individual versus social complexity, with particular reference to ant

Biol. Rev. (2001), 76, pp. 211237 Printed in the United Kingdom # Cambridge Philosophical Society 211

Individual versus social complexity, with

particular reference to ant colonies

CARL ANDERSON* and DANIEL W. MSHEA

Department of Biology, Duke University, Durham, NC 277080338, USA

(Received 6 April 2000, revised 14 November 2000; accepted 14 November 2000)

ABSTRACT

Insect societies colonies of ants, bees, wasps and termites vary enormously in their social complexity.Social complexity is a broadly used term that encompasses many individual and colony-level traits andcharacteristics such as colony size, polymorphism and foraging strategy. A number of earlier studies haveconsidered the relationships among various correlates of social complexity in insect societies ; in this review,we build upon those studies by proposing additional correlates and show how all correlates can beintegrated in a common explanatory framework. The various correlates are divided among four broadcategories (sections). Under polyphenism we consider the differences among individuals, in particularfocusing upon caste and specialization of individuals. This is followed by a section on totipotency in whichwe consider the autonomy and subjugation of individuals. Under this heading we consider various aspectssuch as intracolony conflict, worker reproductive potential and physiological or morphological restrictionswhich limit individuals capacities to perform a range of tasks or functions. A section entitled organizationof work considers a variety of aspects, e.g. the ability to tackle group, team or partitioned tasks, foragingstrategies and colony reliability and efficiency. A final section, communication and functional integration,considers how individual activity is coordinated to produce an integrated and adaptive colony. Within eachsection we use illustrative examples drawn from the social insect literature (mostly from ants, for which thereis the best data) to illustrate concepts or trends and make a number of predictions concerning how aparticular trait is expected to correlate with other aspects of social complexity. Within each section we alsoexpand the scope of the arguments to consider these relationships in a much broader sense of sociality bydrawing parallels with other social entities such as multicellular individuals, which can be understood as societies of cells. The aim is to draw out any parallels and common causal relationships among thecorrelates. Two themes run through the study. The first is the role of colony size as an important factoraffecting social complexity. The second is the complexity of individual workers in relation to the complexityof the colony. Consequently, this is an ideal opportunity to test a previously proposed hypothesis that individuals of highly social ant species are less complex than individuals from simple ant species in lightof numerous social correlates. Our findings support this hypothesis. In summary, we conclude that, ingeneral, complex societies are characterized by large colony size, worker polymorphism, strong behaviouralspecialization and loss of totipotency in its workers, low individual complexity, decentralized colony controland high system redundancy, low individual competence, a high degree of worker cooperation when tacklingtasks, group foraging strategies, high tempo, multi-chambered tailor-made nests, high functional integration,relatively greater use of cues and modulatory signals to coordinate individuals and heterogeneous patternsof worker-worker interaction.

Key words : Ants, insect societies, individual complexity, social complexity, polyphenism, totitpotency, workorganization, functional integration, sociality.

* Current address : LS Biologie I, Universita$ t Regensburg, Universita$ tsstrasse 31, D-93040 Regensburg, Germany.Correspondence: C. Anderson, LS Biologie I, Universita$ t Regensburg, Universita$ tsstrasse 31, D-93040 Regensburg,

Germany.Tel : 49 (0)941 943 3293;Fax: 49 (0)941 943 3304.E-mail : carl.anderson!biologie.uni-regensburg.de

212 Carl Anderson and Daniel W. McShea

CONTENTS

I. Introduction ............................................................................................................................ 211II. Polyphenism............................................................................................................................. 214

(1) Polymorphism ................................................................................................................... 214(2) A continuum of specialization........................................................................................... 216(3) Differentiation................................................................................................................... 216

III. Totipotency ............................................................................................................................. 217(1) Totipotency, morphological skew and intracolony conflict............................................... 217(2) Physiological and morphological constraint...................................................................... 218(3) Individual complexity ....................................................................................................... 218

IV. Organization of work............................................................................................................... 219(1) Groups and teams ............................................................................................................. 220(2) Task partitioning............................................................................................................... 221(3) Nest complexity................................................................................................................. 221(4) Defence.............................................................................................................................. 222(5) Foraging strategies ............................................................................................................ 223(6) Tempo, reliability and efficiency....................................................................................... 223(7) Intermediate-level parts .................................................................................................... 226

V. Communication and functional integration............................................................................. 227(1) Signal range and system connectedness ............................................................................ 227(2) Signals versus cues .............................................................................................................. 228(3) Modulatory signals............................................................................................................ 229(4)Integration and connectedness............................................................................................ 229

VI. Discussion ................................................................................................................................ 230VII. Conclusions .............................................................................................................................. 232

VIII. Acknowledgements .................................................................................................................. 232IX. Appendix ................................................................................................................................. 233X. References................................................................................................................................ 233

I. INTRODUCTION

In insect societies colonies of ants, bees, wasps andtermites significant correlations exist among cer-tain properties of the colony and of the individualsthat comprise it. For example, the degree ofbehavioural specialization and intracolony colonyconflict among the workers both correlate withcolony size and colonies in which the nest is simple instructure also tend to have simple communicationsystems. In the most recent review of the subject,Bourke (1999) used the term simple to describecolonies with few or no morphological differencesbetween reproductive individuals and workers, nophysical caste polymorphism among the workersand relatively simple nests and communicationsystems. (The converse characteristics apply forcomplex colonies.) Here, we initially adopt hisview of complexity as a small cluster of correlates.However, as we proceed through the various sectionswe extend the investigation and consequently ex-pand Bourkes (1999) cluster to include aspects ofcolony life not previously considered.

The above usage of complexity is generallyconsistent with that elsewhere in the social insect

literature, where the term been used to refer to anumber of variables, such as mean size of behav-ioural repertoire (Cole, 1985), body size and percapita productivity (Karsai & Wenzel, 1998), num-ber of castes (Oster & Wilson, 1978), negentropy ofnestmate recognition signals (Jaffe, 1987; Jaffe &Perez, 1989), degree of cooperation and coordinationamong workers (Beckers et al., 1989; C. Anderson,N. R. Franks & D. W. McShea, in preparation),reproductive dimorphism (Peeters, 1997; Bourke,1999) and colony size (Wilson, 1971; Peeters, 1997;Karsai & Wenzel, 1998; Bourke, 1999). This studyexpands the scope of certain related treatments ofsocial insects (e.g. the above references andMichener, 1974; Bonner, 1988; Alexander, Noonan& Crespi, 1991; Seeley, 1995) and our usage of theterm social complexity is more inclusive ; indeed,our hope is to develop a common scheme that showshow many of these variables, along with others, arerelated to each other.

Our usage also suggests a connection with thebroader literature on hierarchy in biology, wherethere has been considerable interest in the factorsaffecting degree of emergence, or individuation, ofentities at various hierarchical levels, such as the

213Individual versus social complexity

Table 1. Social correlates of individual and social complexity in insect societies. The correlates are intended to illustrategeneral trends only and not to encompass the behaviour of every social insect species. All of these trends are discussed inthe text. Numbers in parentheses refer to the subsections of the sections (in bold) where the particular trait or correlate isprimarily discussed. (0) refers to the introduction of a section

Simple societies U Complex societies

I. Introduction(0) Colony size Low High

II. Polyphenism(1) Worker polymorphism Low High(2) Individual specialization NoneUbehaviouralUphysiologicalUmorphological(2) Type of specialization Temporary Permanent

III. Totipotency(1) Functionality of ovaries High Low(1) Morphological skew Low High(1) Worker policing Absent Present(1) Intracolony conflict High Low(2) Physiological constraint Low High(3) Individual complexity High Low

IV. Organization of work(0) Colony control Centralised Decentralised(0) System redundancy Low High(0) Homeostasis Low High(1) Groups and teams Absent Present(2) Task partitioning Absent Present(3) Nest complexity Low High(3) Colony-constructed nest No Yes(3) Number of chambers One Many(4) Foraging strategy IndividualU tandem runningUmassU trunk trailU group hunting(5) Defence Generalist non-sacrificial workers Specialist sacrificial defenders(6) Tempo Low (cool ) High (hot )(6) Individual competence High Low(6) Most complex task type IndividualU groupU team & partitioned(6) Efficiency High Low

V. Communication and functional integration(1) Average system connectedness High Low(2) Use of cues Low High(3) Use of modulatory signals Low High(4) Heterogeneity of interaction Low High

eukaryotic cell and multicellular organisms as well asentities at the colony level [e.g. Spencer, 1904;Beklemishev, 1969; Pattee, 1970; Boardman &Cheetham, 1973; Leigh, 1983; Salthe, 1985, 1993;Buss, 1987; Bonner, 1988, 1998; Wimsatt, 1994;Maynard Smith & Szathma! ry, 1995; Simon, 1962;Keller, 1999; Michod, 1999; Wilson, 1999; McShea,in press b ; see also supplement issue to AmericanNaturalist (1997) 150(1)]. In present terms, aeukaryotic cell can be understood as a complexsociety of prokaryotic cells and a highly individu-ated multicellular organism as a complex society ofeukaryotic cells. Here, we explore this connectionfurther, raising the possibility that at least some of

the same principles governing colony complexity inthe social insects apply and that a similar cluster ofcorrelates will be found across the hierarchicalspectrum.

