Reference: Biol. Bull. 202: 2S3-288. (June 20(12) Self-Organization and Natural Selection in the Evolution of Complex Despotic Societies C. K. HEMELRIJK Department ofInformation Technology andAnthropological Institute and Museum. University ofZurich. Switzerland Abstract. Differences between related species are usually side effect of the change of one single trait proper to an explained as separate adaptations produced by individual se- egalitarian society, in the real world a despotic society may lection. Idiscussinthispaperhow relatedspecies, whichdiffer alsohave arisen as asideeffectofthe mutation ofasingle trait in many respects, may evolve by a combination of individual ofan egalitarian species. selection, self-organization, and group-selection, requiring an Ifgroups withdifferent intensitiesofaggression evolve in evolutionary adaptation ofonly a single trait. In line with the this way, they will also have different gradients of hierar- supposed evolution ofdespotic species of macaques, we take chy. When foodis scarce, groups with the steepest hierarchy as a stalling point an ancestral species that is egalitarian and may have the bestchance to survive, because at leastasmall mildly aggressive. We suppose it to live in an environment number of individuals in such a group may succeed in with abundant food and we put the case that, iffood becomes producing offspring, whereas in egalitarian societies every scarce and more clumped, natural selection at the level ofthe individual is at risk ofbeing insufficiently fed to reproduce. individual will favor individuals with a more intense aggres- Therefore, intrademic group selection (selection within an sion (implying, for instance, biting and fierce fighting). interbreeding group) may have contributed to the evolution Using an individual-centered model, called DomWorld, I of despotic societies. show whathappenswhen the intensity ofaggression increases. In DomWorld,grouplifeisrepresentedby artificial individuals Introduction that live in a homogeneous world. Individuals are extremely simple: all they do is flock together and, upon meeting one The assumption that evolution occurs through a single another, they may perform dominance interactions in which evolutionary process is no longer tenable (e.g., see Plotkin tihneteenfsfietcytsofofawgignrensisnigonanidnltohseinmgoadreelselisf-irnecinrfeoarsceidn,g.aWchoemnpltehxe oa1rn9id9e4s;OdhMlaaivyneng-asSlrmodweleSy,mibt1e9hc81ao)nm,deaSmnzdoartmheumlaatrciycp,leep-1tl9ee9dv5e;(le.Mgsi.et.lteeHcltodigooernfwetahgne,d- feedback between the hierarchy and spatial structure results; Wilson, 2000). Multiple-level selection processes may in- via self-organization, this feedback causes the egalitarian so- clude some, or all. of the following factors: the multi-level ciety to change into a despotic one. The many differences character of biological systems and natural selection oper- between the two types ofartificial society closely correspond ating on them (Lewontin, 1970; Hogeweg, 1994; Sober and tothose between despotic and egalitarian macaques in the real Wilson, 1998), self-organization and its consequences for world. Given that, in the model, the organization changes as a evolution (Boerlijst and Hogeweg, 1991), and nonlinear genotype-phenotype mappings (Kauffman, 1993; Huynen E-mail: [email protected] and Hogeweg. 1994; Kauffman. 1995). This paper was originally presented at a workshop titled The Limits in Within this framework of multiple-level selection theo- Self-Organization inBiologicalSystems.Theworkshop,whichwasheldat ries, I present in this paper an example ofthe way in which the J. Erik Jonsson Center ofthe National Academy ofSciences, Woods a certain type of society may evolve. In studies of animal HfoorleA.dvMaasnscaecdhuSsteutdtise,sfirnotmhe11S-p1a3ceMaLiyfe20S0c1i.enwceassastpothnesoMraerdinbey tBhieolCoegnitcealr behavior, a distinction is usually made between twotypes of Laboratory, and funded by the National Aeronautics and Space Adminis- societies egalitarian and despotic. In her studies of birds, tration underCooperative Agreement NCC 2-896. Vehrencamp (1983) distinguishes between these two on the 283 284 C. K. HEMELRIJK basis of reproductive success. In egalitarian societies, the Dominance Rules Grouping reproductive success of all female group members is more or less the same, whereas in despotic ones there is a great variation in the reproductive success of individual females. When the terms egalitarian and despotic are used for other taxa, such as primates, these terms denote their so-called dom- inance style rather than their reproductive success (Hand, (OJHERSMAXVIEW 1986). Dominance style depends on the gradient of the hier- / ~v~ archy (van Schaik. 