the social brain hypothesis - evergreen state...

ARTICLES

The Social Brain HypothesisRobin I.M. Dunbar

The concensus view has tradition-ally been that brains evolved to pro-cess information of ecological rel-evance. This view, however, ignores animportant consideration: Brains areexceedingly expensive both to evolveand to maintain. The adult humanbrain weighs about 2% of body weightbut consumes about 20% of total en-ergy intake.2 In the light of this, it isdifficult to justify the claim that pri-mates, and especially humans, needlarger brains than other species merelyto do the same ecological job. Claimsthat primate ecological strategies in-volve more complex problem-solv-ing3,4 are plausible when applied tothe behaviors of particular species,such as termite-extraction by chimpan-zees and nut-cracking by Cebus mon-keys, but fail to explain why allprimates, including those that are con-ventional folivores, require larger brainsthan those of all other mammals.

An alternative hypothesis offeredduring the late 1980s was that pri-mates’ large brains reflect the compu-

tational demands of the complex so-cial systems that characterize theorder.5,6 Prima facie, this suggestionseems plausible: There is ample evi-dence that primate social systems aremore complex than those of otherspecies. These systems can be shownto involve processes such as tacticaldeception5 and coalition-formation,7,8

which are rare or occur only in sim-pler forms in other taxonomic groups.Because of this, the suggestion wasrapidly dubbed the Machiavellian intel-ligence hypothesis, although there is agrowing preference to call it the socialbrain hypothesis.9,10

Plausible as it seems, the social brainhypothesis faced a problem that wasrecognized at an early date. Specifi-cally, what quantitative empirical evi-dence there was tended to favor one orthe other of the ecological hypoth-eses,1 whereas the evidence adducedin favor of the social brain hypothesiswas, at best, anecdotal.6 In this article,I shall first show how we can testbetween the competing hypothesesmore conclusively and then considersome of the implications of the socialbrain hypothesis for humans. Finally,I shall briefly consider some of theunderlying cognitive mechanisms thatmight be involved.

TESTING BETWEEN ALTERNATIVEHYPOTHESES

To test between the competing hy-potheses, we need to force the hypoth-eses into conflict in such a way thattheir predictions are mutually contra-dictory. This allows the data to dis-

criminate unequivocally betweenthem. In the present case, we can dothis by asking which hypothesis bestpredicts the differences in brain sizeacross the primate order. To do so, weneed to identify the specific quantita-tive predictions made by each hypoth-esis and to determine an appropriatemeasure of brain size.

The four classes of hypotheses thathave been put forward to explain pri-mate brain evolution are epiphenom-enal, developmental, ecological, andsocial in orientation (Table 1). Theepiphenomenal and developmental hy-potheses share the assumption thatevolution of the brain (or brain part) isnot a consequence of external selec-tion pressures but rather simply aconsequence of something to do withthe way biological growth processesare organized. The epiphenomenal hy-potheses thus argue that brain evolu-tion is a mere byproduct of body sizeevolution, and that brain part size is,in turn, a byproduct of total brainevolution.11

The developmental versions differonly in that they provide a more spe-cific mechanism by presuming thatmaternal metabolic input is the criti-cal factor influencing brain develop-ment. This claim is given credibility bythe fact that the bulk of brain growthin mammals occurs prenatally. In-deed, it appears to be the completionof brain development that precipitatesbirth in mammals, with what littlepostnatal brain growth occurs beingcompleted by the time an infant isweaned. From this, the conclusion isdrawn that brain evolution must beconstrained by the spare energy, overand above her basal metabolic require-ments, that the mother has to channelinto fetal development.12–14 Some evi-dence in support of this claim comesfrom the fact that frugivorous pri-mates have larger adult brains relative

Robin Dunbar is Professor of EvolutionaryPsychology and Behavioural Ecology atthe University of Liverpool, England. Hisresearch primarily focuses on the behav-ioral ecology of ungulates and human andnonhuman primates, and on the cognitivemechanisms and brain components thatunderpin the decisions that animals make.He runs a large research group, with gradu-ate students working on many differentspecies on four continents.

Key Words: brain size, neocortex, social brainhypothesis, social skills, mind reading, primates

Conventional wisdom over the past 160 years in the cognitive and neuroscienceshas assumed that brains evolved to process factual information about the world.Most attention has therefore been focused on such features as pattern recognition,color vision, and speech perception. By extension, it was assumed that brainsevolved to deal with essentially ecological problem-solving tasks.1

178 Evolutionary Anthropology

to body size than do folivorous pri-mates.1 This has been interpreted asimplying that frugivores have a richerdiet than folivores do and thus havemore spare energy to divert into fetalgrowth. Large brains are thus seen asa kind of emergent epigenetic effect ofspare capacity in the system.

Both kinds of explanations sufferfrom the problem that they ignore afundamental principle of evolutionarytheory, which is that evolution is theoutcome of the balance between costsand benefits. Because the cost of main-taining a large brain is so great, it isintrinsically unlikely that large brainswill evolve merely because they can.Large brains will evolve only when theselection factor in their favor is suffi-cient to overcome the steep cost gradi-ent. Developmental constraints are un-doubtedly important, but rather thanbeing causal their role is that of aconstraint that must be overcome iflarger brains are to evolve. In addition,Pagel and Harvey15 have shown that

the energetic arguments do not addup: Precocial mammals do not havehigher metabolic rates than do altri-cial mammals despite the fact thatthey have neonatal brain sizes thatare, on average, twice as large. Wetherefore do not need to consider ei-ther epiphenomenal or developmentalhypotheses any further in the contextof this article.

This does not necessarily mean thatthese explanations are wrong. Bothkinds of explanation may be true inthe sense that they correctly identifydevelopmental constraints on braingrowth, but they do not tell us whybrains actually evolved as they did.They may tell us that if you want toevolve a large brain, then you mustevolve a large body in order to carrythe energetic costs of doing so or a diet

that ensures sufficient energy to pro-vide for fetal brain development. Nosuch allometric argument can everimply that you have to evolve a largebrain or a large body. The large brainor brain part is a cost that animalsmust factor into their calculationswhen considering whether or not alarge body or a large brain is a sensiblesolution to a particular ecologicalproblem. Shifts to more energy-rich ormore easily processed diets may beessential precursors of significant in-creases in brain or brain part size.2

This would explain why frugivoreshave larger brains than folivores doand why hominids have larger brainsthan great apes do.

This leaves us with just two classesof hypotheses, the ecological and thesocial. At least three versions of theformer can be identified, which I willterm the dietary, mental maps, andextractive foraging hypotheses. In es-sence, these argue, respectively, thatprimate species will need larger brainsif (i) they are frugivorous because fruitsare more ephemeral and patchy intheir distribution than leaves are, andhence require more memory to find;(ii) they have larger ranges because ofthe greater memory requirements oflarge-scale mental maps; or (iii) theirdiet requires them to extract resourcesfrom a matrix in which they are em-bedded (e.g., they must remove fruitpulp from a case, stimulate gum flowfrom a tree, extract termites from atermitarium, or hunt species that arecryptic or behave evasively).

