The Social Brain Hypothesis

178 Evolutionary Anthropology
ARTI CLES
The Social Brain Hypothesis
Robin I.M. Dunbar
Conventional wisdom over the past 160 years in the cognitive and neurosciences
has 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 brains
evolved to deal with essentially ecological problem-solving tasks.1
Th e con cen su s view h as tradition ally been th at brain s evolved to process in for m a tion of ecologica l r elevan ce. Th is view, h owever, ign ores an
im portan t con sideration : Brain s are
exceedin gly expen sive both to evolve
an d to m ain tain . Th e adu lt h u m an
brain weigh s abou t 2% of body weigh t
bu t con su m es abou t 20% of total en ergy in take.2 In th e ligh t of th is, it is
difficu lt to ju stify th e claim th at prim ates, an d especially h u m an s, n eed
larger brain s th an oth er species m erely
to do th e sam e ecological job. Claim s
th at prim ate ecological strategies in volve m or e com p lex p r ob lem -solvin g 3,4 are plau sible wh en applied to
th e beh aviors of particu lar species,
su ch as term ite-extraction by ch im pan zees an d n u t-crackin g by Cebu s m on keys, b u t fa il to exp la in wh y a ll
prim ates, including those that are conventional folivores, require larger brains
than those of all other m am m als.
An altern ative h ypoth esis offered
du rin g th e late 1980s was th at prim ates’ large brain s reflect th e com pu -
Robin Dunbar is Professor of Evolutionary
Psychology and Behavioural Ecology at
the University of Liverpool, England. His
research primarily focuses on the behavioral ecology of ungulates and human and
nonhuman primates, and on the cognitive
mechanisms and brain components that
underpin the decisions that animals make.
He runs a large research group, with graduate students working on many different
species on four continents.
Key Words: brain size, neocortex, social brain
hypothesis, social skills, mind reading, primates
tation al dem an ds of th e com plex socia l system s th a t ch a r a cter ize th e
order.5,6 Prim a facie, th is su ggestion
seem s plau sible: Th ere is am ple eviden ce th at prim ate social system s are
m ore com plex th an th ose of oth er
species. Th ese system s can be sh own
to in volve processes su ch as tactical
deception 5 an d coalition -form ation ,7,8
wh ich are rare or occu r on ly in sim pler form s in oth er taxon om ic grou ps.
Becau se of th is, th e su ggestion was
rapidly du bbed th e Mach iavellian in telligen ce h ypoth esis, alth ou gh th ere is a
growin g preferen ce to call it th e social
brain h ypoth esis.9,10
Plau sible as it seem s, th e social brain
h ypoth esis faced a problem th at was
recogn ized at an early date. Specifically, wh at qu an titative em pirical eviden ce th ere was ten ded to favor on e or
th e oth er of th e ecological h ypoth eses,1 wh ereas th e eviden ce addu ced
in favor of th e social brain h ypoth esis
was, at best, an ecdotal.6 In th is article,
I sh all first sh ow h ow we can test
between th e com petin g h ypoth eses
m ore con clu sively an d th en con sider
som e of th e im plication s of th e social
brain h ypoth esis for h u m an s. Fin ally,
I sh all briefly con sider som e of th e
u n derlyin g cogn itive m ech an ism s th at
m igh t be in volved.
TESTING BETWEEN ALTERNATIVE
HYPOTHESES
To test between th e com petin g h ypoth eses, we n eed to force th e h ypoth eses in to con flict in su ch a way th at
th eir prediction s are m u tu ally con tradictory. Th is allows th e data to dis-
cr im in a te u n equ ivoca lly b etween
th em . In th e presen t case, we can do
th is by askin g wh ich h ypoth esis best
predicts th e differen ces in brain size
across th e prim ate order. To do so, we
n eed to iden tify th e specific qu an titative prediction s m ade by each h ypoth esis an d to determ in e an appropriate
m easu re of brain size.
Th e fou r classes of h ypoth eses th at
h ave been pu t forward to explain prim ate brain evolu tion are epiph en om en al, developm en tal, ecological, an d
social in orien tation (Table 1). Th e
epiph en om en al an d developm en tal h ypoth eses sh are th e assu m ption th at
evolu tion of th e brain (or brain part) is
n ot a con sequ en ce of extern al selection pressu res bu t rath er sim ply a
con sequ en ce of som eth in g to do with
th e way biological growth processes
are organ ized. Th e epiph en om en al h ypoth eses th u s argu e th at brain evolu tion is a m ere byprodu ct of body size
evolu tion , an d th at brain part size is,
in tu rn , a byprodu ct of total brain
evolu tion .11
Th e developm en tal version s differ
on ly in th at th ey provide a m ore specific m ech an ism by presu m in g th at
m atern al m etabolic in pu t is th e critical factor in flu en cin g brain developm en t. Th is claim is given credibility by
th e fact th at th e bu lk of brain growth
in m am m als occu rs pren atally. In deed, it appears to be th e com pletion
of brain developm en t th at precipitates
birth in m am m als, with wh at little
postn atal brain growth occu rs bein g
com pleted by th e tim e an in fan t is
wean ed. From th is, th e con clu sion is
drawn th at brain evolu tion m u st be
con strain ed by th e spare en ergy, over
an d above h er basal m etabolic requ irem en ts, th at th e m oth er h as to ch an n el
in to fetal developm en t.12–14 Som e eviden ce in su pport of th is claim com es
from th e fact th at fru givorou s prim ates h ave larger adu lt brain s relative
ARTI CLES
Evolutionary Anthropology 179
TABLE 1. Hypotheses Used to Explain the
Evolution of Large Brains in Primates
Hypothesis
A. Epiphenomenal
hypotheses
1. Large brains (or brain parts)
are an unavoidable consequence of having a large
body (or brain)
B. Ecological hypotheses
2. Frugivory imposes higher
cognitive demands than
folivory does
3. Brain size constrains the size
of the mental map:
(a) constraint on size of
home range
(b) constraint on inertial
navigation (day journey
length)
4. Extractive foraging
hypothesis
C. Social hypotheses
5. Brain size constrains size of
social network (group size):
(a) constraint on memory
for relationships
(b) constraint on social skills
to manage relationships
D. Developmental
hypotheses
6. Maternal energy constraints determine energy
capacity for fetal brain
growth
Sources
11, 70
1, 65
1
3, 4
6, 71,
72
12, 13
46, 55,
73
to body size th an do folivorou s prim ates.1 Th is h as been in terpreted as
im plyin g th at fru givores h ave a rich er
diet th an folivores do an d th u s h ave
m ore spare en ergy to divert in to fetal
growth . Large brain s are th u s seen as
a kin d of em ergen t epigen etic effect of
spare capacity in th e system .
Both kin ds of explan ation s su ffer
from th e problem th at th ey ign ore a
fu n dam en tal prin ciple of evolu tion ary
th eory, wh ich is th at evolu tion is th e
ou tcom e of th e balan ce between costs
an d ben efits. Becau se th e cost of m ain tain in g a large brain is so great, it is
in trin sically u n likely th at large brain s
will evolve m erely becau se th ey can .
Large brain s will evolve on ly wh en th e
selection factor in th eir favor is sufficien t to overcom e th e steep cost gradien t. Developm en tal con strain ts are u n dou btedly im portan t, bu t rath er th an
bein g cau sal th eir role is th at of a
con strain t th at m u st be overcom e if
larger brain s are to evolve. In addition ,
Pagel an d H arvey15 h ave sh own th at
th e en ergetic argu m en ts do n ot add
u p: Precocial m am m als do n ot h ave
h igh er m etabolic rates th an do altricial m am m als despite th e fact th at
th ey h ave n eon atal brain sizes th at
are, on average, twice as large. We
th erefore do n ot n eed to con sider eith er epiph en om en al or developm en tal
h ypoth eses an y fu rth er in th e con text
of th is article.