1989): despotism signifies that the domi- [ IYES] [NO] nance hierarchy is steep (which means that there is a great TURN difference in the success in fighting between individuals); ISEARCH-ANGLEI egalitarianism implies that the hierarchy is weakly developed. Furthermore, in primates, particularly ofthe genus macaques, Figure 1. Flow chart fur the behavioral rules ofagents. these societies differ in a number of other characteristics: in despotic societies aggression is more unidirectional, social in order to rejoin its group. In this way individuals tend to behavioriscorrelatedmorestrongly withdominance,grouping remain in a group. is less cohesive (de Waal and Luttrell. 1989; Thierry, 1990), If, however, an agent sees another agent close by, within mate choice is more selective, and male migration is more its personal space (PersSpace = 2 or 4). a dominance frequent (Caldecott, 1986). Whereas each ofthese differences interaction takes place. The likelihood that an agent initiates between related species is usually explained as a separate an aggressive interaction increases with its chance to defeat adaptation shapedby individual selection, I proposetoshow in its opponent (Hemelrijk, 2000). The agent's capacity to be this paper how a despotic species may evolve from an egali- victorious (reflected in its dominance value) depends on tarian one by a combination of individual selection, self-orga- chance, on its current dominance value (which initially is nization, and group-selection, involving an evolutionary adap- the same for all individuals), and on the self-reinforcing tation in one single trait only. effect of the outcome (winning a fight increases the proba- How individual selection and self-organization may op- bility ofwinning the next one and losing decreases it). This erate, I will explain with the help of an individual-centered is known as the "winner and loser" effect and has been model of a group-living species, called "DomWorld" empirically established in many animal species (e.g., see (Hemelrijk, 1999a,b. 2000). How group selection may favor Chase, 1985). After victory, the dominance value of the the survival of groups with the steepest hierarchy above victorious agent increases and that of its defeated opponent groups with weakergradients ofthe hierarchy. I will explain is reduced by the same amount. When, unexpectedly, an with the help of an ecological model at an evolutionary agent defeats a higher ranking opponent, the dominance time-scale designed by Ulbrich ct til. ( 1996). values of both opponents are changed by a greater amount than when an agent conquers, as expected, a lower ranking opponent(thisresultconforms todetailedbehavioral studies The Model DomWorld (Dominance World) on bumblebees by Honk and Hogeweg. 1981). In this way the model allows forrankreversals. Afterafight, the winner AsummaryofDomWorldmay suffice(foramorecomplete chases the opponent, and the defeated agent flees. description, see Hemelrijk ( 1999b. 2000)). The model isbased Groups usually consist of8 to 10 individuals. The behav- ononlyasmall numberofessentialsofsocial life. It represents ior of the agents is analyzed by means of behavioral units a homogeneous virtual world inhabited by agents that are and statistical methods similar to those used for observing provided with only two tendencies: to group (right halfofFig. real animals. 1) and to perform dominance interactions (left pail ofFig. 1). Why agents actually do group (whether this is to avoid pred- Natural Selection and Self-Organization ators or because resources are clumped) is not specified and irrelevant to the model. The same holds for dominance inter- Group life of primates (including macaques) is generally actions. They reflect competition for resources (such as food supposed to have evolved as a protection against predators and mates), but these resources are not specified. (van Schaik and van Hooff. 