For obvious reasons, I used the per-centage of fruit in the diet as an appro-priate index for the dietary hypothesis.I used the size of the range area andthe length of the day journey as alterna-tive indices for the mental mappinghypothesis, though I present the dataonly for the first of these here. The firstindex corresponds to the case in whichanimals have to be able to manipulateinformation about the locations of re-sources relative to themselves in aEuclidean space (an example wouldbe the nut-cracking activities of theTaı chimpanzees16); the second corre-sponds to the possibility that the con-straint lies in the needs of some aspectof inertial navigation. The extractiveforaging hypothesis is less easy to char-acterize in quantitative terms becausethere is no objective measure of thedegree to which diets vary in theirextractiveness. However, Gibson4 pro-vided a classification of primate spe-cies into four categories of diet thatdiffer in their degree of extractiveness.We can test this hypothesis by askingwhether there is a consistent variationin brain size among these four catego-ries, with the species having the moreextractive diets having larger brainsthan those with the less extractivediets.

Finally, we need an index of socialcomplexity. In my original analyses, Iused social group size as a simplemeasure of social complexity. Althoughat best rather crude, this measurenonetheless captures one aspect of the

TABLE 1. Hypotheses Used to Explain theEvolution of Large Brains in Primates

Hypothesis Sources

A. Epiphenomenalhypotheses

1. Large brains (or brain parts)are an unavoidable conse-quence of having a largebody (or brain) 11, 70

B. Ecological hypotheses2. Frugivory imposes higher

cognitive demands thanfolivory does 1, 65

3. Brain size constrains the sizeof the mental map: 1

(a) constraint on size ofhome range

(b) constraint on inertialnavigation (day journeylength)

4. Extractive foraginghypothesis 3, 4

C. Social hypotheses5. Brain size constrains size of

social network (group size):6, 71,

72(a) constraint on memory

for relationships(b) constraint on social skills

to manage relationshipsD. Developmental

hypotheses6. Maternal energy con-

straints determine energycapacity for fetal braingrowth

12, 1346, 55,73

Because the cost ofmaintaining a large brainis so great, it isintrinsically unlikely thatlarge brains will evolvemerely because theycan. Large brains willevolve only when theselection factor in theirfavor is sufficient toovercome the steep costgradient.

ARTICLES Evolutionary Anthropology 179

complexity of social groups, the factthat information-processing demandscan be expected to increase as thenumber of relationships involved in-creases. More importantly, perhaps,this measure has the distinct merit ofbeing easily quantified and widelyavailable. Although it is possible toconceive of a number of better mea-sures of social complexity, the appro-priate data are rarely available formore than one or two species.

The second problem concerns themost appropriate measure of brainevolution. Hitherto, most studies haveconsidered the brain as a single func-tional unit. This view has been rein-forced by Finlay and Darlington,11 whoargued that the evolution in brain partsize closely correlates with the evolu-tion of total brain size and can beexplained simply in terms of allome-tric consequences of increases in totalbrain size. However, Finlay and Dar-lington failed to consider the possibil-ity that changes in brain size mightactually be driven by changes in itsparts rather than in the whole brain.This is especially true of the neocor-tex, for its volume accounts for 50% to80% of total brain volume in primates.Thus, changes in the volume of theneocortex inevitably have a large di-rect effect on apparent change in brainvolume that may be quite unrelated tochanges in other brain components.This point is given weight by the factthat Finlay and Darlington themselves

showed that neocortex size is an expo-nential function of brain size, whereasother brain components are not.

Finlay and Darlington11 notwith-standing, there is evidence that brainevolution has not been a history ofsimple expansion in total volume.Rather, brain evolution has been mo-saic in character, with both the rateand the extent of evolution havingvaried between components of the sys-tem. MacLean17 pointed out manyyears ago that primate brain evolutioncan be viewed in terms of three majorsystems (his concept of the triunebrain). These systems correspond tothe basic reptilian brain (hind- andmidbrain systems), the mammalianbrain (palaeocortex, subcortical sys-tems), and the primate brain (broadly,the neocortex). A more importantpoint, perhaps, is that variations canbe found within these broad catego-ries in the rates at which differentcomponents expanded, which, in atleast some cases, have been shown tocorrelate with ecological factors.10 Par-tialling out the effects of body size onthe size of brain components suggeststhat the story may be more complexthan Finlay and Darlington11 sup-posed, with some remodeling of braingrowth patterns occurring in the tran-sitions between insectivores, prosim-ians, and anthropoids.18

The important point in the presentcontext is that, as Passingham19 noted,relative to the more primitive parts of

the brain such as the medulla, theneocortex shows dramatic and increas-ing expansion across the range of pri-mates (Fig. 1). The neocortex is ap-proximately the same size as themedulla in insectivores; however, it isabout 10 times larger than the me-dulla in prosimians and 20–50 timeslarger in the anthropoids, with thehuman neocortex being as much as105 times the size of the medulla.

This suggests that rather than look-ing at total brain size, as previousstudies have done, we should in fact beconsidering the brain system, namelythe neocortex, that has been mainlyresponsible for the expansion of theprimate brain. From the point of viewof all the hypotheses of primate brainevolution, this makes sense: The neo-

cortex is generally regarded as beingthe seat of those cognitive processesthat we associate with reasoning andconsciousness, and therefore may beexpected to be under the most intenseselection from the need to increase orimprove the effectiveness of these pro-cesses.

One additional problem needs to beresolved. In his seminal study of brainevolution, Jerison20 argued that brainsize can be expected to vary with bodysize for no other reason than funda-mental allometric relationships associ-ated with the need to manage thephysiological machinery of the body.What is of interest, he suggested, isnot absolute brain size, but the sparebrain capacity over and above thatneeded to manage body mechanisms.

Figure 1. Neocortex volume as a ratio of medulla volume in different groups of primates (afterPassingham19). Source: Stephan et al29

. . . brain evolution hasnot been a history ofsimple expansion in totalvolume. Rather, brainevolution has beenmosaic in character,with both the rate andthe extent of evolutionhaving varied betweencomponents of thesystem.

180 Evolutionary Anthropology ARTICLES

For this reason, Jerison derived hisencephalization quotient. All subse-quent studies have used body size asthe appropriate baseline against whichto measure relative deviations in brainsize. However, a problem has sinceemerged: Brain size is determinedearly in development and, comparedto many other body systems, appearsto be highly conservative in evolution-ary terms. As a result, body size canoften change dramatically both ontoge-netically across populations in re-sponse to local environmental condi-tions21 and phylogenetically22,23

without corresponding changes inbrain size. This is particularly con-spicuous in the case of phyletic dwarfs(e.g., callitrichids and perhaps mod-ern humans and hylobatids22) and spe-cies in which body size may haveincreased in response to predationpressure following the occupation ofmore open terrestrial habitats (e.g.,papionids24).

The lability of body size thereforemakes it a poor baseline, though onethat probably is adequate for analyseson the mouse-elephant scale. Conse-quently, it is necessary to find an inter-nally more consistent baseline for taxo-nomically fine-grained analyses.Willner22 suggested that either molartooth size or brain size may be suit-able because both are developmen-tally conservative. Because we are con-cerned with brain part size, someaspect of brain size seems the mostappropriate.