Th is does n ot n ecessarily m ean th at
th ese explan ation s are wron g. Both
kin ds of explan ation m ay be tru e in
th e sen se th at th ey correctly iden tify
developm en tal con strain ts on brain
growth , bu t th ey do n ot tell u s wh y
brain s actu ally evolved as th ey did.
Th ey m ay tell u s th at if you wan t to
evolve a large brain , th en you m u st
evolve a large body in order to carry
th e en ergetic costs of doin g so or a diet
Because the cost of
maintaining a large brain
is so great, it is
intrinsically unlikely that
large brains will evolve
merely because they
can. Large brains will
evolve only when the
selection factor in their
favor is sufficient to
overcome the steep cost
gradient.
th at en su res sufficien t en ergy to provide for fetal brain developm en t. No
su ch allom etric argu m en t can ever
im ply th at you h ave to evolve a large
brain or a large body. Th e large brain
or brain part is a cost th at an im als
m u st factor in to th eir calcu lation s
wh en con siderin g wh eth er or n ot a
large body or a large brain is a sen sible
solu tion to a p a r ticu la r ecologica l
problem . Sh ifts to m ore en ergy-rich or
m ore easily processed diets m ay be
essen tial precu rsors of sign ifican t in creases in brain or brain part size.2
Th is wou ld explain wh y fru givores
h ave larger brain s th an folivores do
an d wh y h om in ids h ave larger brain s
th an great apes do.
Th is leaves u s with ju st two classes
of h ypoth eses, th e ecological an d th e
social. At least th ree version s of th e
form er can be iden tified, wh ich I will
term th e dietary, m en tal m aps, an d
extractive foragin g h ypoth eses. In essen ce, th ese argu e, respectively, th at
prim ate species will n eed larger brain s
if (i) th ey are fru givorou s becau se fru its
are m ore eph em eral an d patch y in
th eir distribu tion th an leaves are, an d
h en ce requ ire m ore m em ory to fin d;
(ii) th ey h ave larger ran ges becau se of
th e greater m em ory requ irem en ts of
large-scale m en tal m aps; or (iii) th eir
diet requ ires th em to extract resou rces
from a m atrix in wh ich th ey are em bedded (e.g., th ey m u st rem ove fru it
pu lp from a case, stim u late gu m flow
from a tree, extract term ites from a
term itariu m , or h u n t species th at are
cryptic or beh ave evasively).
For obviou s reason s, I u sed th e percen tage of fru it in th e diet as an appropriate in dex for th e dietary h ypoth esis.
I u sed th e size of th e ran ge area an d
th e len gth of th e day jou rn ey as altern ative in dices for th e m en tal m appin g
h ypoth esis, th ou gh I presen t th e data
on ly for th e first of th ese h ere. Th e first
in dex correspon ds to th e case in wh ich
an im als h ave to be able to m an ipu late
in form ation abou t th e location s of resou rces relative to th em selves in a
E u clidean space (an exam ple wou ld
be th e n u t-crackin g activities of th e
Taı̈ ch im pan zees 16 ); th e secon d correspon ds to th e possibility th at th e con strain t lies in th e n eeds of som e aspect
of in ertial n avigation . Th e extractive
foragin g h ypoth esis is less easy to ch aracterize in qu an titative term s becau se
th ere is n o objective m easu re of th e
degree to wh ich diets vary in th eir
extractiven ess. H owever, Gibson 4 provided a classification of prim ate species in to fou r categories of diet th at
differ in th eir degree of extractiven ess.
We can test th is h ypoth esis by askin g
wh eth er th ere is a con sisten t variation
in brain size am on g th ese fou r categories, with th e species h avin g th e m ore
extractive diets h avin g larger brain s
th an th ose with th e less extractive
diets.
Fin ally, we n eed an in dex of social
com plexity. In m y origin al an alyses, I
u sed social grou p size as a sim ple
m easu re of social com plexity. Alth ou gh
at best rath er cru de, th is m easu re
n on eth eless captu res on e aspect of th e
ARTI CLES
180 Evolutionary Anthropology
th e brain su ch as th e m edu lla, th e
n eocortex sh ows dram atic an d in creasin g expan sion across th e ran ge of prim ates (Fig. 1). Th e n eocortex is app r oxim a tely th e sa m e size a s th e
m edu lla in in sectivores; h owever, it is
abou t 10 tim es larger th an th e m edu lla in prosim ian s an d 20–50 tim es
larger in th e an th ropoids, with th e
h u m an n eocortex bein g as m u ch as
105 tim es th e size of th e m edu lla.
Th is su ggests th at rath er th an lookin g at total brain size, as previou s
stu dies h ave don e, we sh ou ld in fact be
con siderin g th e brain system , n am ely
th e n eocortex, th at h as been m ain ly
respon sible for th e expan sion of th e
prim ate brain . From th e poin t of view
of all th e h ypoth eses of prim ate brain
evolu tion , th is m akes sen se: Th e n eoFigure 1. Neocortex volume as a ratio of medulla volume in different groups of primates (after
Passingham19). Source: Stephan et al29
com plexity of social grou ps, th e fact
th at in form ation -processin g dem an ds
can be expected to in crease as th e
n u m ber of relation sh ips in volved in creases. More im portan tly, perh aps,
th is m easu re h as th e distin ct m erit of
b ein g ea sily qu a n tified a n d wid ely
available. Alth ou gh it is possible to
con ceive of a n u m ber of better m easu res of social com plexity, th e appropriate data are rarely available for
m ore th an on e or two species.
Th e secon d problem con cern s th e
m ost appropriate m easu re of brain
evolu tion . H ith erto, m ost stu dies h ave
con sidered th e brain as a sin gle fu n ction al u n it. Th is view h as been rein forced by Fin lay an d Darlin gton ,11 wh o
argu ed th at th e evolu tion in brain part
size closely correlates with th e evolu tion of total brain size an d can be
explain ed sim ply in term s of allom etric con sequ en ces of in creases in total
brain size. H owever, Fin lay an d Darlin gton failed to con sider th e possibility th at ch an ges in brain size m igh t
actu ally be driven by ch an ges in its
parts rath er th an in th e wh ole brain .
Th is is especially tru e of th e n eocortex, for its volu m e accou n ts for 50% to
80% of total brain volu m e in prim ates.
Th u s, ch an ges in th e volu m e of th e
n eocortex in evitably h ave a large direct effect on apparen t ch an ge in brain
volu m e th at m ay be qu ite u n related to
ch an ges in oth er brain com pon en ts.
Th is poin t is given weigh t by th e fact
th at Fin lay an d Darlin gton th em selves
sh owed th at n eocortex size is an expon en tial fu n ction of brain size, wh ereas
oth er brain com pon en ts are n ot.
Fin la y a n d Da r lin gton 11 n otwith stan din g, th ere is eviden ce th at brain
evolu tion h as n ot been a h istory of
sim p le exp a n sion in tota l volu m e.