1983; van Schaik, 1989). Egali- When an individual is activated and it does not see tarian primate societies,at least thoseofmacaques, arethought another agent close by (within its personal space, see Pers- to precede despotic ones evolutionarily (Matsumura, 1999: Space in Fig. I ), then grouping rules come into effect. It Thierry ct til., 2000). Correspondingly, let us suppose that a starts looking for others at greater and greater distances group-living egalitarian species lives in an environment in (Near View = 24 and Max View = 50 units). Ifeven then which food is abundant, and therefore its competition is low no one else is in sight, it turns over a SearchAngle (Fig. 1 ) and its aggression mild (consisting of, for instance, threats and EVOLUTION OF COMPLEX DESPOTIC SOCIETIES 285 286 C. K. HEMELRIJK animals when unfamiliar individuals are put in a group to- Contest Competition gether, but the decline of aggression has so far never been Food Scarcity attributed to a widening of the group.) A consequence of the (2) declineofaggression isthatthehierarchybecomesmore stable ((3) in Fig. 3). Further, because low-ranking individuals flee fromeveryone else, they end up at the periphery ofthe group; this automatically leaves the dominants in the center ((5) in FoodAbundance + Fig. 3). This spatial structuredevelops in spiteofatotal lackof any preference ofthe individuals to be in the center. Yet such Most DespoticGroups a preference (a so-called centripetal instinct) is assumed in the with Steepest "selfish herd" theory by Hamilton (1471). It is supposed to HierarchiesSurvive have been evolved because individuals are better protected in the center, where they are shielded on all sides from possible predators. However, the model shows thateven without such a centripetal instinct, whenever the hierarchy is steep, we must expect a spatial structure with dominants in the center. Spatial centrality, in turn, stahili/.es the hierarchy and supports its differentiation ((5) and (6) in Fig. 3). This arises because the spatial structure causes individuals to be close to partners of similar dominance rank: therefore, ifinciden- tally a dominance-reversal occurs, it is usually between individuals that are similar in dominance, and thus the extent to which dominance values are changed is only a minor one. In this way, the spatial structure strengthens the hierarchy ((?) and (6) in Fig. 3). Thisbecomes evident when we eliminate spatial centrality. We can dothis by decreasing the SearchAngle. If the SearchAngle for returning to the group is made smaller, so that individuals return more slowly and the group therefore spreads out more and more, no spatial structure develops. In this case (for the same number of dominance interactions), the hierarchy becomes weaker (i.e., dominance values differentiate less) than in cohesive groups (Hemelrijk. 1999a). The steeper hierarchy in cohesive groups is thus partly due to a feedback rein- forcement of the hierarchical development under the influ- ence of spatial structure. We mayalsoincludethesexesinthe model;forinstance,by making two classes of individuals that differ in initial domi- nance value and intensity ofaggression (both are made higher lor"males" than lor"females"). Unexpectedly, it turns out that at ahigh intensity ofaggression, femaledominanceovermales is greater than at a low one. This arises because a stronger differentiation of the hierarchy causes some females to reach high dominance and some males to become very low in rank. Consequently, some females become dominant over some males. This is of interest, because in comparative studies between egalitarian macaques (with mild aggression) and des- potic ones (with tierce aggression). Thierry (1990) notes that despotic adult females remain dominant over fast-growing adolescents longer than females in egalitarian species do (which is in accordance with DomWorld). He attributes this to astrongercoalitionary tendency among kin-related individuals in despotic species than in egalitarian ones. DomWorld, how- ever, shows that greaterfemale dominance may also arise as a EVOLUTION OF COMPLEX DESPOTIC SOCIETIES 287 model represents, among other things, colony survival, repro- notype and the phenotype). These results of DomWorld may ductivebehavior,thekindofcompetitionforfood("contest"or alsoberelevanttoresultsofaselectionexperimentwith fishby "scramble" competition), and feeding behavior. If food is Ruzzante and Doyle (1991. 