We divide the whole suite of correlates consideredin this study among four major sections. Section II,polyphenism, considers the differences amongindividuals and in particular considers the conceptof caste. Section III, totipotency, considers thecomplexity of individuals within a colony (asopposed to the complexity of the colony itself),especially their totipotency with respect to behav-ioural and reproductive capacities. Section IV,organization of work, considers how the nature of


work itself may change with increasing colonycomplexity and the important changes observed inthe way that individuals cooperate in order to tacklecolony tasks. Lastly, Section V, communication,considers how individuals communicate with eachother and coordinate their activities. These cate-gories are by no means exclusive and the fact thatmany of the different characters and traits arecorrelated means that there is some overlap acrossthe sections. Within each section, we identify anumber of new correlates of interest and, withexamples, expand Bourkes (1999) cluster. Whereverpossible we attempt to identify any causal relation-ships and put these correlations or causal mech-anisms in a broad context of sociality withcomparisons to other social systems such asmulticellular organisms. The principal concepts,trends and predictions are summarized in Table 1.

Two major themes are found linking all thesesections. The first is the influence of colony size, oneof the most important correlates when consideringsocial complexity (e.g. Wilson, 1971; Peeters, 1997;Karsai & Wenzel, 1998; Bourke, 1999). The secondis the relationship between the complexity of theindividual and that of the colony. This presents anideal opportunity to test Jaffe & Hebling-Beraldos(1990) hypothesis that individuals of highly socialant species are less complex than individuals fromsimple ant societies , in a much broader context, i.e.set of correlates, than was previously possible. Wesuggest that, generally speaking, as social complexityincreases, there is indeed a correlated decrease inindividual complexity. In other words, there is adecline in independence or autonomy of the in-dividual and in its ability to function on its own.

Some explanatory notes are required beforeproceeding. First, describing the many pair-wiserelationships within even a moderate-sized group ofcorrelates in an orderly way is difficult. We havetaken an approach in which we begin with a smallnumber of basic correlates in the first section (e.g.colony size and differentiation) and build upincrementally, in each section adding a small suite ofadditional variables and discussing their relations-hips to those in previous (sub)sections. Importantly,the order in which variables are discussed does notreflect causal primacy.

Second, ideally we would like to be able to mapsome of the social correlates onto a phylogeny. Thiswould enable us for instance to evaluate whether simple maps onto primitive insect societies andcomplex maps onto more derived and advancedsocieties. A phylogeny could also be used to assess

whether complexity has any tendency to increasein evolution, a tendency that is widely acknow-ledged but for which little empirical evidence exists(McShea, 1996 and references therein). Unfortu-nately, at the present time a suitable phylogeny doesnot exist and as such thus we shall avoid the twovalue-laden, yet often-used, terms primitive andadvanced (see Sherman et al., 1995). Third, ourtreatment of complexity in social insects is restrictedmainly to ant societies, where the best data seemto be available.

Fourth, implicit in our understanding is our viewof social complexity as a continuum running fromsimple to complex. However, occasionally, it will beconvenient to frame the discussion solely as a contrastof these two extremes without discussing the wholesuite of intermediates.

II. POLYPHENISM

The following section considers differentiationamong individuals, i.e. temporary and permanentdifferences in morphology (polymorphism), physi-ology and behaviour among individuals.

(1) Polymorphism

Simple societies are composed of monomorphicindividuals whereas complex societies contain apolymorphic workforce (Michener, 1974; Wheeler,1986; Ho$ lldobler & Wilson, 1990; Peeters, 1997;Bourke, 1999). Polymorphism has evolved inde-pendently at least eight times (Ho$ lldobler & Wilson,1990) but only 15% (44}297) of all ant genera ex-hibit some degree of worker polymorphism (Oster& Wilson, 1978: p. 4). Importantly, polymorphismtends only to be associated with species that havelarge colonies (Bourke, 1999: Table 1). AnalyzingJaffes (1987: Table 1) data (although not takinginto account phylogenetic relationships) we found ahighly significant relationship between log

"!(colony

size) and log#(variation of worker size), specifically :

log"!

(colony size) 1.11log#(variation of worker

size) 1.85 (F 19.78, P! 0.0002; solid line in Fig.1A). Jaffe (1987) did not specify how he quantifiedthis worker size variation. Non-linear regressionsgive better fits to the data and the linear fit is stillsignificant when the two obvious outliers areremoved (F 8.15, d.f. 22, P! 0.01; dashed linein Fig. 1A).

It is unfortunate that the analysis cannot easily beextended to quantify the degree of polymorphism


12

10

8

6

4

2

1 2 3 4 5 6 7Log10 (colony size)

Log

2 (v

aria

tio

n in

wo

rker

siz

e)A

70

60

50

40

30

20

10


Pro

po

rtio

n (

%)

C

GH

TT

M

GM

TR

I1 2 3 4 5 6 7

Log10 (colony size)

Fora

gin

g s

trat

egy

B

8

6

4

2


Ru

nn

ing

sp

eed

(cm

/s)

D

Acts self-grooming:Time inactive:

7

Fig. 1. (A) Log#(variation in worker size) versus log

"!(colony size) (data from Jaffe, 1987: Table 1). A small amount

of scatter has been added to the data to separate points. The solid line is the least-squares fit to the data while thedashed line is the least-squares fit when the two obvious outliers have been removed. (B) Foraging strategy changesfrom individual foraging (I) to group hunting (GH) as colony size increases (data from Beckers et al., 1989). Otherforaging strategies : TR, tandem running; GM, group}mass foraging; M, mass foraging; TT, trunk trail. (C) Prop-ortion of time spent inactive and proportion of acts spent self-grooming both decrease with increasing colony size(data from Schmid-Hempel, 1990). Reprinted by permission ! University of Chicago Press. (D) Running speedincreases with log

"!(colony size). Running speed data are listed in the Appendix.

more precisely. In ants, morphological castes areproduced developmentally by altering the relativegrowth rates of different body parts (see Bourke &Franks, 1995 and references therein). The relation-ship between just two body parts may be isometric,di- or tri-phasic allometric, or even completelydimorphic (Wilson, 1953; Wheeler, 1991) and it isnot even always possible to distinguish differentcastes in a supposedly multi-modal species (e.g.Daceton armigerum, Moffett & Tobin, 1991). Inaddition, within a species different relationships mayexist between many pairs of body parts (includinginternal organs such as brains : see Jaffe & Perez,1989) may exist between many pairs of body partsmaking any overall measure of degree of poly-morphism a challenging, but not necessarily in-tractable, multidimensional concept.

Only very recently have researchers begun toappreciate that polymorphism does not necessarily

mean a difference in size among individuals : sizemay be similar but shape may differ. This is clearlythe case when considering winged queens versuswingless workers of similar size (A. F. G. Bourke,personal communication). However, shape may alsovary among individuals in which both queens andworkers are winged. This was recently shown to bethe case in Apoica pallens, a polistine wasp in whichthe overall size of queens and workers does not differ,but queens are smaller than workers anteriorly yetsignificantly larger posteriorly (Jeanne, Graf &Yandell, 1995). Importantly, this research suggeststhat morphological castes may evolve without adivergence in body size. Such studies may show thatmorphological differences and caste in insectsocieties may be greater and more widespread thanpreviously thought. We predict that the greatestdifferences among individuals not just in size(including isometric allometry) but also in shape


will be found in the species with the largest colonysize and suggest that polymorphism is a significantfacet of social complexity (contra Wilson, 1953: p.153).

(2) A continuum of specialization

An additional facet of polyphenism is workerspecialization. We suggest that simple societies arecomposed of generalist workers whereas complexsocieties contain workers that are specialized in avariety of ways, i.e. behaviourally, physiologically,or morphologically (Oster & Wilson, 1978;Anderson, 1998), to allow a certain amount ofdivision of labour. Consequently, the ant Amblyoponepallipes, which completely lacks division of labour(Traniello, 1978), excepting reproductive division oflabour, must class as one of the simplest societiesfrom this aspect. We suggest that the various forms ofspecialization form a continuum (Table 1). At thelower extreme are true generalists, species such asAmblyopone pallipes, in which the probability ofperforming a task is independent of age. In otherspecies, however, individuals concentrate their ac-tivities on a subset of the colonys repertoire and thusexhibit temporal specialization and division oflabour. This temporal specialization may only betemporary age or temporal polyethism may occurin which individuals tend to pass through anaverage sequence of tasks over time, or individualsmay switch tasks through recruitment when the needarises but we still consider them specialists.

