcollectivereactionsin
marinefish.Rev.FishBiol.Fish.,1–17.
Rountree,R.A.,Sedberry,G.R.,2009.Atheoreticalmodelofshoalingbehavior
basedonaconsiderationofpatternsofoverlapamongthevisualfieldsof
individualmembers.ActaEthol.12,61–70,
http://dx.doi.org/10.1007/s10211-009-0057-6.
Santos,R.G.,Pinheiro,H.T.,Martins,A.S.,Riul,P.,Bruno,S.C.,Janzen,F.J.,Ioannou,
C.C.,2016.Theanti-predatorroleofwithin-nestemergencesynchronyinsea
turtlehatchlings.Proc.R.Soc.Lond.BBiol.Sci.283,20160697.
Sasaki,T.,Granovskiy,B.,Mann,R.P.,Sumpter,D.J.T.,Pratt,S.C.,2013.Antcolonies
outperformindividualswhenasensorydiscriminationtaskisdifficultbutnot
whenitiseasy.Proc.Natl.Acad.Sci.110,13769–13773,http://dx.doi.org/10.
1073/pnas.1304917110.
Seehausen,O.,Mayhew,P.J.,VanAlphen,J.J.M.,1999.Evolutionofcolourpatterns
inEastAfricancichlidfish.J.Evol.Biol.12,514–534,http://dx.doi.org/10.1046/
j.1420-9101.1999.00055.x.
Seeley,T.D.,2009.TheWisdomoftheHive:TheSocialPhysiologyofHoneyBee
Colonies.HarvardUniversityPress.
Seeley,T.D.,2010.HoneybeeDemocracy.PrincetonUniversityPress.
Seyfarth,R.M.,Cheney,D.L.,Marler,P.,1980.Vervetmonkeyalarmcalls:semantic
communicationinafree-rangingprimate.Anim.Behav.28,1070–1094.
Shaffer,Z.,Sasaki,T.,Pratt,S.C.,2013.Linearrecruitmentleadstoallocationand
flexibilityincollectiveforagingbyants.Anim.Behav.86,967–975,http://dx.
doi.org/10.1016/j.anbehav.2013.08.014.
Simons,A.M.,2004.Manywrongs:theadvantageofgroupnavigation.TrendsEcol.
Evol.19,453–455.
Slabbekoorn,H.,Verzijden,M.,May,C.,2012.In:Popper,A.N.,Hawkins,A.(Eds.),
CichlidCourtshipAcoustics:SignalsandNoiseInfluenceReproductive
BehaviorBT–TheEffectsofNoiseonAquaticLife.SpringerNewYork,New
York,NY,pp.407–409,http://dx.doi.org/10.1007/978-1-4419-7311-593.
Smith,M.F.L.,Warburton,K.,1992.Predatorshoalingmoderatestheconfusion
effectinblue-greenchromis,Chromisviridis.Behav.Ecol.Sociobiol.30,
103–107.
Solé,R.V.,Montoya,M.,2001.Complexityandfragilityinecologicalnetworks.Proc.
R.Soc.Lond.Ser.BBiol.Sci.268,2039LP–2045LP.
Strandburg-Peshkin,A.,Twomey,C.R.,Bode,N.W.F.,Kao,A.B.,Katz,Y.,Ioannou,
C.C.,Rosenthal,S.B.,Torney,C.J.,Wu,H.S.,Levin,S.A.,Couzin,I.D.,2013.Visual
sensorynetworksandeffectiveinformationtransferinanimalgroups.Curr.
Biol.23,R709–R711.
Sumpter,D.J.T.,Pratt,S.C.,2009.Quorumresponsesandconsensusdecision
making.Philos.Trans.R.Soc.BBiol.Sci.364,743–753,http://dx.doi.org/10.
1098/rstb.2008.0204.
Sumpter,D.J.T.,Krause,J.,James,R.,Couzin,I.D.,Ward,A.J.W.,2008.Consensus
decisionmakingbyfish.Curr.Biol.18,1773–1777.
Sumpter,D.J.T.,Mann,R.P.,Perna,A.,2012.Themodellingcycleforcollective
animalbehaviour.InterfaceFocus2,764–773,http://dx.doi.org/10.1098/rsfs.
2012.0031.
Sumpter,D.J.T.,2006.Theprinciplesofcollectiveanimalbehaviour.Philos.Trans.
R.Soc.BBiol.Sci.361,5–22.
Taraborelli,P.,Gregorio,P.,Moreno,P.,Novaro,A.,Carmanchahi,P.,2012.
Cooperativevigilance:theguanaco’s(Lamaguanicoe)keyantipredator
mechanism.Behav.Processes91,82–89,http://dx.doi.org/10.1016/j.beproc.
2012.06.002.
Treherne,J.E.,Foster,W.A.,1981.Grouptransmissionofpredatoravoidance
behaviourinamarineinsect:thetrafalgareffect.Anim.Behav.29,911–917.
Trompf,L.,Brown,C.,2014.Personalityaffectslearningandtrade-offsbetween
privateandsocialinformationinguppies,Poeciliareticulata.Anim.Behav.88,
99–106,http://dx.doi.org/10.1016/j.anbehav.2013.11.022.
Tsuboi,M.,Gonzalez-Voyer,A.,Kolm,N.,2014.Phenotypicintegrationofbrainsize
andheadmorphologyinLakeTanganyikacichlids.BMCEvol.Biol.14,39,
Vogel,D.,Nicolis,S.C.,Perez-Escudero,A.,Nanjundiah,V.,Sumpter,D.J.T.,
Dussutour,A.,2015.Phenotypicvariabilityinunicellularorganisms:from
calciumsignallingtosocialbehaviour.Proc.R.Soc.Lond.BBiol.Sci.282,
20152322.
Wang,M.-Y.,Brennan,C.H.,Lachlan,R.F.,Chittka,L.,2015.Speed–accuracy
trade-offsandindividuallyconsistentdecisionmakingbyindividualsand
dyadsofzebrafishinacolourdiscriminationtask.Anim.Behav.103,277–283,
http://dx.doi.org/10.1016/j.anbehav.2015.01.022.
Ward,A.J.W.,Webster,M.M.,2016.Sociality:TheBehaviourofGroupLiving
Animals.SpringerInternationalPublishing,
http://dx.doi.org/10.1007/978-3-319-28585-6.
Ward,A.J.W.,Webster,M.M.,Hart,P.J.B.,2007.Socialrecognitioninwildfish
populations.Proc.R.Soc.Lond.BBiol.Sci.274,1071–1077.
Ward,A.J.W.,Herbert-Read,J.E.,Sumpter,D.J.T.,Krause,J.,2011.Fastandaccurate
decisionsthroughcollectivevigilanceinfishshoals.Proc.Natl.Acad.Sci.108,
2312–2315,http://dx.doi.org/10.1073/pnas.1007102108.
Wilson,E.O.,Hölldobler,B.,2009.TheSuperorganism.WWNorton&Co.,New
York,NY.
Wong,G.,Liu,W.,Liu,Y.,Zhou,B.,Bi,Y.,Gao,G.F.,2016.MERS,SARS,andebola:the
roleofsuper-spreadersininfectiousdisease.CellHostMicrobe18,398–401,