At this point, three options are avail-able. One is to compare the neocortex,the brain part of interest, with thewhole brain; the second is to use therest of the brain other than the part ofinterest; the third is to use some lessvariable primitive component of thebrain, such as the medulla, as a base-line. Two options are in turn availableas mechanisms for controlling forbrain size in each of these cases. Oneis to use residuals from a commonregression line against the baseline(e.g., the residual of neocortex volumeon total brain volume or medula vol-ume). The other choice is to use ratios.

We have considered and tested allthese options10,24 (see Box 1). The re-sults are virtually identical irrespec-tive of which measure is used. Oneexplanation for this may be that allthese measures actually index the same

thing, absolute neocortex size, mainlybecause the neocortex is such a largecomponent of the primate brain. In-deed, the use of absolute neocortexsize produces results that are similarto those obtained from relativized indi-ces of neocortex volume.24,25 Thismakes some sense in computationalterms: As Byrne26 has pointed out, a10% increase in the processing capac-ity of a small computer is worth agreat deal less in information-process-ing terms than is a 10% increase in alarge computer. Although residualsfrom a common regression line wouldconventionally be considered the saf-est measure, and have been used inmany recent analyses,27,28 I shall con-

tinue to use my original ratio indexbecause it provides the best predictor(see Box 1).

Finally, it is now widely appreciatedthat comparative analyses need to con-trol for the effects of phylogeneticinertia. Closely related species can beexpected to have similar values formany anatomical and behavioral di-mensions merely by virtue of havinginherited them from a recent commonancestor. In such cases, plotting rawdata would result in pseudoreplica-tion, artificially inflating the samplesize by assuming that closely relatedspecies are actually independent evolu-tionary events. The ways of dealing

with this problem include plottingmeans for higher taxonomic units, per-forming nested analyses of varianceusing phylogenetic levels as factors,comparing matched pairs of species,and making independent contrasts thatcontrol directly for phylogeny. Eachmethod has its own advantages anddisadvantages, but the first and thirdprocedures are particularly associatedwith loss of information and smallsample sizes. I shall use the first andlast method, the last because it allowsindividual species to be compared, butthe first because it allows grade shiftswithin data sets to be identified (aproblem that independent-contrastsmethods have difficulty dealing with).I shall take the genus as a suitablebasis for analysis because genera typi-cally represent different reproductiveor ecological radiations and thus aremore likely to constitute independentevolutionary events.

The resulting analyses are relativelystraightforward: Figure 2 presents thedata for neocortex ratio for the anthro-poid primate species in the data baseof Stephan, Frahm, and Baron.29 Neo-cortex size, however measured, doesnot correlate with any index of theecological hypotheses, but does corre-late with social group size. Similarfindings were reported by Sawaguchiand Kudo,30 who found that neocortexsize correlated with mating system inprimates. Barton and Purvis31 haveconfirmed that using both residuals ofneocortex volume on total brain vol-ume and the method of independentcontrasts yields the same result. BothBarton10 and T. Joffe (unpublished)have repeated the analyses using themedulla as the baseline for compari-son. More importantly, Barton andPurvis31 have shown that while rela-tive neocortex volume correlates withgroup size but not the size of theranging area, the reverse is true ofrelative hippocampus size. A correla-tion between range area and hippo-campus size is to be expected becauseof hippocampal involvement in spatialmemory.32,33 This correlation demon-strates that it is not simply total brainsize that is important (a potential prob-lem, given the overwhelming size of

The neocortex isgenerally regarded asbeing the seat of thosecognitive processes thatwe associate withreasoning andconsciousness, andtherefore may beexpected to be underthe most intenseselection from the needto increase or improvethe effectiveness ofthese processes.


the primate neocortex). Moreover, itpoints to the specific involvement ofthe neocortex.

The validity of this relationshipcould be tested directly by using it topredict group sizes in a sample of

species for which brain volumetricdata were not available in the originalsample of Stephan, Frahm, and

TABLE 2. Stepwise Regression Analysisof Indices of Brain Component Volume

as Predictors of Group Size inAnthropoid Primates, Based on

Independent Contrasts Analysis*

Independent Variable t P

Absolute brain volume 21.69 0.107Residual of brain volume

on body mass 20.91 0.371Absolute telencephalon

volume 21.72 0.101Residual of telen-

cephalon volume onbody mass 0.61 0.546

Residual of telen-cephalon volume onbrain volume 1.69 0.107

Absolute neocortexvolume 21.70 0.104

Residual of neocortexvolume on body mass 0.56 0.583

Residual of neocortexvolume on brainvolume 1.60 0.136

Neocortex ratio (againstrest of brain) 3.79 0.001

*Sample: 24 species of anthropoid pri-mates from Stephan, Frahm, andBaron.29

Box 1. How to Measure Brains

R. Dunbar and Tracey H. JoffeThe different ways of measuring

relative brain size have raised doubtsas to the most appropriate techniqueto use.65 Many researchers have pre-ferred to use residuals from the com-mon regression line of best fit for thedata set concerned. This provides ameasure of the extent to which brain(or brain part) volume deviates fromwhat would be expected for an aver-age member of the relevant taxon ofthe appropriate size. Although ratioshave been used to compare the rela-tive size of brain components,29,66 thishas been criticized on the groundsthat trade-offs within the brain maymean that a given index simply mea-sures total brain size (or the size of abrain part) and thus does not removethe effects of absolute size. Ratiosmay also be prone to autocorrelationeffects, especially when the baselineis taken to be the whole brain and, asin the case of the neocortex, the part inquestion is a major volumetric compo-nent of the brain.

Although there are likely to besome trade-offs of this kind within thebrain, the fact that neocortex volumeincreases progressively across the pri-mate order suggests that such con-straints are less likely to have a signifi-cant effect on a ratio measure. Ofcourse the residuals procedure is itselfa ratio: Encephalization-type indicesare calculated as actual volume di-vided by predicted volume (which,when data are logged, becomes theconventional actual minus predictedvalues). Thus, ratio measures per semay not be the problem. Rather, thesubstantive objection is whether or nota ratio partials out the allometric ef-fects of body size. In fact, it seems thatneocortex ratios are not correlatedwith the basal brain (i.e., brain volumeexcluding the neocortex) within majortaxonomic groups (unpublished analy-ses). Consequently, this criticism hasless force than it might appear to have

on first sight. Moreover, any index thatuses the whole brain as its base islikely to suffer from autocorrelationeffects. Because the neocortex is sucha large proportion of the brain in pri-mates, residuals of neocortex fromtotal brain size may simply be a mea-sure of neocortex plotted against it-self.

To consider the problem in moredetail, we ran a stepwise regressionanalysis on the 24 species of anthro-poid primates, including humans, onthe data base of Stephan, Frahm, andBaron,29 with group size as the depen-dent variable and nine indices of rela-tive brain or brain-part volume as inde-pendent variables. In addition toneocortex ratio, these included totalbrain volume as well as telencephalonand neocortex volume, each taken asabsolute volume and as a residualfrom both body mass and brain vol-ume. All variables were log10-trans-formed for analysis. In both cases,neocortex ratio was selected as thevariable of first choice. We carried outboth regressions on generic plots andindependent contrast analyses. Forthe contrasts analysis, the best fitleast-squares regression equationthrough the origin was:

Contrast in log10 (group size) 5 3.834

* Contrast in log10 (neocortex ratio)

(r 2 5 0.395, F1,22 5 15.39,

P 5 0.001).