We suggest that a greater degree of specializationoccurs when physiology imposes temporary behav-ioural specialization by making individuals par-ticularly well adapted to certain tasks. For instance,between the 6th and 14th day of their lives, honeybee workers produce royal jelly (Manning, 1975)which effectively forces them to specialize in feedingqueen larvae. As the workers age, special cells intheir abdomens start to secrete wax and thus theseindividuals become naturally suited to a comb-building role. Finally, the extreme of this continuumis associated with morphological castes. In thissituation, individuals are not subject to temporarychanges in their abilities to perform certain tasks butto permanent differences in skills and abilities. Forinstance, a morphologically specialized soldier islikely to be inherently better at defence than otheractivities (e.g. brood care) throughout its life. InCrematogaster smithi ants there are two worker castes : large and small . Large workers appear to bespecialized in the production of trophic eggs, i.e.

unfertilized worker-laid eggs used as food. Effect-ively, these workers specialize in turning proteinfrom perishable food such as insect prey intointernally stored eggs that they may store for severalweeks (Heinze, Cover & Ho$ lldobler, 1995; Heinze,Foitzig & Oberstadt, 1999).

(3) Differentiation

Three main arguments support the notion thatdifferentiation should be correlated with colony size.As will be seen, they are quite general applying inprinciple to complex societies of prokaryotic cells(i.e. eukaryotic cells) and of eukaryotic cells (i.e.multicellular organisms) as well as to colonies ofmulticellular individuals. First, in a large aggregateof generalist individuals, performance can often beimproved if individuals differentiate in order tospecialize on particular tasks. In particular, dif-ferentiation is expected when individual specialistsare able to perform functions more efficiently thannon-specialists (Smith, 1776; Bonner, 1988) or whensynergistic interactions among specialists improveperformance (Corning, 1983). A requirement is thatit must be possible for individuals to share resources(Harvell, 1994) and for tasks to be coordinated(Oster & Wilson, 1978; Corning, 1983; Bell, 1985).What factors control the degree of differentiation ornumber of specialist types? One possible constraintfollows from a model developed by Oster & Wilson(1978) relating numbers of castes and tasks in socialinsects. Generalizing their result, selection is ex-pected to favour differentiation of types until thenumber of types equals the number of tasks to beperformed; it follows that the number of types couldbe limited by the number of tasks. However, thenumber of tasks doubtless reflects the number ofselection pressures, including the number of selectiveopportunities and we have no reason think that theseare sufficiently limited so as to impose a significantconstraint, especially considering that tasks can oftenbe readily subdivided into subtasks (see below).

Another possibility is that the number of types islimited by the number of individuals available todifferentiate, so that each individual is specializedfor a unique function. In this case, the number oftypes would be expected to increase directly with thenumber of individuals and therefore with colonysize. However, other constraints tend to reduce therate of increase. For example, in large aggregates, aspecialized type is rarely a single individual, possiblybecause single individuals are often too small to


make a significant contribution to the whole (Bell &Mooers, 1997). Also, in small aggregates, fluct-uations in demand for specialized tasks on theaggregate as a whole are often severe, in relativeterms and as a result, specialized individuals willsometimes be left idle. Thus, differentiation is notexpected even to begin until colonies reach at least amoderate size (Bell & Mooers, 1997). Consistentwith this, in volvocacean algae (e.g. Volvox), dif-ferentiation of cells (into somatic and germ cells) isevident only in species with colony sizes of 64 orgreater (e.g. Eudorina, Bell, 1985). Differentiationmay also be limited by constraints imposed bydevelopment (Oster & Wilson, 1978), environment(Schopf, 1973), life history, etc.

A second argument is that differentiation mayactually be required at large group sizes. Bonner(1988) proposed that increased size reduces surface-area-to-volume ratios, which changes the perform-ance requirements of certain tasks, which in turnfavours the evolution of specialized devices composedof differentiated lower-level individuals. For ex-ample, single-celled organisms can obtain food andoxygen by diffusion alone but larger multicelledorganisms have cells differentiated to function inspecialized devices, circulatory systems, for food andoxygen delivery.

Also, a connection between differentiation andgroup size is a prediction of Spencers (1904)metaphysic, in particular of what he calls theinstability of the homogeneous. Spencer argues thatin an aggregate, the component elements are subjectto different forces by virtue of their differing locationswithin the aggregate internal versus external, forexample and these differences in forces producedifferences in structure. The greater the size of theaggregate, the greater the range of variation ofenvironments within it and therefore also of theforces producing differentiation. Thus, even in theabsence of selection, larger systems are expected tobecome more differentiated.

The relationship between aggregate size anddifferentiation has been studied in multicellularorganisms and the results are consistent withexpectations. In particular, Bell & Mooers (1997)have demonstrated a significant correlation betweenbody size and number of cell types (see also Bonner,1988). They further found that the number of celltypes increased very slowly with body size ; assuminga log-linear relationship, they estimated an exponentof 0.056, much less than 1.

So far we have discussed the complexity of thecolony in the sense of degree of differentiation among

individuals within it. We now turn to the question ofthe complexity of the individuals themselves.

III. TOTIPOTENCY

Workers in complex societies tend to be lesstotipotent (Crespi & Yanega, 1995; Bourke, 1999).Crespi & Yanega (1995: p. 110) define totipotencyas the potential, throughout life, to express the fullbehavioural repertoire of the population (even ifnever actually expressed) and the ability to produceoffspring like oneself, exhibiting the full behaviouralrepertoire of the population without help. Thus,reduction in totipotency includes not only loss ofreproductive potential but also of behaviouralrepertoire. A distinction must be made between twotypes of reduction in repertoire. First, individualsmay be subject to socially mediated and temporarybehavioural specialization in which they retain theirbehavioural flexibility and can potentially switch toother tasks (see Section II). Second, an individualsrepertoire may be reduced due to physiological andmorphological constraint. That is, differentiation, interms of morphology or physiology, causes a dif-ference in abilities to perform certain tasks. Here,only the latter is part of totipotency.

(1) Totipotency, morphological skew andintracolony conflict

One of the most significant losses of totipotency is theloss of functional ovaries and in some cases,complete loss in workers of some complex societies.Oster & Wilson (1978: p. 101; see also Noll, 1999)demonstrated a positive relationship between mono-morphism of the colony and possession of ovaries byworkers concluding that to develop into an extremecaste is to surrender reproductive potential . Thus, itappears there is a correlation between different-iation, specifically polymorphism (Section II) andtotipotency, in the sense of worker reproductivepotential.

Bourke (1999) studied morphological skew ininsect societies. Morphological skew is the degree ofmorphological dimorphism, usually external, be-tween reproductives and workers. It is a consequenceof specialization of queen and worker roles andreduced worker reproductive potential. For instance,workers lacking ovaries, a form of extreme skew,have been reported in the genera Solenopsis, Pheidole,Monomorium, Tetramorium and Eciton and in the


subfamilies Myrmicinae and Ecitoninae (see Villet,Crewe & Duncan, 1991). Bourke (1999) showed thatspecies with large colonies had a high degree ofmorphological skew. Alexander et al. (1991) pro-posed a causal relationship between colony size,worker reproductive potential and morphologicalskew. Briefly, in large colonies, due to competitionfrom the many other workers, any particularindividual has a small chance of replacing the queenand thus selection maintaining reproductive po-tential will be weak. Bourke (1999) extended theseideas to consider the relationship between colonysize and one aspect of intra-colony conflict : workerpolicing. Worker policing is the set of behaviours inwhich workers search for and eat eggs laid by otherworkers (Ratnieks, 1988; Ratnieks & Reeve, 1992;Keller & Reeve, 1999). Worker policing is acomponent of totipotency because it prevents work-ers from reproducing. Bourke (1999) argued thatindependent of mating frequency (which affectsaverage worker-worker relatedness), it is easier forworker policing to invade large colonies. That is, insmall colonies it is easier for an individual tomonopolise reproduction and so there is greaterconflict among individuals fighting for reproduction(Ratnieks & Reeve, 1992). However, in largecolonies, in which each individual has a small chanceof personal reproduction, individuals are more likelyto favour worker-policing and fitness gains throughinclusive fitness from the queen(s) than personalreproduction. Bourke & Ratnieks (1999) consideradditional conflict in insect societies : worker sup-pression of new queen production, which is predictedto increase with a reduction in reproductive skew.

(2) Physiological and morphologicalconstraint

Other forms of physiological differentiation, inaddition to loss of ovaries, can impose behaviouralspecialization. So, honey bees producing royal jellyshould specialize upon feeding queen larvae andhoney bee workers producing wax should specializeon comb building. It is conceivable that non wax-producing workers could remove the wax from thewax-producing bees and use it themselves to buildthe comb. [This is the situation with propolis, asticky tree resin, used by honey bees. Propolisforagers cannot unload themselves and require otherbees to unload them. These unloaders and not theforagers, use the material to mend cracks in the nest(Ratnieks & Anderson, 1999a).] However, it is likelyto be far more efficient if the wax-producing bees

who can remove the wax themselves remove andbuild with the material themselves. In effect, as anindividual becomes physiologically specialized, itbecomes more subjugated and thus less of anindependent individual. This is more so withmorphological differentiation. In this situation,individuals are subject to permanent restrictions intheir abilities. One of the more extreme examples ofreduction of individuality of workers occurs whensoldiers physically cannot feed themselves and mustrely on trophallaxis for survival (e.g. Daceton arm-igerum majors ; Wilson, 1962).