Rath er, brain evolu tion h as been m osaic in ch aracter, with both th e rate
an d th e exten t of evolu tion h avin g
varied between com pon en ts of th e system . MacLean 17 poin ted ou t m an y
years ago th at prim ate brain evolu tion
can be viewed in term s of th ree m ajor
system s (h is con cept of th e triu n e
brain ). Th ese system s correspon d to
th e basic reptilian brain (h in d- an d
m idbrain system s), th e m am m alian
brain (palaeocortex, su bcortical system s), an d th e prim ate brain (broadly,
th e n eocor tex). A m or e im p or ta n t
poin t, perh aps, is th at variation s can
be fou n d with in th ese broad categories in th e rates at wh ich differen t
com pon en ts expan ded, wh ich , in at
least som e cases, h ave been sh own to
correlate with ecological factors. 10 Partiallin g ou t th e effects of body size on
th e size of brain com pon en ts su ggests
th at th e story m ay be m ore com plex
th a n Fin la y a n d Da r lin gton 11 su p posed, with som e rem odelin g of brain
growth pattern s occu rrin g in th e tran sition s between in sectivores, prosim ian s, an d an th ropoids.18
Th e im portan t poin t in th e presen t
con text is th at, as Passin gh am 19 n oted,
relative to th e m ore prim itive parts of
. . . brain evolution has
not been a history of
simple expansion in total
volume. Rather, brain
evolution has been
mosaic in character,
with both the rate and
the extent of evolution
having varied between
components of the
system.
cortex is gen erally regarded as bein g
th e seat of th ose cogn itive processes
th at we associate with reason in g an d
con sciou sn ess, an d th erefore m ay be
expected to be u n der th e m ost in ten se
selection from th e n eed to in crease or
im prove th e effectiven ess of th ese processes.
On e addition al problem n eeds to be
resolved. In h is sem in al stu dy of brain
evolu tion , Jerison 20 argu ed th at brain
size can be expected to vary with body
size for n o oth er reason th an fu n dam en tal allom etric relation sh ips associated with th e n eed to m an age th e
ph ysiological m ach in ery of th e body.
Wh at is of in terest, h e su ggested, is
n ot absolu te brain size, bu t th e spare
brain capacity over an d above th at
n eeded to m an age body m ech an ism s.
ARTI CLES
For th is reason , Jerison derived h is
en ceph alization qu otien t. All su bsequ en t stu dies h ave u sed body size as
th e appropriate baselin e again st wh ich
to m easu re relative deviation s in brain
size. H owever, a problem h as sin ce
em erged: Brain size is determ in ed
early in developm en t an d, com pared
to m an y oth er body system s, appears
to be h igh ly con servative in evolu tion ary term s. As a resu lt, body size can
often ch an ge dram atically both on togen etica lly a cr oss p op u la tion s in r espon se to local en viron m en tal con dition s 21
and
p h ylogen etica lly 22,23
with ou t cor r esp on d in g ch a n ges in
brain size. Th is is particu larly con spicu ou s in th e case of ph yletic dwarfs
(e.g., callitrich ids an d perh aps m odern h u m an s an d h ylobatids 22 ) an d species in wh ich body size m ay h ave
in creased in respon se to predation
pressu re followin g th e occu pation of
m ore open terrestrial h abitats (e.g.,
papion ids 24 ).
Th e lability of body size th erefore
m akes it a poor baselin e, th ou gh on e
th at probably is adequ ate for an alyses
on th e m ou se-eleph an t scale. Con sequ en tly, it is n ecessary to fin d an in tern ally m ore con sisten t baselin e for taxon om ica lly fin e-gr a in ed a n a lyses.
Willn er 22 su ggested th at eith er m olar
tooth size or brain size m ay be su itable becau se both are developm en tally con servative. Becau se we are con cern ed with brain part size, som e
aspect of brain size seem s th e m ost
appropriate.
At th is poin t, th ree option s are available. On e is to com pare th e n eocortex,
th e brain part of in terest, with th e
wh ole brain ; th e secon d is to u se th e
rest of th e brain oth er th an th e part of
in terest; th e th ird is to u se som e less
variable prim itive com pon en t of th e
brain , su ch as th e m edu lla, as a baselin e. Two option s are in tu rn available
a s m ech a n ism s for con tr ollin g for
brain size in each of th ese cases. On e
is to u se residu als from a com m on
regression lin e again st th e baselin e
(e.g., th e residu al of n eocortex volu m e
on total brain volu m e or m edu la volu m e). Th e oth er ch oice is to u se ratios.
We h ave con sidered an d tested all
th ese option s 10,24 (see Box 1). Th e resu lts are virtu ally iden tical irrespective of wh ich m easu re is u sed. On e
explan ation for th is m ay be th at all
th ese m easu res actu ally in dex th e sam e
Evolutionary Anthropology 181
th in g, absolu te n eocortex size, m ain ly
becau se th e n eocortex is su ch a large
com pon en t of th e prim ate brain . In deed, th e u se of absolu te n eocortex
size produ ces resu lts th at are sim ilar
to th ose obtain ed from relativized in dices of n eocor tex volu m e. 24,25 Th is
m akes som e sen se in com pu tation al
term s: As Byrn e 26 h as poin ted ou t, a
10% in crease in th e processin g capacity of a sm all com pu ter is worth a
great deal less in in form ation -processin g term s th an is a 10% in crease in a
large com pu ter. Alth ou gh residu als
from a com m on regression lin e wou ld
con ven tion ally be con sidered th e safest m easu re, an d h ave been u sed in
m an y recen t an alyses,27,28 I sh all con -
The neocortex is
generally regarded as
being the seat of those
cognitive processes that
we associate with
reasoning and
consciousness, and
therefore may be
expected to be under
the most intense
selection from the need
to increase or improve
the effectiveness of
these processes.
tin u e to u se m y origin al ratio in dex
becau se it provides th e best predictor
(see Box 1).
Fin ally, it is n ow widely appreciated
th at com parative an alyses n eed to con trol for th e effects of ph ylogen etic
in ertia. Closely related species can be
expected to h ave sim ilar valu es for
m an y an atom ical an d beh avioral dim en sion s m erely by virtu e of h avin g
in h erited th em from a recen t com m on
an cestor. In su ch cases, plottin g raw
data wou ld resu lt in pseu doreplication , artificially in flatin g th e sam ple
size by assu m in g th at closely related
species are actu ally in depen den t evolu tion ary even ts. Th e ways of dealin g
with th is problem in clu de plottin g
m ean s for h igh er taxon om ic u n its, perform in g n ested an alyses of varian ce
u sin g ph ylogen etic levels as factors,
com parin g m atch ed pairs of species,
an d m akin g in depen den t con trasts th at
con trol directly for ph ylogen y. E ach
m eth od h as its own advan tages an d
disadvan tages, bu t th e first an d th ird
procedu res are particu larly associated
with loss of in form ation an d sm all
sam ple sizes. I sh all u se th e first an d
last m eth od, th e last becau se it allows
in dividu al species to be com pared, bu t
th e first becau se it allows grade sh ifts
with in data sets to be iden tified (a
problem th at in depen den t-con trasts
m eth ods h ave difficu lty dealin g with ).
I sh all take th e gen u s as a su itable
basis for an alysis becau se gen era typically represen t differen t reprodu ctive
or ecological radiation s an d th u s are
m ore likely to con stitu te in depen den t
evolu tion ary even ts.
Th e resu ltin g an alyses are relatively
straigh tforward: Figu re 2 presen ts th e
data for n eocortex ratio for th e an th ropoid prim ate species in th e data base
of Steph an , Frah m , an d Baron .29 Neocortex size, h owever m easu red, does
n ot correlate with an y in dex of th e
ecological h ypoth eses, bu t does correlate with social grou p size. Sim ilar
fin din gs were reported by Sawagu ch i
an d Ku do,30 wh o fou n d th at n eocortex
size correlated with m atin g system in
prim ates. Barton an d Pu rvis 31 h ave
con firm ed th at u sin g both residu als of
n eocortex volu m e on total brain volu m e an d th e m eth od of in depen den t
con trasts yields th e sam e resu lt. Both
Barton 10 an d T. Joffe (u n pu blish ed)
h ave repeated th e an alyses u sin g th e
m edu lla as th e baselin e for com parison . More im portan tly, Barton an d
Pu rvis 31 h ave sh own th at wh ile relative n eocortex volu m e correlates with
grou p size bu t n ot th e size of th e
ran gin g area, th e reverse is tru e of
relative h ippocam pu s size. A correlation between ran ge area an d h ippocam pu s size is to be expected becau se
of h ippocam pal in volvem en t in spatial
m em ory.32,33 Th is correlation dem on strates th at it is n ot sim ply total brain
size th at is im portan t (a poten tial problem , given th e overwh elm in g size of
ARTI CLES
182 Evolutionary Anthropology
Box 1. How to Measure Brains
R. Dunbar and Tracey H. Joffe
The different ways of measuring
relative brain size have raised doubts
as to the most appropriate technique
to use.65 Many researchers have preferred to use residuals from the common regression line of best fit for the
data set concerned. This provides a
measure of the extent to which brain
(or brain part) volume deviates from
what would be expected for an average member of the relevant taxon of
the appropriate size. Although ratios
have been used to compare the relative size of brain components,29,66 this
has been criticized on the grounds
that trade-offs within the brain may
mean that a given index simply measures total brain size (or the size of a
brain part) and thus does not remove
the effects of absolute size. Ratios
may also be prone to autocorrelation
effects, especially when the baseline
is taken to be the whole brain and, as
in the case of the neocortex, the part in
question is a major volumetric component of the brain.