1993). After two generations of clumped, it leads to contest competition, whereby dominants selection for speed of growth, an increase in growth speed get more than subordinates; ifit is scattered and unpredictable, among fish fed on clumped food (leading to intense competi- everyone gets aboutthe same (so-called scramblecompetition, tion) was accompanied by three effects: a decrease in aggres- see Nicholson. 1967). Results show that during levels offood siveness,anincreaseindensityofschooling,andan increasein scarcity, colonies where there is contest competition survive social tolerance. According to the authors, selection for fast longerthan colonies in which competition is ofthe scrambling growth results in a high threshold for aggression, and this type.This isaconsequenceofthe steeperdominance hierarchy threshold also genetically influences the other two features of and, therefore, greatervariety in body-size ofmembers, which social behavior cohesion and social tolerance. These find- allows a small number of dominant individuals to eat all the ings resemble those from DomWorld; but in DomWorld only food so that at least some of the females are sufficiently the intensity of aggression is changed "genetically." and all nourishedtoreachreproductiveage. Inthisway,becausesome other changes of social behavior are mere side effects. Such individuals in more intensely despotic groups are likely to parsimoniousexplanations may be relevant to many speciesof survive and reproduce, the group itselfmay survive (although despotic animals that have been studied and possibly even to the number ofits members decreases during periods ot dimin- plants, where akind ofdespotism is alsodescribed (forastudy ished food availability). In contrast, in groups with a weaker on a hemiparasite, see Prati et ai, 1997). hierarchy, food will be distributed more evenly and no single Groupswithdifferentdegreesofintensityofaggression may individual may reach reproductive success (see also Rypstra. be liabletogroupselection similartothat suggestedforspiders 1993). Thus, severe shortage offood in clumped patches may by Ulbrich et til. ( 1996). Group selection has always been a result in groupselection in favorofthe most despotic societies controversialissue.Theonly studiesinwhichanyevidencehas ((3) to (4) in Fig. 4). been produced (Bradley. 1999) concern invertebrates (namely, Note that in macaques, adult males usually migrate from aspeciesofvirus,social spider,andant). However,designating their natal group to other groups. Therefore, the degree of group selection as a useful explanation for behavior of (non- despotism in macaques is defined on the basis of the dom- human) primates is usually avoided (Bradley, 1999). Yet. inance hierarchy among the females. Also, the survival and primates utter alarm calls that, though they are at the expense extinction of"the group" implies the core group offemales of the fitness of the individual that uses them, may serve to only, because they are the resident sex. Among such female protect group members (both kin and non-kin). When alarm- resident core-groups, we may expect that when food is calls are more beneficial for the group than they are costly for scarce, group selection will lead to better survival of those the individual, they will evolve by group selection, as ex- groups that have a steeper hierarchy (in accordance with the plained by differential selection among groups for the same findings in social spiders). trait so-called intrademic group selection (Wilson. 2001). Computer models show how such an intrademic group selec- Discussion tion favors altruistic traits, particularly under harsh conditions (Mitteldorfand Wilson. 2000; Pepper and Smuts. 