The greatest loss of individuality must surely bethat associated with suicidal and sacrificial de-fenders. These traits are only expected to occur incomplex societies. Sting autotomy self-amputationof a sting (usually barbed) is found in varioussocial insect taxa among the ants, bees and wasps(Hermann, 1971, 1981). One predicts that barbedstings will be used in defence only in species thatform new colonies in swarms such as honey bees andsome tropical wasps [i.e. complex societies]. In verysmall colonies [simple societies] workers are toovaluable for suicidal attacks to be beneficial (Alexander et al., 1991: p. 31). However, alternativesacrificial and suicidal behaviour does occur in antssuch as in the walking bombs (Oster & Wilson,1978: p. 226) of Globitermes sulfureus and Camponotussaundersi ants who literally explode in front of theirattackers. These ants use hydrostatic pressure toburst their gaster and mandibular glands releasingstick fluids, a process termed autothysis byMaschwitz & Maschwitz (1974). Although theseadaptations, such as barbed stings and speciallyadapted mandibular glands, may be morphologicaland physiological adaptations, they are present inthe whole workforce and therefore are not part ofdifferentiation (Section II). Our point is that thepossession and use of these adaptations constitutes areduction in individual autonomy and that theseadaptations are expected only in complex societies.

(3) Individual complexity

Reduction of totipotency has a number of possiblecorrelates, which we predict will be manifest at allhierarchical levels. First, there is a direct correlationwith colony size. In small colonies, each individualin the colony is a scarce and valuable resource andtherefore remains totipotent. Second and moregenerally, we expect a correlation between loss ofbehavioural and reproductive capacities on the onehand and loss of morphological structures and


physiological capacities on the other. Further, all ofthese should be correlated with colony size. As sizeincreases and individuals specialize, they becomedependent on others, or more precisely on the colonyas a whole, for the performance of various survival-related functions. An individual that is specializedfor defence, for example, relies on other specialists forforaging, colony-homeostatic activities and so on. Asa result, the number of different functional demandson specialist individuals is reduced, along with theselection pressures maintaining the morphologicalstructures, behaviours and physiological mechanismsassociated with these activities. [An underlyingassumption is that functions tend to be localized insuch structures and mechanisms (McShea, 2000).]Indeed, selection is expected to favour the loss ofthese structures, behaviours and mechanisms in theinterest of economy (McShea, in press a).

Third, the emergence of function at the level of thecolony requires the loss of some degrees of freedom atthe level of individuals (Guttman, 1969; Pattee,1973; Bar-Yam, 1997). Buss (1987; see alsoMaynard Smith & Szathma! ry, 1995) has pointedout that selection on the colony favours loss ofreproductive capacities in some individuals, asdiscussed above in ants. But more generally, selectionshould act to reduce individuals degrees of freedomin order to limit their interference with the co-ordinated function of the whole. Thus, individualsspecialized for metabolism must not reproduce;likewise, reproductive specialists should not havemetabolic capability, as their metabolic processesare unlikely to be coordinated with and indeed arelikely to interfere with, those of the specialists. Theloss of degrees of freedom may be achieved either bypolicing (Michod, 1999) self-policing or totali-tarian control or, more effectively, by the loss ofmorphological structures, behaviours and physio-logical mechanisms associated with superfluouscapacities.

In addition to the findings in social insectsdiscussed above, a number of observations in colonialmarine invertebrates are consistent with these ex-pectations. For example, Wood, Zhuravlev &Debrenne (1992) noted that individuals in coloniesare often smaller and less complex than relatedsolitary individuals and Beklemishev (1969) notedthat zooids in colonial forms are simpler comparedthose in related free-living species. At the cellularlevel, metazoan and land plant cells contain fewermacroscopic morphological structures than free-living eukaryotic cells (McShea, in press a). Extremeexamples include mature human haemocytes, which

contain almost no macroscopic structures and free-living Euglena, which have a variety of structures(e.g. nucleus, mitochondria, plastids, a flagellum,contractile vacuoles and so on). These extreme casesaside, even the more typical metazoan cells are, atleast impressionistically, less complex in this sense,on average, than typical free-living cells (Gerhart &Kirschner, 1997: p. 242).

These observations are based on comparisonsbetween extreme cases of social complexity, betweenfree-living forms in which social complexity isessentially zero and individuals in colonies, in whichit is present to some degree. There has been nodemonstration yet that losses of structure accompanyincremental increases in social complexity. In otherwords, it has not been established that the two arecorrelated as continuous variables, or that such lossesare connected with any of the various correlates ofsocial complexity, such as degree of differentiation.Also, most of the evidence for losses is based onmorphology (e.g. ovaries in ants, parts in cells, etc.),but the expectation of loss at the level of theindividual extends to behaviours and physiologicalmechanisms as well. Individual zooids in marineinvertebrate colonies and cells in metazoans andland plants show little motor behaviour, but possiblelosses of physiological mechanisms could be investi-gated. Finally, in social insects, we predict thatindividuals in complex colonies should have fewerbehaviours, on average, than those in simple col-onies. Methods for counting behaviours have beendeveloped (e.g. Wilson & Fagen, 1974; Fagen &Goldman, 1977) and thus testing could be straight-forward.

IV. ORGANIZATION OF WORK

The complexity of the colony also has consequencesfor the organization of work within the colony. Inaddition to worker policing (see Section III),reduced intracolony conflict in complex societies hasanother important consequence: the type of socialcontrol. In small simple societies, as in the ponerineant Dinoponera quadriceps (Monnin & Peeters, 1999),there is much aggression and direct control of colonyactivity by the queen or gamergate . [Gamergatesare mated reproductive workers in queenless ants(Peeters, 1997).] In effect, the reproductives cen-trally control simple conflict-ridden societies. How-ever, in the relatively harmonious larger societies,colony control and decision-making tends to bedecentralized. That is, workers react to local in-


formation and configurations and so are self-organized (e.g. The! raulaz & Bonabeau, 1995;Bonabeau et al., 1997; Bonabeau, Dorigo &The! raulaz, 1999; Camazine et al., in press). Cen-tralized control would be difficult or impossible inlarge colonies. However, decentralized control canbe very adaptive at the colony level, even with verylarge colony sizes and importantly does not necess-arily require complexity at the individual level (e.g.Bonabeau, 1998; Bonabeau et al., 1999; Anderson &Bartholdi, 2000).

Interestingly, there are theoretical reasons tosuppose that a large decentralized colony can beboth efficient and reliable (high probability thattasks are completed) even when the individualstackling the tasks are simple and not especiallycompetent. Specifically, when activity is concurrentand individual competence is low, efficiency andreliability can be achieved by having redundancy replication of parts at the subunit level rather thanthe system level (Barlow & Proschan, 1975; Oster &Wilson, 1978; Herbers, 1981; see below). Thus, weexpect many (spare) individuals in the vicinity of atask who are able to tackle that task. Even though allthe individuals may not be fruitfully employed, atleast the task will be completed.

It has been noted that complex societies tend tohave a perennial cycle and in wasps and bees areswarm-founded rather than independent-founded(Bourke, 1999) and so do not go through a bottleneckof small colony size. Not only is colony size lessvariable through the year in more complex societiesbut larger systems are inherently more homeostatic(Wilson, 1971). For instance, there are automaticbenefits to group foraging such as reduced variationin food influx rate (Wenzel & Pickering, 1991) anda greater potential for food storage to supply thecolony during forage dearths (Jeanne, 1991). Insummary, complex societies have a more constantcolony size and structure, have less intracolonyconflict, are inherently more homeostatic and relymore on decentralized control. This has hugeimplications in the organization of work as shownbelow.

(1) Groups and teams

One very apparent aspect of work organization incomplex societies is the degree of cooperation amongindividuals when tackling some tasks. We suggestthat in complex societies tasks may be completed bygroups and teams rather than by individuals.Anderson & Franks (2001) suggest that there arefour types of task in insect societies : individual,

group, team and partitioned tasks. The first threetypes are defined and reviewed in Anderson &Franks (2001). The fourth task type, partitionedtasks, are considered in the next subsection. All fourtask types are summarized in Table 2 and are furtherexplored in C. Anderson, N. R. Franks & D. W.McShea (in preparation).

Briefly, individual tasks are tasks that can sat-isfactorily be performed by an individual, e.g.feeding a larva within a cell. Group tasks requiremany individuals working together for successfultask completion. In a group task, there is no divisionof labour and the behaviour of the individual is thesame as that of the group. For instance, groupambush and bivouac construction in army ants(Schneirla, 1971; Gotwald, 1995), intimidation ofcompetitors inMyrmecocystusmimicusants (Ho$ lldobler,1976, 1979), and stigmergic nest construction intermites (Deneubourg, 1977; Bonabeau et al., 1998b ;see also Karsai, 1999) are all examples of grouptasks. By contrast, Anderson & Franks (2001)propose that a team task requires different subtasksto be performed concurrently, i.e. there is division oflabour. They propose that a task is an item of workthat potentially makes a positive contribution,however small, to inclusive fitness (i.e. direct andindirect fitness) . However, sometimes a subset of thebehaviours required to complete a task may appearas a unit in themselves but cannot contribute toinclusive fitness when performed in isolation. Thisunit is a subtask.