Although there are likely to be
some trade-offs of this kind within the
brain, the fact that neocortex volume
increases progressively across the primate order suggests that such constraints are less likely to have a significant effect on a ratio measure. Of
course the residuals procedure is itself
a ratio: Encephalization-type indices
are calculated as actual volume divided by predicted volume (which,
when data are logged, becomes the
conventional actual minus predicted
values). Thus, ratio measures per se
may not be the problem. Rather, the
substantive objection is whether or not
a ratio partials out the allometric effects of body size. In fact, it seems that
neocortex ratios are not correlated
with the basal brain (i.e., brain volume
excluding the neocortex) within major
taxonomic groups (unpublished analyses). Consequently, this criticism has
less force than it might appear to have
th e prim ate n eocortex). Moreover, it
poin ts to th e specific in volvem en t of
th e n eocortex.
on first sight. Moreover, any index that
uses the whole brain as its base is
likely to suffer from autocorrelation
effects. Because the neocortex is such
a large proportion of the brain in primates, residuals of neocortex from
total brain size may simply be a measure of neocortex plotted against itself.
To consider the problem in more
detail, we ran a stepwise regression
analysis on the 24 species of anthropoid primates, including humans, on
the data base of Stephan, Frahm, and
Baron,29 with group size as the dependent variable and nine indices of relative brain or brain-part volume as independent variables. In addition to
neocortex ratio, these included total
brain volume as well as telencephalon
and neocortex volume, each taken as
absolute volume and as a residual
from both body mass and brain volume. All variables were log10-transformed for analysis. In both cases,
neocortex ratio was selected as the
variable of first choice. We carried out
both regressions on generic plots and
independent contrast analyses. For
the contrasts analysis, the best fit
least-squares regression equation
through the origin was:
dently of all other confounding measures, suggests that more detailed
consideration needs to be given to its
significance and meaning. It may be,
for example, that body size, rather
than being a determinant20 is simply a
constraint on neocortex size: A species can evolve a large neocortex only
if its body is large enough to provide
the spare energy capacity through
Kleiber’s relationship for basal metabolic rate to allow for a larger than
average brain. This interpretation is
implied by the Aiello and Wheeler2
‘‘expensive tissue hypothesis.’’ It would
also be in line with Finlay and Darlington’s11 claim that in mammals the evolution of brain-part size is driven, developmentally at least, by the evolution of
the whole brain, thus generating very
tight correlations between brain-part
size and total brain size.
Contrast in log10 (group size) 5 3.834
21.72 0.101
* 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 a
significant contribution to the variability in group size in either analysis.
Table 2 gives the results for the independent contrasts analysis.
Neocortex ratio is thus the single
most powerful predictor of group size
in these species. While the biological
significance of this variable remains
open to interpretation, the fact that it
provides the best predictor, indepen-
Th e va lid ity of th is r ela tion sh ip
cou ld be tested directly by u sin g it to
predict grou p sizes in a sam ple of
TABLE 2. Stepwise Regression Analysis
of Indices of Brain Component Volume
as Predictors of Group Size in
Anthropoid Primates, Based on
Independent Contrasts Analysis*
Independent Variable
t
P
Absolute brain volume
Residual of brain volume
on body mass
Absolute telencephalon
volume
Residual of telencephalon volume on
body mass
Residual of telencephalon volume on
brain volume
Absolute neocortex
volume
Residual of neocortex
volume on body mass
Residual of neocortex
volume on brain
volume
Neocortex ratio (against
rest of brain)
21.69 0.107
20.91 0.371
0.61 0.546
1.69 0.107
21.70 0.104
0.56 0.583
1.60 0.136
3.79 0.001
*Sample: 24 species of anthropoid primates from Stephan, Frahm, and
Baron.29
species for wh ich brain volu m etric
data were n ot available in th e origin al
sa m p le of Step h a n , Fr a h m , a n d
ARTI CLES
Evolutionary Anthropology 183
Figure 2. Relative neocortex size in anthropoid primates plotted against (a) percentage of fruit in the diet, (b) mean home-range size scaled as
the 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.)
Baron .29 I did th is by exploitin g th e
fact th at n eocortex ratios can be predicted from total brain volu m e,34 a
resu lt th at, in fact, follows directly
from th e Fin lay an d Darlin gton 11 fin din gs. Th e resu lt was a sign ifican t fit
between predicted n eocortex ratio an d
observed m ean grou p size for a sam ple of 15 New an d Old World m on key
species.35
Ba r ton 27 n oted th a t th e or igin a l
an alyses of Du n bar 24 seem ed to im ply
th at variation in n eocortex size was
m u ch greater th an variation in grou p
size in th e prosim ian s. Usin g Du n bar’s 24 data on grou p size, Barton su ggested th at th e relation sh ip between
n eocortex an d grou p size did n ot apply in th e case of prosim ian s. H owever, th e data on prosim ian grou p
sizes in th is sam ple su ffered from a
pau city of data, particu larly for th e
n octu rn al species. Becau se m an y of
th ese are described as sem i-solitary, it
was con servatively assu m ed in th e
Du n bar 24 database th at th eir grou p
size was on e. More recen t field stu dies
h ave produ ced m arkedly im proved es-
tim ates of th e sizes of social grou ps
an d, in th e case of th e sem i-solitary
species, daytim e n est grou ps.36 Rean alysis of th e data for prosim ian s
u sin g th ese im proved estim ates of social grou p size su ggests th at th ese
species do in fact adh ere to th e sam e
relation sh ip between n eocortex an d
grou p size as th at wh ich pertain s for
oth er prim ates.25 More im portan tly,
th e regression lin e for th is taxon is
parallel to, bu t sh ifted to th e left of,
th at for oth er an th ropoid prim ates
(Fig. 3).
Th is r ela tion sh ip h a s n ow b een
sh own to h old for at least fou r oth er
m am m alian orders: bats,10 carn ivores
an d in sectivores,37,38 an d odon tocete
cetacean s.39,40 In th e case of th e in sectivores, th e data poin ts are sh ifted far
to th e left of th ose for th e prim ates, as
m igh t be expected of a taxon om ic
grou p th at is con sidered to be broadly
represen tative of th e an cestral m am m als.37 H owever, th e relation sh ip is
weak in th is case, probably becau se
estim ates of grou p size are particu larly u n certain for in sectivores.
Su rprisin gly, th e data for th e carn ivores m ap directly on to th ose for th e
sim ian prim ates, th at is, th e regression lin es for th e two data sets do n ot
differ sign ifican tly. H owever, th e carn ivores do n ot exh ibit as wide a ran ge of
n eocortex ratios or grou p sizes as do
an th ropoid prim ates. Th e fact th at th e
prosim ian s lie to th e left of both th ese
taxon om ic grou ps im plies th at th e carn ivores represen t an in depen den t evolu tion ary developm en t alon g th e sam e
prin ciples as th e an th ropoid prim ates,
th e differen ce bein g th at th ey ju st h ave
n ot taken it as far as prim ates h ave.