With all other variables held constant,none of the other eight indices made asignificant contribution to the variabil-ity in group size in either analysis.Table 2 gives the results for the inde-pendent contrasts analysis.

Neocortex ratio is thus the singlemost powerful predictor of group sizein these species. While the biologicalsignificance of this variable remainsopen to interpretation, the fact that itprovides the best predictor, indepen-

dently of all other confounding mea-sures, suggests that more detailedconsideration needs to be given to itssignificance and meaning. It may be,for example, that body size, ratherthan being a determinant20 is simply aconstraint on neocortex size: A spe-cies can evolve a large neocortex onlyif its body is large enough to providethe spare energy capacity throughKleiber’s relationship for basal meta-bolic rate to allow for a larger thanaverage brain. This interpretation isimplied by the Aiello and Wheeler2

‘‘expensive tissue hypothesis.’’ It wouldalso be in line with Finlay and Darling-ton’s11 claim that in mammals the evo-lution of brain-part size is driven, devel-opmentally at least, by the evolution ofthe whole brain, thus generating verytight correlations between brain-partsize and total brain size.


Baron.29 I did this by exploiting thefact that neocortex ratios can be pre-dicted from total brain volume,34 aresult that, in fact, follows directlyfrom the Finlay and Darlington11 find-ings. The result was a significant fitbetween predicted neocortex ratio andobserved mean group size for a sam-ple of 15 New and Old World monkeyspecies.35

Barton27 noted that the originalanalyses of Dunbar24 seemed to implythat variation in neocortex size wasmuch greater than variation in groupsize in the prosimians. Using Dun-bar’s24 data on group size, Barton sug-gested that the relationship betweenneocortex and group size did not ap-ply in the case of prosimians. How-ever, the data on prosimian groupsizes in this sample suffered from apaucity of data, particularly for thenocturnal species. Because many ofthese are described as semi-solitary, itwas conservatively assumed in theDunbar24 database that their groupsize was one. More recent field studieshave produced markedly improved es-

timates of the sizes of social groupsand, in the case of the semi-solitaryspecies, daytime nest groups.36 Re-analysis of the data for prosimiansusing these improved estimates of so-cial group size suggests that thesespecies do in fact adhere to the samerelationship between neocortex andgroup size as that which pertains forother primates.25 More importantly,the regression line for this taxon isparallel to, but shifted to the left of,that for other anthropoid primates(Fig. 3).

This relationship has now beenshown to hold for at least four othermammalian orders: bats,10 carnivoresand insectivores,37,38 and odontocetecetaceans.39,40 In the case of the insec-tivores, the data points are shifted farto the left of those for the primates, asmight be expected of a taxonomicgroup that is considered to be broadlyrepresentative of the ancestral mam-mals.37 However, the relationship isweak in this case, probably becauseestimates of group size are particu-larly uncertain for insectivores.

Surprisingly, the data for the carni-vores map directly onto those for thesimian primates, that is, the regres-sion lines for the two data sets do notdiffer significantly. However, the carni-vores do not exhibit as wide a range ofneocortex ratios or group sizes as doanthropoid primates. The fact that theprosimians lie to the left of both thesetaxonomic groups implies that the car-nivores represent an independent evo-lutionary development along the sameprinciples as the anthropoid primates,the difference being that they just havenot taken it as far as primates have.One reason for this may be that thecarnivore social world is olfaction-dominated rather than vision-domi-nated, as in the case of the primates.Barton10,27,41 has pointed out that theshift to a diurnal lifestyle based oncolor vision, perhaps initially diet-driven, but leading to a shift into vi-sion-based communication, may bethe key feature that has spurred on thedramatic development of the primateneocortex.

Figure 2. Relative neocortex size in anthropoid primates plotted against (a) percentage of fruit in the diet, (b) mean home-range size scaled asthe residual of range size regressed on body weight (after Dunbar24), (c) types of extractive foraging (after Gibson4), and (d) mean group size.((a), (b), and (d) are redrawn from Dunbar24, Figures 6, 2 and 1, respectively; (c) is from Dunbar,35 Figure 2.)


REFINING THE RELATIONSHIPThe social brain hypothesis implies

that constraints on group size arisefrom the information-processing ca-pacity of the primate brain, and thatthe neocortex plays a major role inthis. However, even this proposal isopen to several interpretations as tohow the relationship is mediated. Atleast five possibilities can be usefullyconsidered. The constraint on groupsize could be a result of the ability torecognize and interpret visual signalsfor identifying either individuals ortheir behavior; limitations on memoryfor faces; the ability to remember whohas a relationship with whom (e.g., alldyadic relationships within the groupas a whole); the ability to manipulateinformation about a set of relation-ships; and the capacity to process emo-tional information, particularly withrespect to recognizing and acting oncues to other animals’ emotional states.These are not all necessarily mutuallyexclusive, but they do identify differ-ent points in the cognitive mechanismthat might be the crucial information-processing bottleneck.

Although visual mechanisms arelikely to be important for social inter-action, and may well have been the

initial kick for the evolution of largebrains in primates,10 it seems intrinsi-cally unlikely that the ultimate con-straint lies in the mechanisms of thevisual system itself.28 Although thereis a correlation between the relativesize of the visual cortex and group sizein anthropoid primates, the fit is muchpoorer, and the slope significantly shal-lower than that between the nonvisualneocortex and group size (r2 5 0.31 vsr2 5 0.61, respectively) (Fig. 4). Partialcorrelation analysis indicates that onlythe correlation for the nonvisual rela-tionship remains significant when theother component is held constant28

(though this is not true for prosim-ians25). A more important point is thatthe volume of the lateral geniculatenucleus, a major subcortical way sta-tion in visual processing, does notcorrelate with group size at all, indicat-ing that pattern recognition per se isunlikely to be the issue.28 It may be ofsome significance that the absolutesize of the visual cortex seems to reachan asymptotic value in the great apeclade, whereas the nonvisual neocor-tex continues to increase in size. Oneinterpretation of this is that visualprocessing does not necessarily con-tinue to improve indefinitely as the

size of the cortical processing machin-ery increases, at least relative to theopportunity cost of taking cortical neu-rons away from other cognitive pro-cesses.

It seems equally unlikely that theproblem lies with a pure memory con-straint, though memory capacity obvi-ously must impose some kind of upperlimit on the number of relationshipsthat an animal can have. There arethree reasons for this claim. First, inhumans at least, memory for faces isan order of magnitude larger than thepredicted cognitive group size: Hu-mans are said to be able to attachnames to around 2,000 faces but havea cognitive group size of only about

150. Second, there is no intrinsic rea-son to suppose that memory per se isthe issue. The social brain hypothesisis about the ability to manipulate infor-mation, not simply to remember it.Third, and perhaps most significantly,memories appear to be stored mainlyin the temporal lobes,42 whereas re-cent PET scan studies implicate theprefrontal neocortex, notably Brod-man area 8, as the area for social skillsand, specifically, theory of mind.43

Frith44 has suggested that memoriesand representations for objects orevents may involve interactions be-tween several levels of the neocortexdepending on the kinds of operations

Figure 3. Mean group size plotted against neocortex ratio for individual genera, shownseparately for prosimian, simian, and hominoid primates. Prosimian group size data, fromDunbar and Joffe,25 include species for which neocortex ratio is estimated from total brainvolume. Anthropoid data are from Dunbar.24 Simians: 1, Miopithecus; 2, Papio; 3, Macaca; 4,Procolobus; 5, Saimiri; 6, Erythrocebus; 7, Cercopithecus; 8, Lagothrix; 9, Cebus; 10, Ateles; 11,Cercocebus; 12, Nasalis; 13, Callicebus; 14, Alouatta; 15, Callimico; 16, Cebuella; 17, Saguinus;18, Aotus; 19, Pithecia; 20, Callicebus. Prosimians: a, Lemur; b, Varecia; c, Eulemur; d, Propithe-cus; e, Indri; f, Microcebus; g, Galago; h, Hapalemur; i, Avahi; j, Perodictus.