Under the above definition of a team task, onesuch example is nest construction in Oecophylla spp.(weaver) ants. To construct the nest, leaves must beglued with silk while pulled together (subtask 1).The silk is produced by larvae (subtask 2) who areheld above the seam by other workers (subtask 3)(Anderson & Franks, 2001). Anderson & Franks(2001) are careful to point out that team tasks do notrequire that individuals in a team must be membersof different castes (contra Oster & Wilson, 1978: p.151). However, they do highlight that polymorphismand thus presumed inherent differences in abilitiesamong individuals to perform different activities,may favour specialization of castes upon certainsubtasks thus enhancing team task performance.

A team task requires more cooperation than agroup task. This is because in a team workers mustcoordinate their different work contributions. Inturn, a group task requires more cooperation andcoordination than an individual task because agroup task requires the concurrent activity ofmultiple individuals (Table 2). These concepts are


Table 2. Characteristics of various task types. This table defines the various task types. For instance, a team taskrequires a number of different individuals (column 2) performing multiple subtasks (column 3) concurrently (column 4).[From Anderson & Franks (2001).]

TaskNumber ofindividuals

Divided intosubtasks?

Organization ofsubtasks

Individual task Single No Group task Multiple No* Partitioned task Multiple Yes SequentialTeam task Multiple Yes Concurrent

* But concurrency is required.

explored in depth in C. Anderson, N. R. Franks &D. W. McShea (in preparation). We hypothesizethat in simple societies, the most coordinated tasksobserved will generally be individual tasks. Highintracolony conflict, monomorphism and small col-ony size will disfavour more coordinated activity insuch colonies. In more complex societies we alsoexpect to observe individual tasks but additionallywe expect the most coordinated activity to be grouptasks. For instance, we may observe group foragingor group nest construction activity. These grouptasks will be favoured by large colony size in thatmany individuals may be available in the immediatevicinity for recruitment to the task. In even morecomplex societies we expect to observe team tasks.These too will be favoured by large colony size.However, they are more likely to be favoured bydifferences in performance efficiency among individ-uals arising from polymorphism (e.g. N. R. Franks,A. Sendova-Franks & C. Anderson, in preparation)and also from the increased potential for behaviouralspecialization due to increased colony homeostasis.

(2) Task partitioning

Task partitioning (Jeanne, 1986; reviewed inRatnieks & Anderson, 1999a), in addition to groupsand teams, is another task type that is only tenablein relatively complex societies. A task is said to bepartitioned when it is split into a number ofsequential stages and material is passed from oneworker to another. In short, the task is split into anumber of sequential subtasks (Table 2) which areexplicitly linked by the act of material transfer. Forinstance, in the obligate termite-hunting antMegaponera foetans minor workers enter the termitenest, hunt for termites and deposit them in a pileoutside the termite nest. Major workers then trans-port the termite prey back to the ants nest(Longhurst & Howse, 1979). Like group and teamtasks, partitioned tasks require the coordination of a

number of individuals for successful task completion(Table 2).

Partitioning a task may involve a number ofadvantages and disadvantages, which are reviewedin Ratnieks & Anderson (1999a). The most relevantto social complexity are as follows. First, taskpartitioning usually occurs when there is a differencein abilities of individuals to perform different parts ofthe task. For instance, small Oecophylla longinodaworkers are better at milking scale insects than arelarger workers. Small workers collect the honeydewand pass it onto larger workers who are better ableto transport it back to the nest (Ho$ lldobler, 1984b).Thus, division of labour enhances task partitioningand is most likely to occur when the workerpopulation is polymorphic, i.e. in complex societies.Second, when direct transfer occurs searching for atransfer partner takes time (a cost) and averagesearch duration decreases exponentially as thenumber of individuals in the working group in-creases. Consequently, when the group is large, thesequeueing and search delays are negligible (Anderson& Ratnieks, 1999a, b). Third, when group size islarge, workers are significantly better able to assessthe status of the system and make correct recruitmentdecisions to optimize worker allocation and efficiency(Ratnieks & Anderson, 1999b). However, becausework demands are inherently more stable in acomplex society anyway, there is less need forindividuals to switch tasks and this increases ef-ficiency (e.g. Gordon, 1989, 1999b ; Karsai &Wenzel, 1998). Overall, task partitioning has thepotential to greatly enhance task performanceefficiency but mostly so in complex societies.

(3) Nest complexity

Individual ability to cooperate as groups and teamsand partition tasks in complex societies may haveimplications for where they live. We suggest thatcomplex ant societies tend to construct nests for


themselves rather than utilizing existing crevices andthat these nests are relatively complex. One possiblereason is that suitable natural nest cavities are morelimiting for large colonies than for small ; unfortun-ately, testing this is not straightforward. Alter-natively, complex colonies have greater potential fortackling tasks that are beyond the abilities of anindividual, such as the construction of tailor-madenests (e.g. Bonabeau et al., 1999; Anderson & Franks,2001).

What constitutes a complex nest? There are manyfeatures associated with increased nest complexity.Foremost, we see partitioning of the nest into two ormore chambers. This is usually associated with somespatial specialization of colony activity. For instance,in Atta leaf cutter ants some chambers contain thefungal garden and brood whilst others solely containrefuse material (e.g. Weber, 1972). In many antspecies, brood is piled in specific areas of the nest andin termites the queen is physically entombed in onechamber and the eggs that she lays are taken to otherchambers (e.g. Wilson, 1971). In the seed harvesterant Pogonomyrmex barbatus (Gordon, 1999a), piles ofseeds accrue only in certain parts of the nest.

In addition to the points made above, this physicalsegregation of workers in a multi-chambered nesthas an important implication in the degree of controla queen has over a complex society. In a multi-chambered nest, any regulatory pheromones thequeen releases are unlikely to diffuse rapidly through-out the nest and so her ability to control colonyactivity will be greatly reduced. In addition, morecomplex societies exhibit a greater degree of mor-phological skew. Thus, workers could potentiallyvery easily exclude the queen(s) from certain areas ofthe nest by restricting the size of entrances toparticular chambers (Howse, 1970), in the extreme,completely entombing the queen in a single cham-ber. Not only will the diffusion of any of the queenspheromones be restricted but this could also be trueof alarm pheromone: recruitment of additionaldefenders when the colony is attacked could belimited by the layout of the nest. Thus, in complexsocieties we expect some localisation of specializeddefenders near the nest entrance(s) as in Camponotus(Colobopsis) fraxinicola ants discussed in Section IV.4.

Complex nests may contain additional adaptivestructural features not found in simple nests. Starr(1991: p. 523) lists a number of nest features that arebelieved to have originated to serve secondaryfunctions. In ant societies, these include fungusgardens (leaf cutter ants), auxiliary silk bowers usedas shelters for homopterans (e.g. weaver ants) and

windows or tunnels used for the escape of sexuals formating flights (fire ants). Here, we also include thespecialized repletes who act as living storage vesselsin the honey pot ants (Myrmecocystus mimicus,Ho$ lldobler & Wilson, 1990) and turrets at the nestentrance of Acromymrex landolti nests which mayprevent inundation during sheet flooding (Navarro& Jaffe, 1985). Despite the apparent complexity ofsome nests, it should be stressed that it is notnecessary to invoke complexity at the individuallevel during their construction: the complexity maybe an emergent property of a collection of interactingsimple builders (e.g. The! raulaz & Bonabeau, 1995;The! raulaz, Bonabeau & Denubourg, 1998; Karsai,1999).

Specific testing of our prediction that complexsocieties construct their own nests and that thesenests are more complex may be difficult. First, thethree-dimensional aspect of an excavated nest willprobably be difficult to preserve. Second, to under-stand how features of the nest impact on colonyactivity requires observation of undisturbed coloniesand the attempt to observe will probably affect itsbehaviour. Third, an insect society may move intoan existing crevice but modify it specifically for theirneeds. Consider a honey bee colony, a complexsociety, whose swarms move into natural tree cavitiesbut add their own wax combs. In the case of thehoney bee, these modifications may be easy todetect, but in other instances they may not ; anenlarged natural chamber in a log may be in-distinguishable from an unenlarged one.

(4) Defence

We suggest that simple ant societies with few workersreact to intruders, usually by stinging, as and whenthe need arises. Stinging is presumably the majorancestral ant defence mechanism as ants evolvedfrom wasps (Ho$ lldobler & Wilson, 1990) andstinging is the main, but by no means only, defencein wasps (e.g. Hermann, 1981). By contrast, in morecomplex societies, which have many individuals, thecolonies can afford to allocate workers to specificdefensive roles and defence mechanisms are morelikely to be varied. For instance, they may postworkers at entrances to check colony membershipof individuals going into the nest [e.g. Camponotus(Colobopsis) fraxinicola ants ; Wilson, 1974]. Altern-atively, they may employ workers to patrol aroundthe nest looking for intruders. For instance, Pseudo-myrmex ferruginea ants patrol the foliage around theAcacia thorns in which they live (Janzen, 1983).