On e reason for th is m ay be th at th e
carn ivore social world is olfaction dom in ated rath er th an vision -dom in ated, as in th e case of th e prim ates.
Barton 10,27,41 h as poin ted ou t th at th e
sh ift to a diu rn al lifestyle based on
color vision , perh aps in itially dietdriven , bu t leadin g to a sh ift in to vision -based com m u n ication , m ay be
th e key featu re th at h as spu rred on th e
dram atic developm en t of th e prim ate
n eocortex.
ARTI CLES
184 Evolutionary Anthropology
size of th e cortical processin g m ach in ery in creases, at least relative to th e
opportu n ity cost of takin g cortical n eu ron s away from oth er cogn itive processes.
It seem s equ ally u n likely th at th e
problem lies with a pu re m em ory con strain t, th ou gh m em ory capacity obviou sly m u st im pose som e kin d of u pper
lim it on th e n u m ber of relation sh ips
th at an an im al can h ave. Th ere are
th ree reason s for th is claim . First, in
h u m an s at least, m em ory for faces is
an order of m agn itu de larger th an th e
predicted cogn itive grou p size: H u m an s are said to be able to attach
n am es to arou n d 2,000 faces bu t h ave
a cogn itive grou p size of on ly abou t
Figure 3. Mean group size plotted against neocortex ratio for individual genera, shown
separately for prosimian, simian, and hominoid primates. Prosimian group size data, from
Dunbar and Joffe,25 include species for which neocortex ratio is estimated from total brain
volume. 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, Propithecus; e, Indri; f, Microcebus; g, Galago; h, Hapalemur; i, Avahi; j, Perodictus.
REFINING THE RELATIONSHIP
Th e social brain h ypoth esis im plies
th at con strain ts on grou p size arise
from th e in form ation -processin g capacity of th e prim ate brain , an d th at
th e n eocortex plays a m ajor role in
th is. H owever, even th is proposal is
open to several in terpretation s as to
h ow th e relation sh ip is m ediated. At
least five possibilities can be u sefu lly
con sidered. Th e con strain t on grou p
size cou ld be a resu lt of th e ability to
recogn ize an d in terpret visu al sign als
for iden tifyin g eith er in dividu als or
th eir beh avior; lim itation s on m em ory
for faces; th e ability to rem em ber wh o
h as a relation sh ip with wh om (e.g., all
dyadic relation sh ips with in th e grou p
as a wh ole); th e ability to m an ipu late
in form ation abou t a set of relation sh ips; an d th e capacity to process em otion al in form ation , particu larly with
respect to recogn izin g an d actin g on
cu es to oth er an im als’em otion al states.
Th ese are n ot all n ecessarily m u tu ally
exclu sive, bu t th ey do iden tify differen t poin ts in th e cogn itive m ech an ism
th at m igh t be th e cru cial in form ation processin g bottlen eck.
Alth ou gh visu a l m ech a n ism s a r e
likely to be im portan t for social in teraction , an d m ay well h ave been th e
in itial kick for th e evolu tion of large
brain s in prim ates,10 it seem s in trin sically u n likely th at th e u ltim ate con strain t lies in th e m ech an ism s of th e
visu al system itself.28 Alth ou gh th ere
is a correlation between th e relative
size of th e visu al cortex an d grou p size
in an th ropoid prim ates, th e fit is m u ch
poorer, an d th e slope sign ifican tly sh allower th an th at between th e n on visu al
n eocortex an d grou p size (r2 5 0.31 vs
r2 5 0.61, respectively) (Fig. 4). Partial
correlation an alysis in dicates th at on ly
th e correlation for th e n on visu al relation sh ip rem ain s sign ifican t wh en th e
oth er com pon en t is h eld con stan t 28
(th ou gh th is is n ot tru e for prosim ian s 25 ). A m ore im portan t poin t is th at
th e volu m e of th e lateral gen icu late
n u cleu s, a m ajor su bcortical way station in visu al processin g, does n ot
correlate with grou p size at all, in dicatin g th at pattern recogn ition per se is
u n likely to be th e issu e.28 It m ay be of
som e sign ifican ce th at th e absolu te
size of th e visu al cortex seem s to reach
an asym ptotic valu e in th e great ape
clade, wh ereas th e n on visu al n eocortex con tin u es to in crease in size. On e
in terpretation of th is is th at visu al
processin g does n ot n ecessarily con tin u e to im prove in defin itely as th e
The social brain
hypothesis implies that
constraints on group size
arise from the
information-processing
capacity of the primate
brain, and that the
neocortex plays a major
role in this. However,
even this proposal is
open to several
interpretations as to how
the relationship is
mediated.
150. Secon d, th ere is n o in trin sic reason to su ppose th at m em ory per se is
th e issu e. Th e social brain h ypoth esis
is abou t th e ability to m an ipu late in form ation , n ot sim ply to rem em ber it.
Th ird, an d perh aps m ost sign ifican tly,
m em ories appear to be stored m ain ly
in th e tem poral lobes,42 wh ereas recen t PE T scan stu dies im plicate th e
prefron tal n eocortex, n otably Brodm an area 8, as th e area for social skills
a n d, specifically, th eory of m in d. 43
Frith 44 h as su ggested th at m em ories
a n d r ep r esen ta tion s for ob jects or
even ts m ay in volve in teraction s between several levels of th e n eocortex
depen din g on th e kin ds of operation s
ARTI CLES
Evolutionary Anthropology 185
th e rates with wh ich tactical deception
a r e u sed cor r ela te with n eocor tex
size.26 Species with large n eocortex
ratios m ake sign ifican tly m ore u se of
tactical deception , even wh en th e differen tial frequ en cies with wh ich th ese
large-brain ed species h ave been stu died are taken in to accou n t.
Th ir d , Pa wlowski, Du n b a r, a n d
Lowen 47 h ave sh own th at am on g polygam ou s prim ates th e m ale ran k correlation with m atin g su ccess is n egatively related to n eocortex size (Fig. 5).
Th is is ju st wh at we wou ld predict if
th e lower ran kin g m ales of species
with larger n eocortices were able to
u se th eir greater com pu tation al capacities to deploy m ore soph isticated social skills, su ch as th e u se of coalition s
a n d ca p ita lizin g on fem a le m a te
ch oice, to u n derm in e or circu m ven t
th e power-based strategies of th e dom in an t an im als.
Figure 4. Independent contrasts in mean group size plotted against contrasts in the visual
cortex and the volume of the rest of the neocortex (nonvisual neocortex) for individual
anthropoid species. Note that the visual cortex is here defined as visual area V1; the nonvisual
cortex is the non-V1 volume of the neocortex and thus includes some higher order visual
processing 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 than
this. (Reprinted from Joffe and Dunbar,28 Fig. 1.)
in volved. Th ese in teraction s cou ld occu r between th e sen sory an d association cortices (perceivin g an object),
between th e association an d fron tal
cortices (rem em berin g an object), an d
am on g all th ree (bein g aware of perceivin g an object). It is worth n otin g in
th is con text th at alth ou gh social skills
are com m on ly disru pted by dam age to
th e p r efr on ta l cor tex, m em or y for
even ts an d people is n ot.42
It seem s u n likely th at em otion al respon ses per se are th e su bstan tive
con strain t. Alth ou gh th e correct em ission an d in terpretation of em otion al
cu es is of sin gu lar im portan ce in th e
m an agem en t of social relation sh ips,45
th ere is little eviden ce th at th e su bcortical areas prin cipally associated with
em otion a l cu in g (for exa m p le, th e
am ygdala in th e lim bic system ) correlate in an y way with social grou p
size.28 In deed, Kevern e, Martel, an d
Nevison 46 poin t ou t th at th ere h as
been progressive redu ction in th e relative sizes of th e ‘‘em otion al’’ cen ters in
th e brain (th e h ypoth alam u s an d septu m ) in favor of th e ‘‘execu tive’’ cen ters (th e n eocortex an d striate cortex)
du rin g prim ate evolu tion . Th ey in ter-
pret th is in term s of a sh ift away from
em otion al con trol of beh avior to m ore
con sciou s, deliberate con trol.