The social brainhypothesis implies thatconstraints on group sizearise from theinformation-processingcapacity of the primatebrain, and that theneocortex plays a majorrole in this. However,even this proposal isopen to severalinterpretations as to howthe relationship ismediated.


involved. These interactions could oc-cur between the sensory and associa-tion cortices (perceiving an object),between the association and frontalcortices (remembering an object), andamong all three (being aware of per-ceiving an object). It is worth noting inthis context that although social skillsare commonly disrupted by damage tothe prefrontal cortex, memory forevents and people is not.42

It seems unlikely that emotional re-sponses per se are the substantiveconstraint. Although the correct emis-sion and interpretation of emotionalcues is of singular importance in themanagement of social relationships,45

there is little evidence that the subcor-tical areas principally associated withemotional cuing (for example, theamygdala in the limbic system) corre-late in any way with social groupsize.28 Indeed, Keverne, Martel, andNevison46 point out that there hasbeen progressive reduction in the rela-tive sizes of the ‘‘emotional’’ centers inthe brain (the hypothalamus and sep-tum) in favor of the ‘‘executive’’ cen-ters (the neocortex and striate cortex)during primate evolution. They inter-

pret this in terms of a shift away fromemotional control of behavior to moreconscious, deliberate control.

The only remaining alternative isthat the mechanisms involved lie inthe ability to manipulate informationabout social relationships themselves.This claim is supported by six addi-tional lines of evidence that point tothe fundamental importance of socialskills in the detailed management ofsocial relationships.

One is the fact that close analysis ofthe data on group size and neocortexvolume suggests that there are, in fact,distinct grades even within the anthro-poid primates (Fig. 3). Apes seem to lieon a separate grade from the monkeys,which in turn lie on a separate gradefrom the prosimians. The slope coeffi-cients on these separate regressionlines do not differ significantly, but theintercepts do. It is as if apes requiremore computing power to manage thesame number of relationships thatmonkeys do, and monkeys in turnrequire more than prosimians do. Thisgradation corresponds closely to theperceived scaling of social complexity.

The second line of evidence is that

the rates with which tactical deceptionare used correlate with neocortexsize.26 Species with large neocortexratios make significantly more use oftactical deception, even when the dif-ferential frequencies with which theselarge-brained species have been stud-ied are taken into account.

Third, Pawlowski, Dunbar, andLowen47 have shown that among po-lygamous primates the male rank cor-relation with mating success is nega-tively related to neocortex size (Fig. 5).This is just what we would predict ifthe lower ranking males of specieswith larger neocortices were able touse their greater computational capaci-ties to deploy more sophisticated so-cial skills, such as the use of coalitionsand capitalizing on female matechoice, to undermine or circumventthe power-based strategies of the domi-nant animals.

The fourth line of evidence is Jof-fe’s48 demonstration that adult neocor-tex size in primates correlates with thelength of the juvenile period, but notwith the length of gestation, lactation,or the reproductive life span, eventhough total brain size in mammalscorrelates with the length of the gesta-tion period.49,50 This suggests that whatis most important in the developmentof a large neocortex in primates is notthe embryological development ofbrain tissue per se, which is associatedmainly with gestation length, butrather the ‘‘software programming’’that occurs during the period of sociallearning between weaning and adult-hood.

Fifth, Kudo, Lowen, and Dunbar51

have shown that grooming clique size,a surrogate variable that indexes alli-

Figure 4. Independent contrasts in mean group size plotted against contrasts in the visualcortex and the volume of the rest of the neocortex (nonvisual neocortex) for individualanthropoid species. Note that the visual cortex is here defined as visual area V1; the nonvisualcortex is the non-V1 volume of the neocortex and thus includes some higher order visualprocessing components (e.g., visual area V2). Unfortunately, the data base of Stephan, Frahm,and Baron29 does not allow us to define our measure of the nonvisual area any more finely thanthis. (Reprinted from Joffe and Dunbar,28 Fig. 1.)

. . . there is no intrinsicreason to suppose thatmemory per se is theissue. The social brainhypothesis is about theability to manipulateinformation, not simply toremember it.


ance size, correlates rather tightly withrelative neocortex and social groupsize in primates, including humans(Fig. 6). The human data derive fromtwo samples: hair-care networksamong female bushmen52 and supportcliques among adults in the UnitedKingdom.53 What is remarkable is howclosely the human data fit with thedata from other primate species.Grooming cliques of this kind invari-ably function as coalitions in primategroups. Coalitions are functionally cru-cial to individuals within these groupsbecause they enable the animals tominimize the levels of harassment andcompetition that they inevitably sufferwhen living in close proximity to oth-ers.54 Coalitions essentially allow pri-mates to manage a fine balancing actbetween keeping other individuals offtheir backs while at the same timeavoiding driving them away altogetherand thereby losing the benefits forwhich the groups formed in the firstplace. These results can thus probablybe interpreted as a direct cognitivelimitation on the number of individu-als with which an animal can simulta-neously maintain a relationship of suf-

ficient depth that they can be relied onto provide unstinting mutual supportwhen one of them is under attack.Because this is the core process that

gives primate social groups their inter-nal structure and coherence, this canbe seen as a crucial basis for primatesociality.

Finally, Keverne, Martel, and Nevi-son46 have suggested that the neocor-tex and striate cortex, those areas ofthe primate brain that are responsiblefor executive function, are under ma-ternally rather than paternally im-printed genes (i.e., genes that ‘‘know’’which parent they came from),whereas the converse is true for thelimbic system, those parts of the brainmost closely associated with emo-tional behavior. They interpret this inrelation to the cognitive demands ofthe more intense social life of femalesin matrilineal female-bonded societ-ies.

IMPLICATIONS FOR HUMANGROUPS

The fact that the relationship be-tween neocortex size and what I willterm the cognitive group size holds upso well in so many different taxonomicgroups raises the obvious question ofwhether or not it also applies to hu-mans. We can easily predict a value forgroup size in humans. Doing so, whichis simply a matter of using the humanneocortex volume to extrapolate avalue for group size from the primate

Figure 5. Independent contrasts in the Spearman rank correlation (rs) between male rank andmating success plotted against contrasts in neocortex size for two different male cohort sizes (4to 8, and 9 to 30 males) for individual species. The regression equations for the two cohort sizesare significantly different from b 5 0. The species sampled are C. apella, P. entellus, C. aethiops,M. fuscata, M. mulatta, M. radiata, M. arctoides, P. cynocephalus, P. anubis, P. ursinus, and P.troglodytes. (Redrawn from Pawlowski, Dunbar, and Lowen,47 Fig. 1.)