Polymorphism, a complex-society trait, opens up thepossibility of producing workers morphologicallyspecialized for defence. Defence is one of the threemain proposed roles of majors in ants, the other twobeing seed milling and food storage (Ho$ lldobler &Wilson, 1990). In many polymorphic species majorspossess particularly large and strong mandibles (e.g.Pheidole pallidula majors who decapitate intruders ;Detrain & Pasteels, 1992). Some species, such asCamponotus truncatus and C. ephippium, possess speciallyshaped phragmotic which can plug nest entrances,either individually or working with others to functionas a living door (Hermann, 1981; Fig. 833 ofHo$ lldobler & Wilson, 1990). In this situation, nestcomplexity itself also plays its part in colony defenceby only having nest entrances that are just oneindividual wide. Similar behaviour is also known insome social wasps, such as in Chartergus chartarius(Jeanne, 1991), in which the passages betweencombs can be blocked by a single individual.

(5) Foraging strategies

As well as the existence of group, team andpartitioned tasks in complex societies, another facetof higher-level functionality is a shift from individualto group foraging strategies. Beckers et al. (1989)identified six foraging strategies in ant colonies : (1) individual foraging foraging without cooperationand communication with others ; (2) tandem run-ning a scout guides one recruit to the food sourcewith or without trail laying; (3) group}massrecruitment the scout guides a group of recruits tothe source, usually laying a trail to the nest ; (4)mass recruitment the scout lays a trail whilereturning to the nest which guides recruits to thefood source; (5) trunk trail semi-permanent trailsguide foragers to long-lasting food sources ; and (6)group hunting a group leaves the nest andforages collectively in a swarm along a well-definedtrail system. It is reasonable to assume that the orderof strategies above (16) represents an increase inreliance on chemical communication between work-ers ; assuming that this ordering is correct, then atrend of increasing use of chemical communicationwith increasing colony size is unmistakable (Beckerset al., 1989; Fig. 1B). (See also Ruano, Tinaut &Jose Soler, 2000 for effects of temperature uponforaging strategy.) In addition, these strategies alsoappear to be correlated with a decrease in theautonomy of the individual foragers themselves (cf.Jaffe & Hebling-Beraldo, 1990 and also see SectionIV.6). That is, there is a shift from information

processing by individuals to emergent properties of aset of essentially probabilistically behaving individ-uals mediated through signals, i.e. a set of trailpheromones. For instance, in an individual foragingstrategy the worker must rely on its own information,navigating back to the nest using the sun or otherlandmarks (e.g. the desert ant Cataglyphis bicolor). Intandem running, a successful returning forager canrecruit just one individual and passes on informationof where the food source is by physically leading therecruit to the source (e.g. Leptothorax spp.). However,with more complex strategies trail pheromones canpass the information not just to one other recruit butto many. There is no need for an individual to beable to navigate back to the nest using the sun or aprominent rock but can simply orient ( smell ) theirway along a chemical trail (e.g. Atta spp.). Despitethe apparent simplicity of this task, foragers ex-perience a constant probability per unit distance oflosing the trail. Seemingly counterintuitive, thisapparently errant behaviour has been shown to bevery adaptive at the group-level (Deneubourg,Pasteels & Verhaeghe, 1983; Deneubourg et al.,1987; Fletcher, Blackwell & Cannings, 1995). Oncelost, these workers become scouts who can search fornew sites. However, it appears that the error rate issufficiently tuned so that enough foragers do not losethe trail and thus can exploit the source whilstenough become scouts enabling a constant supply ofnew sources. [Parallel behaviour is known in honeybee foraging in which the directional information inwaggle dances is imprecise (Weidenmu$ ller & Seeley,1999).] It seems that the complexity emerges at thelevel of the trail network (or group) which canadaptively adjust to fluctuating food dispersion ordensity (e.g. Bernstein, 1975). Thus, the foragers area group-level adaptive unit (sensu Seeley, 1995,1997; see also Bonabeau, 1998).

(6) Tempo, reliability and efficiency

There are some interesting relationships betweencolony tempo, individual reliability, foraging strat-egy, metabolic rate and social complexity, inparticular, colony size. Oster & Wilson (1978),without supporting data, suggested that colonytempo increases with colony size. Tempo was definedsimply as activity level and thus is presumed toincorporate both amount of activity, e.g. proportionof time spent inactive and the speed and efficiency oftask completion when active. Oster & Wilson (1978)suggested that workers of cool species, such asmany ponerine, dacetine and basicerotine ants


which we suggest are simple societies are slow anddeliberate and conservative in their use of energy.Hot species, on the other hand, such as army ants,fire ants and members of the genera Conomyrma,Forelius and Iridomyrmex which we suggest arecomplex societies bustle with activity, seethe withrapid motion and waste substantial amounts oftime canceling out anothers actions (Oster &Wilson, 1978: p. 282). On this basis, we canformulate two non-exclusive hypotheses. First, theamount of inactivity decreases with increasingcolony size (hypothesis 1). Second, running speedincreases with colony size (hypothesis 2).

Hypothesis 1 was examined by Schmid-Hempel(1990). He demonstrated a highly significant de-crease in the proportion of time spent inactive andthe proportion of acts of self-grooming (a low-energyactivity), with increase in colony size for a variety ofant species. See Fig. 1C. Running speed was shownto increase with colony size in Aphaenogaster rudis andMyrmica punctiventris by Leonard & Herbers (1986)and in colony fragments of Pheidole dentata byBurkhardt (1998). Thus, there are limited datasupporting the hypothesis that within a species,running speed is correlated with colony size. Wecompared running speed among 26 taxonomicallydiverse ant species (from 11 genera) and found ahighly significant relationship between runningspeed (cm}s) and log

"!(colony size) (F 5.03, r#

0.326, N 26, P! 0.02) thus confirming hypothesis2 (Fig. 1D). In response to Tschinkels (1991) pleafor publication of sociometric data, the raw dataappear in the Appendix. (Two desert-dwellingspecies, Cataglyphis bicolor and Ocymyrmex barbiger,which live in very hot climates and run fast tominimize time spent outside in the intense heat, werenot included in the analysis. Their running speedsare an order of magnitude greater than most of theother species studied.) Thus, the available data although uncorrected for body size, phylogeneticrelationships and ambient temperature appear tosupport Oster & Wilsons (1978) assertions.

Why should tempo and colony size be correlated?There are at least three main hypotheses. The firstconsiders foraging strategy and reliability theory(Barlow & Proschan, 1975; Oster & Wilson, 1978;Herbers, 1981). The second considers thermo-dynamics of far-from-equilibrium systems (Jaffe &Hebling-Beraldo, 1990, 1993). The third considerspheromone concentration.

The first argument is based upon Oster & Wilsons(1978) proposal that the correlation between tempoand colony size can be explained by diet. Their

argument, now backed with more recent research(Herbers, 1981), goes as follows: in Section IV.5 weexplained that there is a correlation between colonysize and the type of foraging strategy, ranging fromindividual foraging in species with small colonies togroup hunting in species with very large colonies.Herbers (1981) used reliability theory to considerselection pressures upon individual competence forfive ant foraging strategies. She demonstrated thatstrategies associated with larger colonies strategiesin which foragers have a high rate of encounter withtheir prey (high tempo) result in high systemreliability, even if individual competence is low. So,species with large colonies, which are catholic intheir selection of prey (i.e. broad diet ; Oster &Wilson, 1978), can afford some inefficiency at theindividual level. Even though retrieval rate perencounter is relatively low because of poor individualcompetence, the net energy influx is high because ofthe sheer numbers of workers involved (Oster &Wilson, 1978). Herbers (1981: p. 185) even arguesthat colony selection will itself result in workers withlow competence (see also Karsai & Wenzel, 1998).Her argument is based on the assumption thatselection maximizes benefit minus cost, where bene-fit is the system reliability and the cost isindividual competence in which it is assumed thatmore reliable individuals are more costly to produce.Species with few foragers on the other hand, tend tohave a more restricted diet and rely on stealth tocatch their prey; Peeters (1997: p. 377) states thatAll primitive ants are armed with a sting andhunt arthropods. The loss of one forager in a smallcolony will have a large effect upon the colonysforaging success and so they are selected to have highindividual competence. In short, species with fewforagers have to be relatively reliable whereasforagers in large colonies tend to be less reliablebecause they are able to exploit such a strategy. Itcould be argued, although little data exist, that lessreliable foragers, such as blind army ants who followchemical trails and overcome prey by sheer weight ofnumbers (e.g. Schneirla, 1971; Franks, 1989) maybe cheaper to produce than highly competentindividuals who stealthily track down and overcomeprey individually. However, incompetent workerscan only pay their way if they are especially active,i.e. are high tempo, so that their rate of encounters,despite their incompetence during an encounter,generates a sufficient supply of prey items.