Th e on ly rem ain in g altern ative is
th at th e m ech an ism s in volved lie in
th e ability to m an ipu late in form ation
abou t social relation sh ips th em selves.
Th is claim is su pported by six addition al lin es of eviden ce th at poin t to
th e fu n dam en tal im portan ce of social
skills in th e detailed m an agem en t of
social relation sh ips.
On e is th e fact th at close an alysis of
th e data on grou p size an d n eocortex
volu m e su ggests th at th ere are, in fact,
distin ct grades even with in th e an th ropoid prim ates (Fig. 3). Apes seem to lie
on a separate grade from th e m on keys,
wh ich in tu rn lie on a separate grade
from th e prosim ian s. Th e slope coefficien ts on th ese separate regression
lin es do n ot differ sign ifican tly, bu t th e
in tercepts do. It is as if apes requ ire
m ore com pu tin g power to m an age th e
sam e n u m ber of relation sh ips th at
m on keys do, an d m on keys in tu rn
requ ire m ore th an prosim ian s do. Th is
gradation correspon ds closely to th e
perceived scalin g of social com plexity.
Th e secon d lin e of eviden ce is th at
. . . there is no intrinsic
reason to suppose that
memory per se is the
issue. The social brain
hypothesis is about the
ability to manipulate
information, not simply to
remember it.
Th e fou rth lin e of eviden ce is Joffe’s 48 dem on stration th at adu lt n eocortex size in prim ates correlates with th e
len gth of th e ju ven ile period, bu t n ot
with th e len gth of gestation , lactation ,
or th e reprodu ctive life span , even
th ou gh total brain size in m am m als
correlates with th e len gth of th e gestation period.49,50 Th is su ggests th at wh at
is m ost im portan t in th e developm en t
of a large n eocortex in prim ates is n ot
th e em b r yologica l d evelop m en t of
brain tissu e per se, wh ich is associated
m a in ly with gesta tion len gth , b u t
rath er th e ‘‘software program m in g’’
th at occu rs du rin g th e period of social
learn in g between wean in g an d adu lth ood.
Fifth , Ku do, Lowen , an d Du n bar 51
h ave sh own th at groom in g cliqu e size,
a su rrogate variable th at in dexes alli-
ARTI CLES
186 Evolutionary Anthropology
gives prim ate social grou ps th eir in tern al stru ctu re an d coh eren ce, th is can
be seen as a cru cial basis for prim ate
sociality.
Fin ally, Kevern e, Martel, an d Nevison 46 h ave su ggested th at th e n eocortex an d striate cortex, th ose areas of
th e prim ate brain th at are respon sible
for execu tive fu n ction , are u n der m atern ally rath er th an patern ally im prin ted gen es (i.e., gen es th at ‘‘kn ow’’
wh ich p a r en t th ey ca m e fr om ),
wh ereas th e con verse is tru e for th e
lim bic system , th ose parts of th e brain
m ost closely a ssocia ted with em otion al beh avior. Th ey in terpret th is in
relation to th e cogn itive dem an ds of
th e m ore in ten se social life of fem ales
in m atrilin eal fem ale-bon ded societies.
IMPLICATIONS FOR HUMAN
GROUPS
Figure 5. Independent contrasts in the Spearman rank correlation (rs) between male rank and
mating success plotted against contrasts in neocortex size for two different male cohort sizes (4
to 8, and 9 to 30 males) for individual species. The regression equations for the two cohort sizes
are 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.)
an ce size, correlates rath er tigh tly with
relative n eocortex an d social grou p
size in prim ates, in clu din g h u m an s
(Fig. 6). Th e h u m an data derive from
two sa m p les: h a ir-ca r e n etwor ks
am on g fem ale bu sh m en 52 an d su pport
cliqu es am on g adu lts in th e Un ited
Kin gdom .53 Wh at is rem arkable is h ow
closely th e h u m an data fit with th e
d a ta fr om oth er p r im a te sp ecies.
Groom in g cliqu es of th is kin d in variably fu n ction as coalition s in prim ate
grou ps. Coalition s are fu n ction ally cru cial to in dividu als with in th ese grou ps
becau se th ey en able th e an im als to
m in im ize th e levels of h arassm en t an d
com petition th at th ey in evitably su ffer
wh en livin g in close proxim ity to oth ers.54 Coalition s essen tially allow prim ates to m an age a fin e balan cin g act
between keepin g oth er in dividu als off
th eir backs wh ile at th e sam e tim e
avoidin g drivin g th em away altogeth er
an d th ereby losin g th e ben efits for
wh ich th e grou ps form ed in th e first
place. Th ese resu lts can th u s probably
be in terpreted as a direct cogn itive
lim itation on th e n u m ber of in dividu als with wh ich an an im al can sim u ltan eou sly m ain tain a relation sh ip of su f-
ficien t depth th at th ey can be relied on
to provide u n stin tin g m u tu al su pport
wh en on e of th em is u n der attack.
Becau se th is is th e core process th at
Th e fact th at th e relation sh ip between n eocortex size an d wh at I will
term th e cogn itive grou p size h olds u p
so well in so m an y differen t taxon om ic
grou ps raises th e obviou s qu estion of
wh eth er or n ot it also applies to h u m an s. We can easily predict a valu e for
grou p size in h u m an s. Doin g so, wh ich
is sim ply a m atter of u sin g th e h u m an
n eocor tex volu m e to extr a p ola te a
valu e for grou p size from th e prim ate
Figure 6. Mean grooming clique size plotted against mean neocortex ratio for individual
primate genera. The square is Homo sapiens. Species sampled are L. catta, L. fulvus, Propithecus, 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.)
ARTI CLES
Evolutionary Anthropology 187
Figure 7. Mean sizes for different types of groups in traditional human societies. Individual
societies 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 settled
horticulturalists from Australia, Africa, Asia, and North and South America. The triangles give
mean group sizes for three contemporary United States samples: mean network size from
small-worlds experiments (N 5 2),67 mean Hutterite community size,68 and the size of an East
Tennessee mountain community.69 The value of 150 predicted by the primate neocortex size
relationship (from Fig. 1d) is indicated by the horizontal line, with 95% confidence intervals
shown as dashed lines.
equ ation , produ ces a valu e in th e order of 150. Th e real issu e is wh eth er
h u m an s really do go arou n d in grou ps
of th is size.
Id en tifyin g th e r eleva n t level of
grou pin g to m easu re in h u m an s is
difficu lt becau se m ost h u m an s live in
a ser ies of h ier a r ch ica lly in clu sive
grou ps. Th is, in itself, is n ot especially
u n u su a l: H ier a r ch ica lly str u ctu r ed
grou ps of th is kin d are ch aracteristic
of prim ates 54 an d m ay be typical of
m an y m am m als an d birds.55 At least
in th e case of th e diu rn al prim ates, it
seem s th at, with a few n otable exception s, th e variou s species’ grou pin g
pattern s exh ibit an overt level of stability at rou gh ly th e sam e position in th e
h ierarch y across a wide ran ge of taxa.
Moreover, becau se th e variou s layers
of th is h ierarch y appear to be in tim ately related to each oth er, probably
th rou gh bein g part of a series of cau sean d-con sequ en ce ch ain s,51 it wou ld
n ot m atter wh ich particu lar grou pin g
level (for exam ple, stable social grou p,
n etwor k, or gr oom in g cliqu e) wa s
taken to be th e grou pin g criterion .