Figure 6. Mean grooming clique size plotted against mean neocortex ratio for individualprimate genera. The square is Homo sapiens. Species sampled are L. catta, L. fulvus, Propithe-cus, Indri, S. sciureus, C. apella, C. torquatus, A. geoffroyi, A. fusciceps, P. badius, P. entellus, P.pileata, P. johnii, C. campbelli, C. diana, C. aethiops, C. mitis, E. patas, M. mulatta, M. fuscata,M. arctoides, M. sylvana, M. radiata, P. anubis, P. ursinus, P. cynocephalus, P. hamadryas, T.gelada, P. troglodytes, P. paniscus. (Redrawn from Kudo, Lowen, and Dunbar,51 Fig. 4a.)


equation, produces a value in the or-der of 150. The real issue is whetherhumans really do go around in groupsof this size.

Identifying the relevant level ofgrouping to measure in humans isdifficult because most humans live ina series of hierarchically inclusivegroups. This, in itself, is not especiallyunusual: Hierarchically structuredgroups of this kind are characteristicof primates54 and may be typical ofmany mammals and birds.55 At leastin the case of the diurnal primates, itseems that, with a few notable excep-tions, the various species’ groupingpatterns exhibit an overt level of stabil-ity at roughly the same position in thehierarchy across a wide range of taxa.Moreover, because the various layersof this hierarchy appear to be inti-mately related to each other, probablythrough being part of a series of cause-and-consequence chains,51 it wouldnot matter which particular groupinglevel (for example, stable social group,

network, or grooming clique) wastaken to be the grouping criterion.

The problem with respect to hu-mans is that it is difficult to identifywhich of the many potential groupinglevels is functionally or cognitivelyequivalent to the particular level ofgrouping that I happened to use forprimates. This difficulty is particularlyintrusive in this case because humanslive in a dispersed social system some-times referred to as a fission-fusionsystem. In order to get around thisproblem, I adopted the converse strat-egy in my original analysis,56 askingwhether there was any group size con-sistently characteristic of humans thatwas of about the requisite size and, ifso, whether its intrinsic psychologicalcharacteristics were similar to thosefound in primate groups.

Because of the structural complex-ity of postagricultural societies, I con-sidered only traditional hunter-gath-erer and small-scale horticulturalsocieties. Although census data on

such societies are limited, those thatare available suggest that there is in-deed a consistent group size in theregion of 150 individuals (Fig. 7). Ex-cept among settled horticulturalists,where the village seems to be therelevant unit, this typically involvesthe set of individuals from whom over-night camps are easily and regularlyformed. Such groups are not oftenconspicuous as physical entities (theydo not often appear together in oneplace at one time), but they do invari-ably have important ritual functionsfor the individuals concerned. AmongAustralian aboriginals, for example,the relevant group is the clan, whichmeets from time to time in jamboreeswhere the rituals of life (marriagesand rites of passage) are enacted andtales of the old times are rehearsed toremind everyone who they are andwhy they hold a particular relation-ship to each other. Indeed, this genu-inely seems to be the largest group ofpeople who know everyone in thegroup as individuals at the level ofpersonal relationships. This is essen-tially the definition that holds in thecase of primates.

A more extensive exploration of hu-man groups in other contexts suggeststhat groupings of this size are wide-spread and form an important compo-nent of all human social systems, be-ing present in structures that rangefrom business organizations to thearrangement of farming communi-ties.56 Estimates of community sizefor two traditional farming communi-ties in the United States, Hutteritesand an East Tennessee mountain com-munity, and of actual social networksizes (from small-worlds experiments)(shown as triangles on the right side ofFig. 7) fit very closely within the rel-evant range of group sizes.

It is easy, of course, to play thenumerologist in this context by find-ing groups that fit whatever group sizeone wishes to promote. The importantfeature to note here, however, is thatthe various human groups that can beidentified in any society seem to clus-ter rather tightly around a series ofvalues (5, 12, 35, 150, 500, and 2,000)with virtually no overlap in the vari-ance around these characteristic val-ues. They seem to represent points ofstability or clustering in the degrees offamiliarity within the broad range of

Figure 7. Mean sizes for different types of groups in traditional human societies. Individualsocieties are ordered along the bottom, with data for three main types of social groups(overnight camps, clans or villages, and tribes). Societies include hunter-gatherer and settledhorticulturalists from Australia, Africa, Asia, and North and South America. The triangles givemean group sizes for three contemporary United States samples: mean network size fromsmall-worlds experiments (N 5 2),67 mean Hutterite community size,68 and the size of an EastTennessee mountain community.69 The value of 150 predicted by the primate neocortex sizerelationship (from Fig. 1d) is indicated by the horizontal line, with 95% confidence intervalsshown as dashed lines.


human relationships, from the mostintimate to the most tenuous.

COGNITIVE MECHANISMSThe suggestion that the mecha-

nisms involved in these processes maybe concerned with social skills raisesthe issue alluded to by the originalMachiavellian intelligence hypothesis,namely to what extent cognitively so-phisticated mechanisms conferring theability to ‘‘mind-read’’ might be in-volved. Tactical deception, in its strongsense, implies the ability to hold falsebeliefs and, thus, the presence of theability known as ‘‘theory of mind’’(ToM). Of course, tactical deception aspracticed by primates on a daily basismay not, as Byrne26 himself haspointed out, be quite as sophisticatedas first impressions suggest. A moreconventional behaviorist accountbased on simple associative learningcan invariably be given for almost allexamples reported in the literature.

Nonetheless, convincing evidencesuggests that humans at least do useToM in executing some of their moremanipulative social activities. Andwhile we may not wish to attribute fullToM to all primates, at least circum-stantial evidence suggests that basicToM is present in great apes and thatmonkeys may aspire to a level thatByrne26 has described as level 1.5 in-tentionality (full ToM being level 2intentionality) (see Box 2). The differ-ence has been summed up rather

graphically by Cheney and Seyfarth’s57

observation that apes seem to be goodpsychologists in that they are good atreading minds, whereas monkeys aregood ethologists in that they are goodat reading behavior—or at least atmaking inferences about intentions inthe everyday sense, even if not in the

philosophical sense of belief states.Evidence that chimpanzees aspire

to at least a basic form of ToM isprovided by their performance on ex-perimental false-belief tasks.58–61 Thesestudies have attempted to develop ana-logues of the classic false-belief tasksused with children.62 Though it is clearthat chimpanzees do not perform tothe level at which fully competentchildren perform, O’Connell’s61 experi-ments at least suggest that they canperform at the level of children who

stand on the threshold of acquiringToM. More importantly, chimpanzeesdo better than autistic adults, one ofwhose defining features is the lack ofToM, on the same tests.