A second explanation for the correlation betweentempo and colony size can be formulated, withoutreference to diet, by considering (neg)entropy and


far-from-equilibrium thermodynamics. The moreordered and complex a system is, the further it isfrom equilibrium (disorder) and the higher theamount of energy required to maintain the system inthat ordered state (e.g. Nicolis & Prigogine, 1977).In a static equilibrium, such as a wood ant colonysnest the equilibrium state being disordered pineneedles lying on the ground the energy required toorder the system is normally only a one-timeinvestment: the effort to construct the nest. How-ever, with a dynamic equilibrium, such as the far-from-equilibrium order of an ant society (i.e. theworkers themselves) the disordered state beinganarchic activity, each individual operating in-dependently and without cooperating with others a constant stream of energy is required to maintainthat state. Jaffe & Hebling-Beraldo (1990, 1993 andreferences therein) argue that this should be reflectedin a higher metabolic rate and greater energyconsumption per unit mass. In short, they argue thata correlation should exist between social complexityand (1) the colonys overall energy content and (2)the rate of energy use by workers, which may bereflected in tempo. They found that basal metabolicrate increased with colony size and the number ofmorphological castes in eight attine species from fourgenera (Jaffe & Hebling-Beraldo, 1993). However,these results hold for the genus level but not whencomparing castes within a species. That is, complexcolonies as a whole should have a higher metabolicrate with increasing colony size, but there may bevariation in the partitioning of energy within thecolony. For instance, colonies may contain minimswho have a high metabolic rate and large soldierswith a low metabolic rate. Karsai & Wenzel (1998)point out that one way of increasing the tempo ofinteraction, assuming that the queens resources forworker production are limited, is to increase colonysize through reduced body size of workers. Wesuggest that if the almost universal 0.75 scaling rule for metabolic rate i.e. metabolic rate scales withthe $

%power of body mass (see Schmidt-Nielsen,

1984) applies to workers, then metabolic rate perunit mass increases as workers are reduced (onaverage) in size and overall metabolic rate for thecolony as a whole increases. In effect, we suggest thatthe increase in colony metabolic rate is achieved bypartitioning a colony into smaller and highermetabolic-rate units. This is supported by thefindings of Calabi & Porter (1989) who showed thatin the fire ant Solenopsis invicta, large workers haveapproximately a 30% lower respiration rate permilligram of tissue than smaller ants.

A third possibility and fairly speculative hy-pothesis is that higher running speed might be asecondary effect of colony size. The suggestion is thatcolony size correlates with a larger absolute numberof workers on a foraging trail, which in turn maycorrelate with total pheromone concentration. Ifrunning speed is positively correlated with phero-mone concentration as an experimental study ofEciton burchelli trails suggests (Franks et al., 1991) then we might expect higher tempo in largercolonies. However, further study is required to testthis hypothesis at the cross-species level because ifmembers of larger and more complex colonies laypheromone with higher volatility (and thus lower percapita concentration) then this could ameliorate oreven reverse this effect. A relationship betweencolony size and trail pheromone volatility has not yetbeen investigated, but doing so should be straight-forward.

Irrespective of the underlying reasons for thecorrelation between tempo and social complexity thesame conclusions can be drawn. Individuals fromlarger colonies are more active and tend to be lessreliable, yet the colonies can afford to waste energyat the individual level. However, as noted earlier,larger systems are inherently more homeostatic andpredictable than equivalent smaller systems (e.g.more predictable food influx) and species with largercolonies tend to rely on recruitment during foraging.Karsai & Wenzel (1998) also argue that parallel-processing systems in insect societies are moreefficient per se in that it allows behavioural special-ization of individuals and greater buffering ofstochastic fluctuations. Schmid-Hempel (1990) notesthat this creates an interesting paradox. Largercolonies may be able to afford a reserve forager force.Some of the workers could remain in the colonysaving energy until demand for their labour isrequired when effective recruitment can be used.[See model and review in Anderson (2001).] How-ever, on energetic grounds workers in larger coloniesare shown empirically to be and from theory may berequired to be, more active than in simple societies,i.e. there is an ergonomic cost. This cost is contraryto complex societies greater potential for efficientcooperative work organization, i.e. an ergonomicbenefit. We suggest that far-from-equilibrium dy-namics that larger colonies waste energy as a costof order may be a causal factor of Micheners(1974) much cited but puzzling decrease in per capitaproductivity with increasing colony size, but this willbe difficult to test.


(7) Intermediate-level parts

Groups, teams and possibly nests are examples ofintermediate-level parts (sensu McShea 2000;McShea & Venit, 2001; C. Anderson & D. W.McShea, in preparation), or adaptive structures orbehaviours in an aggregate which are larger than asingle lower-level individual but a subset of thewhole. Such structures occur at the level of theorganism as well as that of the colony and arepossible only where there is a high degree ofcooperation and coordination among lower-levelindividuals. They are important for several reasons.First, they confer a higher level of functionality,meaning that they allow individuals or colonies toachieve things collectively that a single individualcannot. Second, they introduce new organizationallevels. This has important implications in the waythat communication and feedbacks are effectedamong individuals (see Section V). Third, some ofthe structures are clearly associated with a reductionin autonomy: in many self-assemblages, each in-dividual is effectively reduced to a constructioncomponent equivalent to a building brick or scaf-folding pole (C. Anderson, G. The! raulaz & J. L.Deneubourg, in preparation). For instance, theprimary role of an individual ant within a Solenopsisinvicta raft or Eciton burchelli bridge is to hold on toother ants and so help to form and maintain thestructure.

Intermediate-level parts can arise in at least threeways. One is as a connected group of lower-levelindividuals, a subgroup of the entire aggregate in thesame manner as an army ant bivouac. Also, mostmetazoan tissues and organs would qualify asintermediate-level parts. Or at the colony level, anon-social-insect example might be the clusters ofnon-feeding zooids (or clusters in which density offeeding zooids is lower), i.e. maculae, which in somecyclostome bryozoan species may function as ex-current chimneys for the colony as a whole (Banta,McKinney & Zimmer, 1974).

Alternatively, an intermediate-level part mayconsist of a single lower-level individual which ishypertrophied or elaborated in various ways toattain intermediate size. For example, in eukaryoticcells, the portion of the cell that is homologous withthe original eubacterial host, plus subsequent ad-ditions e.g. microtubular structures, Golgi appar-atus and so on constitutes an elaborated lower-level individual. Notice that this set does not includethe former endosymbionts, such as mitochondriaand chloroplasts, which (historically) are different

individuals. At the colony level, modern chondro-phorines, such as Velella and Porpita, consist of a largepneumatophore, or float, overlying a central gas-trozooid, ringed by smaller medusa-shaped gono-zooids (Hyman, 1940; Mackie, 1959); the gastro-zooid by itself constitutes an intermediate-level part.This type of intermediate-level part appear to beabsent in insect societies. This may be so because ofseveral reasons. First, developmental constraints maylimit the size and form that ants may take (Wheeler,1991). Second, large elaborate individuals may betoo costly. Calabi & Porter (1989) analyzed theenergetic costs longevity, respiration rates andenergy content of worker tissue in various sizes ofworkers in the fire ant Solenopsis invicta. Theyconcluded that although larger workers have a lowerrespiration rate per milligram of tissue their largerbiomass means that one large worker must provideservices equivalent to at least four small workers, tojustify the colonys energy investment (Calabi &Porter, 1989: p. 643). Third, elaborated individualsmay reduce the increased reliability effects of systemredundancy. That is, assuming that these individualsare relatively rare, they become key individuals (see Robson & Traniello, 1999). Failure by a singlekey individual may have a large effect upon theoverall efficiency of the group and makes the systemless robust and reliable.

Finally, many of the various inanimate structuresproduced by higher-level entities would count asintermediate-level parts. At very low hierarchicallevels, such structures might include the sheaths thatenclose filamentous cyanobacterial colonies, or atthe eukaryotic cell level, the nucleus (assuming it isnot a former endosymbiont). In multicellular organ-isms, shells and exoskeletons and in colonies inaddition to nests in social insects the various non-zooidal support structures (Boardman & Cheetham,1973) count as intermediate level.

Bonners (1988) argument that size increasesfavours internal division of labour and thereforedifferentiation, can be extended to predict inter-mediate-level parts. As lower-level individuals ag-gregate to form a whole, selection favours theemergence of functional capabilities in the whole.But as the whole grows, lower-level individualsbecome less competent to perform higher-levelfunctions on account of their small size (Bell &Mooers, 1997).Thus, selection favours collaborationsamong a number of lower-level individuals (perhapswith subdivision of labour), the elaboration of singleones, or the production of larger inanimate structuresin order to magnify their effects. Also, the argument


for the production of heterogeneity among indi-viduals, namely the various advantages that flowfrom the cooperative division of labour, predicts thatlabour will be divided among intermediate-levelparts and therefore they will become differentiatedfrom each other. Notice that the main differencebetween this argument and Bonners original is thatintermediate-level parts unlike lower-level indi-viduals (cells in multicellular individuals, or colonymembers in colonies) have no autonomous exist-ence. A team in an insect colony, or a macula in abryozoan colony, has no autonomous existence; eventhe elaborated archaebacterial host has lost much ofits historical autonomy. However, autonomy is not aprerequisite for collaboration and differentiation.