Th e problem with respect to h u m an s is th at it is difficu lt to iden tify
wh ich of th e m an y poten tial grou pin g
levels is fu n ction ally or cogn itively
equ ivalen t to th e particu lar level of
grou pin g th at I h appen ed to u se for
prim ates. Th is difficu lty is particu larly
in tru sive in th is case becau se h u m an s
live in a dispersed social system som etim es referred to as a fission -fu sion
system . In order to get arou n d th is
problem , I adopted th e con verse strategy in m y origin al an alysis,56 askin g
wh eth er th ere was an y grou p size con sisten tly ch aracteristic of h u m an s th at
was of abou t th e requ isite size an d, if
so, wh eth er its in trin sic psych ological
ch aracteristics were sim ilar to th ose
fou n d in prim ate grou ps.
Becau se of th e stru ctu ral com plexity of postagricu ltu ral societies, I con sidered on ly tradition al h u n ter-gath er er a n d sm a ll-sca le h or ticu ltu r a l
societies. Alth ou gh cen su s data on
su ch societies are lim ited, th ose th at
are available su ggest th at th ere is in deed a con sisten t grou p size in th e
region of 150 in dividu als (Fig. 7). E xcept am on g settled h orticu ltu ralists,
wh ere th e village seem s to be th e
relevan t u n it, th is typically in volves
th e set of in dividu als from wh om overn igh t cam ps are easily an d regu larly
form ed. Su ch grou ps are n ot often
con spicu ou s as ph ysical en tities (th ey
do n ot often appear togeth er in on e
place at on e tim e), bu t th ey do in variably h ave im portan t ritu al fu n ction s
for th e in dividu als con cern ed. Am on g
Au stralian aborigin als, for exam ple,
th e relevan t grou p is th e clan , wh ich
m eets from tim e to tim e in jam borees
wh ere th e ritu als of life (m arriages
an d rites of passage) are en acted an d
tales of th e old tim es are reh earsed to
rem in d everyon e wh o th ey are an d
wh y th ey h old a particu lar relation sh ip to each oth er. In deed, th is gen u in ely seem s to be th e largest grou p of
people wh o kn ow everyon e in th e
grou p as in dividu als at th e level of
person al relation sh ips. Th is is essen tially th e defin ition th at h olds in th e
case of prim ates.
A m ore exten sive exploration of h u m an grou ps in oth er con texts su ggests
th at grou pin gs of th is size are widespread an d form an im portan t com pon en t of all h u m an social system s, bein g presen t in stru ctu res th at ran ge
from bu sin ess organ ization s to th e
arran gem en t of farm in g com m u n ities.56 E stim ates of com m u n ity size
for two tradition al farm in g com m u n ities in th e Un ited States, H u tterites
an d an E ast Ten n essee m ou n tain com m u n ity, an d of actu al social n etwork
sizes (from sm all-worlds experim en ts)
(sh own as trian gles on th e righ t side of
Fig. 7) fit very closely with in th e relevan t ran ge of grou p sizes.
It is easy, of cou rse, to play th e
n u m erologist in th is con text by fin din g grou ps th at fit wh atever grou p size
on e wish es to prom ote. Th e im portan t
featu re to n ote h ere, h owever, is th at
th e variou s h u m an grou ps th at can be
iden tified in an y society seem to clu ster rath er tigh tly arou n d a series of
valu es (5, 12, 35, 150, 500, an d 2,000)
with virtu ally n o overlap in th e varian ce arou n d th ese ch aracteristic valu es. Th ey seem to represen t poin ts of
stability or clu sterin g in th e degrees of
fam iliarity with in th e broad ran ge of
ARTI CLES
188 Evolutionary Anthropology
Box 2. A Beginner’s Guide to Intensionality
Computers can be said to know
things because their memories contain information; however, it seems
unlikely that they know that they know
these things, in that we have no evidence that they can reflect on their
states of ‘‘mind.’’ In the jargon of the
philosophy of mind, computers are
zero-order intensional machines. Intensionality (with an 2s) is the term
that philosophers of mind use to refer
to the state of having a state of mind
(knowing, believing, thinking, wanting,
understanding, intending, etc).
Most vertebrates are probably capable of reflecting on their states of
h u m an relation sh ips, from th e m ost
in tim ate to th e m ost ten u ou s.
COGNITIVE MECHANISMS
Th e su ggestion th a t th e m ech a n ism s in volved in th ese processes m ay
be con cern ed with social skills raises
th e issu e allu ded to by th e origin al
Mach iavellian in telligen ce h ypoth esis,
n am ely to wh at exten t cogn itively soph isticated m ech an ism s con ferrin g th e
ability to ‘‘m in d-read’’ m igh t be in volved. Tactical deception , in its stron g
sen se, im plies th e ability to h old false
beliefs an d, th u s, th e presen ce of th e
ability kn own as ‘‘th eory of m in d’’
(ToM). Of cou rse, tactical deception as
practiced by prim ates on a daily basis
m a y n ot, a s Byr n e 26 h im self h a s
poin ted ou t, be qu ite as soph isticated
as first im pression s su ggest. A m ore
con ven tion a l b eh a vior ist a ccou n t
based on sim ple associative learn in g
can in variably be given for alm ost all
exam ples reported in th e literatu re.
Non eth eless, con vin cin g eviden ce
su ggests th at h u m an s at least do u se
ToM in execu tin g som e of th eir m ore
m a n ip u la tive socia l a ctivities. An d
wh ile we m ay n ot wish to attribu te fu ll
ToM to all prim ates, at least circu m stan tial eviden ce su ggests th at basic
ToM is presen t in great apes an d th at
m on keys m ay aspire to a level th at
Byrn e 26 h as described as level 1.5 in ten tion ality (fu ll ToM bein g level 2
in ten tion ality) (see Box 2). Th e differen ce h a s b een su m m ed u p r a th er
mind, at least in some crude sense:
they know that they know. Organisms
of this kind are first-order intensional.
By extension, second-order intensional organisms know that someone
else knows something, and third-order
intensional organisms know that someone else knows that someone else
knows something. In principle, the sequence can be extended reflexively
indefinitely, although, in practice, humans rarely engage in more than
fourth-order intensionality in everyday
life and probably face an upper limit at
sixth-order (‘‘Peter knows that Jane
believes that Mark thinks that Paula
graph ically by Ch en ey an d Seyfarth ’s 57
observation th at apes seem to be good
psych ologists in th at th ey are good at
readin g m in ds, wh ereas m on keys are
good eth ologists in th at th ey are good
at readin g beh avior—or at least at
m akin g in feren ces abou t in ten tion s in
th e everyday sen se, even if n ot in th e
. . . apes seem to be
good psychologists in
that they are good at
reading minds, whereas
monkeys are good
ethologists in that they
are good at reading
behavior . . .
ph ilosoph ical sen se of belief states.
E viden ce th at ch im pan zees aspire
to at least a basic form of ToM is
provided by th eir perform an ce on experim en tal false-belief tasks.58–61 Th ese
stu dies h ave attem pted to develop an alogu es of th e classic false-belief tasks
u sed with ch ildren .62 Th ou gh it is clear
th at ch im pan zees do n ot perform to
th e level at wh ich fu lly com peten t
ch ildren perform , O’Con n ell’s 61 experim en ts at least su ggest th at th ey can
perform at th e level of ch ildren wh o
wants Jake to suppose that Amelia
intends to do something’’).
A minimum of fourth-order intensionality is required for literature that
goes beyond the merely narrative (‘‘the
writer wants the reader to believe that
character A thinks that character B
intends to do something’’). Similar abilities may be required for science, since
doing science requires us to ask
whether the world can be other than it
is (a second-order problem at the very
least) and then ask someone else to
do the same (an additional order of
intensionality).
stan d on th e th resh old of acqu irin g
ToM. More im portan tly, ch im pan zees
do better th an au tistic adu lts, on e of
wh ose defin in g featu res is th e lack of
ToM, on th e sam e tests.