2000). Changing only one trait intensity ofaggression in the Similarly, I suggest that group selection may operate on a model leads to a great number of phenotypic differences at non-altruistic trait, namely intensity of aggression, in primate the level of the individual and of the group. Thus, by societies in which there is sharp contest competition for food changing the intensity of aggression in DomWorld, as sug- (during harshconditions). Such strongercompetition leadstoa gested by the hypothesis for egalitarian and despotic ma- more asymmetrical distribution offood intake, and by guaran- caques (McKenna. 1979; Thierry. 1985a. b. 1990). we may teeing that at least some individuals ofa group reproduce, aids switch from an egalitarian society to a despotic one. Inten- group survival (ofthe core group of the resident sex). sity ofaggression is not the only variable, however, thatcan Note that in the othertransition, fromdespotic societiesto produce such a cascade of effects. If we increase cohesion egalitarian ones (see Fig. 4). progressively milder aggres- under one and the same level of intensity of aggression sion would be favored by selection on the level of the (which must lie somewhere between medium and high individual and the group, because less energy is wasted on values), the society becomes more despotic in all its char- conflicts. In this case, selection at t'te level ofthe group and acteristics (Hemelrijk. 1999a). The question of whether the individual will be weak, and self-organization will not stronger cohesion ofgroups is also accompanied by greater be operative. Thus, a different numi'fr of processes are at despotism in real primates must be studied empirically. work depending on the direction of t'io transition between Thus, the model shows how self-organization causes non- egalitarian and despotic societies. linearity in theconnection between thebehavioral rulesand the The different ways in which selection may act (namely, observed behavior (which respectively correspond to the ge- on self-organized patterns, and on the level ofthe individual 2X8 C. K. HEMELRUK and the group) illustrate how we may envision a multiple- Kaut'fman, S. A. 1993. The Origins of Order: Self-organisation and level selection theory for the creation ofa despotic society. Selection in Evolution. Oxford University Press, New York. Although, for the sake of clarity in the sketch I have given Lewontin. R. C. 1970. The units ofselection. Anim. Rev. Eeol. S\st. 1: 1-18. above. I have made individual selection, self-organization, Matsumura. S. 1999. The evolution of 'Egalitarian' and 'Despotic' and group selection function one after the other, in reality social systems among macaques. Primates 40: 23-31. they operate mostly simultaneously in different proportions MaynardSmith.J..andE.Szathmary. 1995. TheMajorTransition.*, in (Lewontin, 1970). Evolution. W. H. Freeman. New York. McKenna, J. J. 1979. The evolution ofallomothering behavior among Acknowledgments Colobine monkeys: function and opportunism in evolution. Am. An- thropol. 81: 818-840. I thank Rolf Pfeiter and Bob Martin for continuous sup- Mittt-ldorf, J. and D. S. Wilson. 2001). Population viscosity and the port. I am grateful to Scott Camazine for inviting me to the evolution of altruism. ./. Theor. Biol. 204: 481-496. workshop on "The Limits to Self-organization in Biological NichSoplrisnotn;,HAa.rbJo.r19S6v7m.p. SQeulafn-ta.djBuisotl.me2n2t:of15p3o-p1u7l3a.tions to change. Cold Systems," and to Bernard Schmidt and Jens Krause for Pepper,J. W.,and B. B.Smuts. 2000. Theevolution ofcooperation in making me aware of similarities between DomWorld and anecological context: anagent-basedmodel. Pp.45-76in Dynamic* in plant and fish communities. This work is partly supported Human anil Primate Societies: Agcnt-hosea1 Mode/ing ofSocial and by a grant from the A. H. Schultz foundation. SpatialProcesses, T. Kohler and G. Gumerman. eds. Oxford Univer- sity Press. New York. Literature Cited Plotkin,H.C.,and F..1.Odling-Smee. 1981. A multiple-level model of evolution and its implications for sociobiology. Behav. Brain Sci. 4: Boerlijst,M. C.,and P. Hogeweg. 1991. Self-structuringandselection: 225-268. spiral wavesasasubstrateforevolution. Pp. 255-276inArtificialLife Prati, D., 1). Matthies,and B. Schmid. 1997. Reciprocal parasitization II. C. G. Langton, C. Taylor. J. D. Farmer, and S. Rusmussen. edv inRhinanthus serotinus: amodel systemof physiological integrationin Addison-Wesley, Redwood City, CA. clonal plants. Oiko.s 78: 221-229. Bradley, B.J. 1999. Levelsofselection, altruism and primate behavior. Ruzzante. D. E., and R. W. Doyle. 