V. COMMUNICATION AND FUNCTIONAL

INTEGRATION

In the previous section, we suggested three differenttask types (group, team and partitioned tasks) inwhich coordination of multiple individuals is re-quired for successful task completion. More gen-erally, most, if not all, aspects of colony life requirecoordination of individuals for successful colonyoperation. Mechanisms exist to coordinate effort,e.g. recruit additional help when required, with theresult that complex colonies can often exhibit group-level adaptive behaviour (see Seeley, 1997;Bonabeau, 1998). It is these mechanisms mech-anisms that are used functionally to integrateindividuals and thus the colony which are con-sidered in this section.

We consider these mechanisms of functionalintegration under Seeleys (1995: p. 248) broaddefinition of communication: information transfervia cues as well as signals . It is important todistinguish cues from signals because, as we showbelow, cues play a relatively more important part incomplex society communication and integrationthan in simple societies. According to Lloyd (1983)a signal is an act of communication that has beenshaped by natural selection. An example of a signalis alarm pheromone. When signaling occurs, naturalselection acts upon both the signal sender, e.g. bycanalizing the behaviour so that it is more stereo-typed and also upon the receiver, e.g. increasingsensitivity to that behaviour pattern. A cue on theother hand, is a structure or behaviour that conveysinformation, but only incidentally and has not beenshaped by natural selection (at least not directly).For example, intranidal temperature and the num-ber of brood are both cues. Clearly, as information

from cues is incidental and may originate frominanimate objects, such as the nest, natural selectionmay only act upon the information receiver.

(1) Signal range and system connectedness

The first signaling characteristic we consider is signalrange: that is, local versus global. Some types ofsignal are expected to be global regardless of socialcomplexity. In particular, we refer to defencebehaviour and alarm pheromones. If a colony isunder attack it is imperative that this information isconveyed to all members of the colony as quickly aspossible so that individuals can become involved incolony defence, or in some cases escape or hide insidethe nest. [In several ant species, young workers reactto an alarm signal by retreating into the nest whileolder workers a disposable caste (Porter &Jorgensen, 1981) who are more likely to havedegenerate ovaries move out and defend the nest(Ho$ lldobler, 1984a).] Consequently, these signalsare expected to be volatile chemical signals ormechanical signals, such as rapping the nest sub-stratum or jittering (Wilson, 1976), with one-to-many effects, which will reach a large proportion ofthe individuals quickly.

Signals for other colony functions, however, areexpected to differ among simple and complexsocieties. In a small simple society, local signals couldpotentially reach a relatively large proportion of thecolony. Also, direct one-to-one interactions withindividuals, such as antennation, as they movearound the nest may also mean that individuals insimple societies interact with a relatively largeproportion of the colony. However, in a complexsociety, individuals are spatially separated. Forinstance, they may be segregated among differentnest chambers. We suggest that this does not implythat signals should be necessarily be long range asone might initially expect. We have argued (SectionsII and IV) that in complex societies, individualsexhibit division of labour and specialization, workmore cooperatively and also that work tends to beconcentrated in different sections of the nest [whichwould favour foraging for work (Tofts 1993;Franks & Tofts, 1994)]. This means that individualsshould primarily interact with others in the localvicinity, i.e. other individuals who may have theskills and opportunity to help with a task. In short,we should expect the existence of cliques : a groupof individuals who preferentially interact with eachother rather than with others outside the group.Thus, the term clique is used here sensu Moritz &


Southwick (1992: p. 145) rather than Ho$ lldobler &Wilsons (1990: p. 343) more stringent and fromour current knowledge of social insect cognitiveabilities, unrealistic definition: a group of workerswhose members recognize one another as individualsto accomplish some task.

It is not clear whether the average signal range,the physical distance, in a simple society is greaterthan in complex societies. Indeed, simple societiesmay rely more on direct one-to-one interaction.However, we do expect signaling behaviour andinteraction in complex societies to be relatively morelocally concentrated and organized. One way ofquantifying this is to consider the average systemconnectedness (or ASC; Moritz & Southwick, 1992).The basis of this analysis is to consider all the possibleinteractions between pairs of individuals, or dyads.Given n individuals, there are n (n1)} 2 possiblepairings. If one can observe interaction behaviourand quantify the number of unique pairs ofindividuals that actually interact within a colony,then the ASC is the observed number of dyadsdivided by the possible number of dyads. ASC valuesrange from 0 to 1; a value of 1 means that allindividuals are fully connected to each other. Wewould expect a high ASC value when consideringalarm pheromones. A low ASC would be expectedwhen interaction patterns are very localized, i.e.cliques occur and members within a clique interactwith each other but cliques have little interactionwith other cliques. On this basis, we make thefollowing prediction: with the exception of im-portant necessarily global signals such as alarmpheromones, ASC for other tasks is expected todecrease with increasing social complexity. This isnot to say that individuals are not highly connectedin a complex society, but simply that for a given task,a relatively small and exclusive group of individualswill be involved in tackling that task. Morespecifically we predict that patterns of interaction, asreflected by ASC, are more heterogeneous incomplex societies than in simple ones. This pattern ofinteraction does not necessarily reflect the totalnumber of interactions among individuals which is afunction of signaling rate, or the strength ofinteraction, simply who is communicating withwhom. Moritz & Southwick (1992: pp. 149150)provide some estimates of ASC for various tasks inthe honey bee. Thus, as expected, an ASC of 1(100%) was observed for defensive behaviour butthe ASC was very low, only 3.1%, for water-handling behaviour. These methods could be used tocompare the ASC in simple versus complex societies.

(2) Signals versus cues

We predict that communication in simple societiesmakes relatively little use of cues whereas they are ofmuch greater importance in more complex societies.Unfortunately, it is not possible to make a moreexplicit prediction, e.g. concerning the ratio ofcommunication that flows through cues rather thansignals since the effects of confounding variables,such as the rate of signaling and existence ofmodulatory signals (see below), are not clear. Wemake the above prediction on the basis that weexpect a vastly different pattern of social regulationin simple versus complex societies. Social regulationin complex societies is based upon feedback fromgroups to individuals and also from groups to groups[see Wilson & Ho$ lldobler (1988) and Seeley (1995)],while regulation in simple societies occurs mostlythrough direct worker-worker interaction and feed-back is primarily from individual to individual.

Wilson & Ho$ lldobler (1988) talk of denseheterarchies and mass communication as the basisof organization in ant colonies. (We suggest that thisapplies only to relatively complex ant societies.)Their findings are mirrored by Seeleys (1995)analysis of honey bee social integration. A heterarchyis a communication network in which interactionoccurs between different levels of the system orsubsystems. Thus, feedback occurs from higher-levelparts (sensu McShea, 2000; McShea & Venit,2001), such as intermediate-level parts, includinggroups and teams, to lower-level parts, such asindividual workers. In turn, lower-level parts haveeffects upon the higher-level parts. For instance, lowfood reserves in the nest could influence a worker toforage and collect more food. The lack of stored foodmay influence many individuals to forage and as theforage reserve within the nest increases this mayinhibit further foraging. Thus, there is interactionfrom the group-level summed foraging effort, i.e. thefood pile (an intermediate-level part) and the lower-level individual foragers. Notice that this exampleused a cue the amount of stored food to influenceindividual activity. Seeley (1995) argues that cuesare relatively important in complex social systems,for at least three reasons. First, the group-level cuesare effectively a summation of many individualsinputs and thus are expected to be accurateindicators. Second, cue-based communication meansthat information may be conveyed to many in-dividuals at once. Third, cues are probably moreeasily sensed than the underlying variables of supplyand demand. For instance, it should be easier to


assess the size of a seed pile than to directly assess theflow of returning foragers adding to the pile andother individuals depleting the resource (seeAnderson & Bartholdi, 2000).

The honey bee society is probably our most well-understood complex insect society. Seeley (1998)reviewed the state of our current knowledge of cuesand signals in honey bee colonies and found thatthere are twice as many known cues as signals (34 vs.17). Unfortunately, the ant literature lacks thisdegree of detail for any ant species and is certainly anarea where more research is needed.

(3) Modulatory signals

Ho$ lldobler (1999) reviewed the complexity of signals(in general) in ant communication. One importantaspect of that review which relates to social com-plexity is the existence of modulatory signals.Ho$ lldobler (1995: p. 20) explicitly states that com-munication in complex social systems is not alwayscharacterized by a deterministic releasing processbut sometimes plays a more subtle role . He proceedsto explain the concept of a modulatory signal asignal that does not necessarily evoke a behaviouralresponse in itself from the recipient, but does alterthe way that the recipient responds to another signal(Markl, 1985). For instance, in Camponotus spp. ants,workers may strike the nest substratum during alarmbehaviour. This signal does not necessarily induceaggressive defensive behaviour per se but appears toact as a modulatory signal. Ants that have firstreceived this drumming signal react more aggres-sively to

individual versus social complexity, with particular reference to ant

Documents