That mind-reading, the basis of ToM,is difficult to do has been shown byexperiments on normal adults testedon ‘‘advanced’’ ToM tasks, up to fifth-order intentionality.62 These data sug-gest that normal humans find tasks ofgreater than fourth-order intentional-ity exceedingly hard to do. The higherror rates at these levels do not reflecta memory retention problem: All sub-jects pass the tests that assess memoryfor the story line. Moreover, the samesubjects show considerable compe-tence on reasoning tasks that involvecausal chains of up to the sixth order.The difficulty seems genuinely to besomething to do with operating withdeeply embedded mental states.

One possibly significant observationin this context is that the visual andnonvisual components of the primateneocortex do not increase isometri-cally. Although initially there is a moreor less linear increase in the visualarea V1 with increasing size of the restof the neocortex, this drops off withinthe great ape clade. From gorillasthrough humans, increases in the sizeof the visual area progress more slowlythan do increases in the size of the restof the neocortex.28 We interpret this asimplying that beyond a certain pointthe acuity of the visual system does

Box 2. A Beginner’s Guide to Intensionality

Computers can be said to knowthings because their memories con-tain information; however, it seemsunlikely that they know that they knowthese things, in that we have no evi-dence that they can reflect on theirstates of ‘‘mind.’’ In the jargon of thephilosophy of mind, computers arezero-order intensional machines. In-tensionality (with an 2s) is the termthat philosophers of mind use to referto the state of having a state of mind(knowing, believing, thinking, wanting,understanding, intending, etc).

Most vertebrates are probably ca-pable of reflecting on their states of

mind, at least in some crude sense:they know that they know. Organismsof this kind are first-order intensional.By extension, second-order inten-sional organisms know that someoneelse knows something, and third-orderintensional organisms know that some-one else knows that someone elseknows something. In principle, the se-quence can be extended reflexivelyindefinitely, although, in practice, hu-mans rarely engage in more thanfourth-order intensionality in everydaylife and probably face an upper limit atsixth-order (‘‘Peter knows that Janebelieves that Mark thinks that Paula

wants Jake to suppose that Ameliaintends to do something’’).

A minimum of fourth-order inten-sionality is required for literature thatgoes beyond the merely narrative (‘‘thewriter wants the reader to believe thatcharacter A thinks that character Bintends to do something’’). Similar abili-ties may be required for science, sincedoing science requires us to askwhether the world can be other than itis (a second-order problem at the veryleast) and then ask someone else todo the same (an additional order ofintensionality).

. . . apes seem to begood psychologists inthat they are good atreading minds, whereasmonkeys are goodethologists in that theyare good at readingbehavior . . .


not increase linearly with size. Be-cause the total size of the neocortex islimited by embryological and ener-getic factors, this means that dispro-portionately more capacity can bededicated to nonvisual areas of theneocortex once the volume is abovethe crucial threshold. This might ex-plain why apes appear to be capable ofthe additional cognitive processing as-sociated with mind reading, whereasmonkeys are not. It might also explainwhy humans are better at it than apes.

For humans, one important aspectof ToM concerns its relevance to lan-guage, a communication medium thatcrucially depends on understandinginterlocutors’ mental states or inten-tions. The kinds of metaphorical usesof language that characterize not onlyour rather telegraphic everyday ex-changes (in which ‘‘you know what Imean?’’ is a common terminal clause)but also lies at the very heart of themetaphorical features of language. Asstudies of pragmatics have amply dem-onstrated,63 a great deal of linguisticcommunication is based on metaphor:Understanding the intentions behinda metaphor is crucial to successfulcommunication. Failure to understandthese intentions commonly results inconfusion or inappropriate responses.Indeed, without these abilities it isdoubtful whether literature, notablypoetry, would be possible. Our conver-sations would be confined to the ba-nally factual; those fine nuances ofmeaning that create both the ambigu-ities of politeness and the subtleties ofpublic relations would not be pos-sible.64

REFERENCES1 Clutton-Brock TH, Harvey PH (1980) Primates,brains and ecology. J Zool Lond 190:309–323.2 Aiello LC, Wheeler P (1995) The expensivetissue hypothesis. Curr Anthropol 36:184–193.3 Parker ST, Gibson KR (1977) Object manipula-tion, tool use and sensorimotor intelligence asfeeding adaptations in great apes and cebusmonkeys. J Hum Evol 6:623–641.4 Gibson KR (1986) Cognition, brain size and theextraction of embedded food resources. In Else J,Lee PC (eds), Primate Ontogeny, Cognition andSocial Behaviour, pp 93–104. Cambridge: Cam-bridge University Press.5 Whiten A, Byrne R (1988) Tactical deception inprimates. Behav Brain Sci 12:233–273.6 Byrne R, Whiten A (eds) (1988) MachiavellianIntelligence. Oxford: Oxford University Press.7 Harcourt AH (1988) Alliances in contests andsocial intelligence. In Byrne R, Whiten A (eds),Machiavellian Intelligence, pp 142–152. Oxford:Oxford University Press.

8 Harcourt AH (1989) Sociality and competitionin primates and non-primates. In Standen V,Foley R (eds), Comparative Socioecology, pp 223–242. Oxford: Blackwell Scientific.

9 Brothers L (1990) The social brain: A projectfor integrating primate behaviour and neurophysi-ology in a new domain. Concepts Neurosci 1:27–251.

10 Barton RA, Dunbar RLM (1997) Evolution ofthe social brain. In Whiten A, Byrne R (eds),Machiavellian Intelligence, Vol. II. Cambridge:Cambridge University Press.

11 Finlay BL, Darlington RB (1995) Linked regu-larities in the development and evolution of mam-malian brains. Science 268:1678– 1684.

12 Martin RD (1981) Relative brain size andmetabolic rate in terrestrial vertebrates. NatureLond 293:57–60.

13 Martin RD (1983) Human brain evolution inan ecological context. 52nd James Arthur Lec-ture, American Museum of Natural History, NewYork.

14 Hofman MA (1983) Evolution of the brain inneonatal and adult placental mammals: A theo-retical approach. J Theoret Biol 105:317–322.

15 Pagel M, Harvey PH (1988) How mammalsproduce large-brained offspring. Evolution 42:948–957.

16 Boesch C, Boesch H (1984) Mental map inchimpanzees. Primates 25:110–170.

17 MacLean PD (1982) On the origin and progres-sive evolution of the triune brain. In ArmstrongE, Falk D (eds), Primate Brain Evolution, pp291–310. New York: Plenum Press.

18 Joffe T, Dunbar RIM (n.d.). Primate brainsystem evolution. Brain Behav Evol, submittedfor publication.

19 Passingham RE (1982) The Human Primate.San Francisco: Freeman.

20 Jersion HJ (1973) Evolution of the Brain andIntelligence. New York: Academic Press.

21 Dunbar RIM (1989) Environmental determi-nants of intraspecific variation in body weight inbaboons. J Zool London 220:167–169.

22 Willner LA (1989) Sexual Dimorphism in Pri-mates. Ph.D. thesis, University of London.

23 Deacon TW (1990) Fallacies of progressionindices in theories of brain-size evolution. Int JPrimatol 12:193–236.

24 Dunbar RIM (1992) Neocortex size as a con-straint on group size in primates. J Hum Evol20:469–493.

25 Dunbar RIM, Joffe TH (n.d.) Neocortex sizeand social group size in prosimians. Primates,submitted for publication.

26 Byrne RB (1995) The Thinking Primate. Ox-ford: Oxford University Press.

27 Barton RA (1995) Neocortex size and behav-ioural ecology in primates. Proc R Soc London B,263:173–177.

28 Joffe TH, Dunbar RIM (1997) Visual andsocio-cognitive information processing in pri-mate brain evolution. Proc R Soc London B,264:1303–1307.