1991. Rapid behavioral changes in Q. Rev. Biol. 74: 171-194. medaka caused by selection for competitive and noneompetitive Caldecott, J. O. 1986. Mating patterns, societies and ecogeography of growth. Evolution 45: 1436-1946. macaques. Anini. Behav. 34: 208-220. Ruzzante,D. K.,and R. \V. Doyle. 1993. Evolutionofsocial behaviour Chase, I. D. 1985. The sequential analysis of aggressive acts during m a resource rich, structured environment: selection experiments with hierarchy formation: an application of the 'jigsaw' pu/zle approach. medaka. Evolution 47: 456-470. ,4/i/Hi. Behav. 33: 86-100. Rypstra,A. L. 1993. Prey size,socialcompetition,andthedevelopment de Waal, F. B. M., and L. M. Luttrell. 1989. Towards a comparative ofreproductivedivisionoflaborin social spidergroups.Am. Nat. 142: socioecology of the genus Macaco: different dominance styles in 868-SXO. rhesus and stumptuil monkeys. Am. J. Primatol. 19: 83-104. Sober, E., and D. S. Wilson. 1998. Unto Others. Harvard University Hamilton,W.D. 1971. Geometry fortheselfishherd.J. Theor. Biol 31: Press. Cambridge. 295-311. Thierry, B. 1985a. Patterns ofagonistic interactions in three species of Hand..1.L. 1986. Resolutionofsocialconflicts:dominance,egahturianism. macaque (Macaco mulatto, M. fascicu/aris. M. tonkeana}. Aggress. spheres ofdominance, andgametheory. Q. Rev. Biol. 61: 201-220. Behav. 11: 223-233. Hemelrijk, C. K. 1990a. A matrix partial correlation test used in inves- Thierry, B. 19851). Social development in three species ofmacaque (Ma- Hemetlleivgreailtj.ikoJ,n.s(T.o'h.feKor.re.ci1p9Bri9oo0lc1.)i.ty14a3Mn:odd4eo0tl5hs-e4ro2fs0.o.cainadl tiensttesrafcotr.iornecpiaptrtoecrintsy,atunaidgirroeucp- Thieemriatrceyaa,qnnuBi.elat1mi9oa9d,0e.Ml.. fJF.aesecTidhcebuoalracr.kisB,liooMlo..pt1bo4en5tk:weea5ne1an]1.-k5iB2ne1sh.haivp.aPnrdocdeosmsi1n1a:nc8e9:-9t5h.e Btieohnaavl.a3n9d: o1t0h1e3r-1s0oc2i4a.l interaction patterns at a group level. Anim. Thierry, B., A. N. Iwaniuk, and S. M. Pellis. 2000. The influence ol Hemelrijk, (.'. K. 1999a. Effects ofcohesiveness on intersexual domi- phytogenyon the social behaviourofmacaques(Primates: Cercopithe- ennatnitciees.relPapt.io5n2sh4i-p5s34andinsApialtviualncis'tsruicntuArretiafimcoianlgLigfreo.upF-ilftihvinEgurvoipretuaanl Ulbrciicdhae,,Kg.,enJu.sR.MaHceancsoc)h.elE,tFh.olJoelgtysc1h0.6:an7d13C-.7W2i8s.sel. 1996. Modelling ConferenceonArtificialLife. Vol. 1674.D.Floreano,J-D. Nicoud.and individual variability in asocial spidercolony (Stegodyphusditmieola: F. Mondada, eds. Springer Verlag, Berlin. Eresidae) in relation to food abundance and its allocation. Proc. 13th Hemelrijk,C. K. 1999h. An individual-oriented model on theemergence Int. Congr. Arachnol., Geneva, 3-8 Sept. 1995. Rev. SuisseZoo/, vol. ofdespoticandegalitariansocieties. Prt>c. R. Sue. Lund. B266: 361-369. hors serie: 661-670. Hemelrijk.C.K.2000. Towardstheintegrationofsocialdominanceand van Schaik. C. P. 1989. The ecology of social relationships amongst spatial structure. Anim. Behav. 59: 1035-1048. female primates. Pp. 195-218 in Comparative Socioecology, the Be- Hogeweg. P. 1994. Multi-level evolution: replicators and the evolution liovioiiral Ecology ofHumans and Other Mammals, V. Standen and ofdiversity. Phyxica D 75: 275-291. G. R. A. Foley, eds. Blackwell. Oxford. Honk,C.v.,andI'.Hogeweg. 1981. Theontogenyofthesocialstructurein vanSchaik.C.P..andJ. A.R.A.M.vanHooff. 1983. On the ultimate acaptiveBomhux lerrestrixcolony. Behav. Eeol. Sociohiol. 9: 111-] 19 causes of primate social systems. Behaviour85: 41-117. Huynen,M.A.,and P. Hogeweg. 1994. Patterngenerationin molecular \ehreiicamp, S. L. 1983. A model for the evolution ofdespotic versus evolution: exploitation of the variation in RNA landscapes. J. Mol. egalitarian societies. Anim. Behav. 31: 667-682. Evol. 39: 71-79. Wilson, I). S. 2001. Cooperation and altruism. Pp. 222-231 in Evolu- K.iiitlin in. S. 1995. At Home in the Universe. The Search for La\vs of tionary Ecology: Concepts and Case Studies. C. W. Fox. D. A. Roff. CoinplcMt\. Penguin Group. London. and D. J. Fairbairn, eds. Oxford University Press. Oxford.