29 Stephan H, Frahm H, Baron G (1981) Newand revised data on volumes of brain structuresin insectivores and primates. Folia Primatol35:1–29.

30 Sawaguchi T, Kudo H (1990) Neocortical devel-opment and social structure in primates. Pri-mates 31:283–290.

31 Barton RA, Purvis A (1994) Primate brainsand ecology: Looking beneath the surface. InAnderson JR, Thierry B, Herrenschmidt N (eds),Current Primatology, pp 1–12. Strasbourg: Univer-sity of Strasbourg Press.

32. O’Keefe J, Nadel L (1978) The Hippocampusas a Cognitive Map. Oxford: Oxford UniversityPress.

33. Krebs JH, Sherry DF, Healy SD, Perry VH,Vaccarino AL (1989) Hippocampal specialisationof food-storing birds. Proc Natl Acad Sci USA86:1488–1492.

34. Aiello LC, Dunbar RIM (1993) Neocortexsize, group size and the evolution of language.Curr Anthropol 34:184–193.

35 Dunbar RIM (1995) Neocortex size and groupsize in primates: A test of the hypothesis. J HumEvol 28:287–296.

36. Bearder SK (1987) Lorises, bushbabies andtarsiers: Diverse societies in solitary foragers. InSmuts B, Cheney D, Seyfarth R, Wrangham R,Struhsaker T (eds), Primate Societies, pp 12–24.Chicago: Chicago University Press.

37 Dunbar RIM, Bever J (n.d.) Neocortex sizedetermines group size in carnivores and insecti-vores. Ethology, in press.

38 Gittleman JH (1986) Carnivore brain size,behavioural ecology and phylogeny. J Mammal67:23–36.

39 Marino L (1996) What can dolphins tell usabout primate evolution? Evol Anthropol 5:81–86.

40 Tschudin A (1996) The Use of Neuroimagingin the Assessment of Brain Size and Structure inOdontocetes. MSc thesis, University of Natal(Durban).

41 Barton RA, Purvis A, Harvey PH (1995) Evolu-tionary radiation of visual and olfactory brainsystems in primates, bats and insectivores. PhilosTrans R Soc B 348:381–392.

42 Kolb B, Wishaw LO (1996) Fundamentals ofHuman Neuropsychology. San Francisco: Free-man.

43 Fletcher P, Happe F, Frith U, Baker SC, DolanRJ, Frakowiak RSJ, Frith CD (1996) Other mindsin the brain: A functional imaging study of ‘‘theoryof mind’’ in story comprehension. Cognition 00:000–000.

44 Frith C (1996) Brain mechanisms for ‘‘havinga theory of mind.’’ J Psychopharmacol 10:9–16.

45 Armstrong E, Clarke MR, Hill EM (1987)Relative size of the anterior thalamic nuclei differ-entiates anthropoids by social system. Brain Be-hav Evol 30:263–271.46 Keverne EB, Martel FL, Nevison CM (1996)Primate brain evolution: Genetic and functionalconsiderations. Proc R Soc Lond B 262:689–696.47 Pawlowski B, Dunbar R, Lowen C (n.d.) Neo-cortex size, social skills and mating success inprimates. Behaviour, in press.48 Joffe TH (1997) Social pressures have selectedfor an extended juvenile period in primates. JHum Evol, in press.49 Bennett PM, Harvey PH (1985) Brain size,development and metabolism in birds and mam-mals. J Zool London 207:491–509.50 Marino L (1997) The relationship betweengestation length, encephalisation and body weightin odontocetes. Marine Mammal Sci 14:143–148.51 Kudo H, Lowen S, Dunbar RIM (n.d.) Neocor-tex size and social network size in primates.Behaviour, submitted for publication.52 Sugawara K (1984) Spatial proximity andbodily contact among the central Kalahari San.Afr Studies Monogr 3.53 Dunbar RIM, Spoor M (1995) Social net-works, support cliques and kinship. Hum Nature6:273–290.54 Dunbar RIM (1988) Primate Social Systems.London: Chapman & Hall.55 Dunbar RIM (1989) Social systems as optimalstrategy sets. In Standen V, Foley R (eds), Com-parative Socioecology, pp 141–149. Oxford: Black-well Scientific.


56 Dunbar RIM (1993) Coevolution of neocorti-cal size, group size and language in humans.Behav Brain Sci 11:681–735.57 Cheney DL, Seyfarth RM (1990) How MonkeysSee the World. Chicago: Chicago University Press.58 Povinelli DJ, Nelson KE (1990) Inferencesabout guessing and knowing in chimpanzees. JComp Psychol 104:203–210.59 Povinelli DJ (1994) What chimpanzees (might)know about the mind. In Wrangham R, McGrewW, de Waal F, Heltne P (eds), Chimpanzee Cul-tures, pp 285–300. Cambridge: Harvard Univer-sity Press.60 O’Connell SM (1996) Theory of Mind in Chim-panzees. Ph.D. thesis, University of Liverpool.61 Perner J, Wimmer D (1985) ‘‘John thinks thatMary thinks that. . . .’’ Attribution of second-order beliefs by 5 and 10 year-old children. J ExpChild Psychol 39:437–471.

62 Kinderman P, Dunbar RIM, Bentall RP (n.d.)Theory of mind deficits and causal attributions.Br J Psychol, in press.63 Desalles J-L (n.d.) Altruism, status and theorigin of relevance. In Hurford JR, Studdert-Kennedy M, Knight C (eds), Evolution of Lan-guage. Cambridge: Cambridge University Press.64 Dunbar RIM (1997) Grooming, Gossip and theEvolution of Language. Cambridge: Harvard Uni-versity Press.65 Harvey PH, Krebs JR (1990) Comparing brains.Science 249:150–156.66 Stephan H (1972) Evolution of primate brains:A comparative anatomical investigation. In TuttleR (ed), Functional and Evolutionary Biology ofPrimates, pp 165–174. Chicago: Aldine-Atherton.67 Killworth PD, Bernard HR, McCarty C (1984)Measuring patterns of acquaintanceship. CurrAnthropol 25:391–397.

68 Mange A, Mange E (1980) Genetics: HumanAspects. W.B. Saunders.

69 Bryant FC (1981) We’re All Kin: A CulturalStudy of a Mountain Neighbourhood. Knoxville:University of Tennessee Press.

70 Gould SJ (1975) Allometry in primates, withemphasis on scaling and the evolution of thebrain. Contrib Primatol 5:244–292.

71 Jolly A (1969) Lemur social behaviour andprimate intelligence. Science 163:501–506.

72 Humphrey NK (1976) The social function ofintellect. In Bateson PPG, Hinde RA (eds) Grow-ing Points in Ethology, pp 303–317. Cambridge:Cambridge University Press.

73 Armstrong E (1985) Relative brain size inmonkeys and prosimians. Am J Phys Anthropol66:263–273.

r 1998 Wiley-Liss, Inc.


the social brain hypothesis - evergreen state...

Documents