Th at m in d-readin g, th e basis of ToM,
is difficu lt to do h as been sh own by
experim en ts on n orm al adu lts tested
on ‘‘advan ced’’ ToM tasks, u p to fifth order in ten tion ality.62 Th ese data su ggest th at n orm al h u m an s fin d tasks of
greater th an fou rth -order in ten tion ality exceedin gly h ard to do. Th e h igh
error rates at th ese levels do n ot reflect
a m em ory reten tion problem : All su bjects pass th e tests th at assess m em ory
for th e story lin e. Moreover, th e sam e
su bjects sh ow con siderable com peten ce on reason in g tasks th at in volve
cau sal ch ain s of u p to th e sixth order.
Th e difficu lty seem s gen u in ely to be
som eth in g to do with operatin g with
deeply em bedded m en tal states.
On e possibly sign ifican t observation
in th is con text is th at th e visu al an d
n on visu al com pon en ts of th e prim ate
n eocortex do n ot in crease isom etrically. Alth ou gh in itially th ere is a m ore
or less lin ear in crease in th e visu al
area V1 with in creasin g size of th e rest
of th e n eocortex, th is drops off with in
th e great ape clade. From gorillas
th rou gh h u m an s, in creases in th e size
of th e visu al area progress m ore slowly
th an do in creases in th e size of th e rest
of th e n eocortex.28 We in terpret th is as
im plyin g th at beyon d a certain poin t
th e acu ity of th e visu al system does
ARTI CLES
Evolutionary Anthropology 189
n ot in crease lin early with size. Becau se th e total size of th e n eocortex is
lim ited by em bryological an d en ergetic factors, th is m ean s th at disprop ortion ately m ore capacity can be
dedicated to n on visu al areas of th e
n eocortex on ce th e volu m e is above
th e cru cial th resh old. Th is m igh t explain wh y apes appear to be capable of
th e addition al cogn itive processin g associated with m in d readin g, wh ereas
m on keys are n ot. It m igh t also explain
wh y h u m an s are better at it th an apes.
For h u m an s, on e im portan t aspect
of ToM con cern s its relevan ce to lan gu age, a com m u n ication m ediu m th at
cru cially depen ds on u n derstan din g
in terlocu tors’ m en tal states or in ten tion s. Th e kin ds of m etaph orical u ses
of lan gu age th at ch aracterize n ot on ly
ou r rath er telegraph ic everyday exch an ges (in wh ich ‘‘you kn ow wh at I
m ean ?’’ is a com m on term in al clau se)
bu t also lies at th e very h eart of th e
m etaph orical featu res of lan gu age. As
stu dies of pragm atics h ave am ply dem on strated,63 a great deal of lin gu istic
com m u n ication is based on m etaph or:
Un derstan din g th e in ten tion s beh in d
a m etaph or is cru cial to su ccessfu l
com m u n ication . Failu re to u n derstan d
th ese in ten tion s com m on ly resu lts in
con fu sion or in appropriate respon ses.
In deed, with ou t th ese abilities it is
dou btfu l wh eth er literatu re, n otably
poetry, wou ld be possible. Ou r con versation s wou ld be con fin ed to th e ban ally factu al; th ose fin e n u an ces of
m ean in g th at create both th e am bigu ities of politen ess an d th e su btleties of
pu blic relation s wou ld n ot be possible.64
REFERENCES
1 Clu tton -Brock TH , H arvey PH (1980) Prim ates,
brain s an d ecology. J Zool Lon d 190:309–323.
2 Aiello LC, Wh eeler P (1995) Th e expen sive
tissu e h ypoth esis. Cu rr An th ropol 36:184–193.
3 Parker ST, Gibson KR (1977) Object m an ipu lation , tool u se an d sen sorim otor in telligen ce as
feedin g adaptation s in great apes an d cebu s
m on keys. J H u m E vol 6:623–641.
4 Gibson KR (1986) Cogn ition , brain size an d th e
extraction of em bedded food resou rces. In E lse J,
Lee PC (eds), Prim ate On togen y, Cogn ition an d
Social Beh aviou r, pp 93–104. Cam bridge: Cam bridge Un iversity Press.
5 Wh iten A, Byrn e R (1988) Tactical deception in
prim ates. Beh av Brain Sci 12:233–273.
6 Byrn e R, Wh iten A (eds) (1988) Mach iavellian
In telligen ce. Oxford: Oxford Un iversity Press.
7 H arcou rt AH (1988) Allian ces in con tests an d
social in telligen ce. In Byrn e R, Wh iten A (eds),
Mach iavellian In telligen ce, pp 142–152. Oxford:
Oxford Un iversity Press.
8 H arcou rt AH (1989) Sociality an d com petition
in prim ates an d n on -prim ates. In Stan den V,
Foley R (eds), Com parative Socioecology, pp 223–
242. Oxford: Blackwell Scien tific.
9 Broth ers L (1990) Th e social brain : A project
for in tegratin g prim ate beh aviou r an d n eu roph ysiology in a n ew dom ain . Con cepts Neu rosci 1:27–
251.
10 Barton RA, Du n bar RLM (1997) E volu tion of
th e social brain . In Wh iten A, Byrn e R (eds),
Machiavellian In telligen ce, Vol. II. Cam bridge:
Cam bridge Un iversity Press.
11 Fin lay BL, Darlin gton RB (1995) Lin ked regu larities in th e developm en t an d evolu tion of m am m alian brain s. Scien ce 268:1678– 1684.
12 Martin RD (1981) Relative brain size an d
m etabolic rate in terrestrial vertebrates. Natu re
Lon d 293:57–60.
13 Martin RD (1983) H u m an brain evolu tion in
an ecological con text. 52n d Jam es Arth u r Lectu re, Am erican Mu seu m of Natu ral H istory, New
York.
14 H ofm an MA (1983) E volu tion of th e brain in
n eon atal an d adu lt placen tal m am m als: A th eoretical approach . J Th eoret Biol 105:317–322.
15 Pagel M, H arvey PH (1988) H ow m am m als
produ ce large-brain ed offsprin g. E volu tion 42:
948–957.
16 Boesch C, Boesch H (1984) Men tal m ap in
ch im pan zees. Prim ates 25:110–170.
17 MacLean PD (1982) On th e origin an d progressive evolu tion of th e triu n e brain . In Arm stron g
E , Falk D (eds), Prim ate Brain Evolu tion , pp
291–310. New York: Plen u m Press.
18 Joffe T, Du n bar RIM (n .d.). Prim ate brain
system evolu tion . Brain Beh av E vol, su bm itted
for pu blication .
19 Passin gh am RE (1982) The Hu m an Prim ate.
San Fran cisco: Freem an .
20 Jersion H J (1973) Evolu tion of the Brain an d
In telligen ce. New York: Academ ic Press.
21 Du n bar RIM (1989) E n viron m en tal determ in an ts of in traspecific variation in body weigh t in
baboon s. J Zool Lon don 220:167–169.
22 Willn er LA (1989) Sexu al Dim orph ism in Prim ates. Ph .D. th esis, Un iversity of Lon don .
23 Deacon TW (1990) Fallacies of progression
in dices in th eories of brain -size evolu tion . In t J
Prim atol 12:193–236.
24 Du n bar RIM (1992) Neocortex size as a con strain t on grou p size in prim ates. J H u m E vol
20:469–493.
25 Du n bar RIM, Joffe TH (n .d.) Neocortex size
an d social grou p size in prosim ian s. Prim ates,
su bm itted for pu blication .
26 Byrn e RB (1995) The Thin kin g Prim ate. Oxford: Oxford Un iversity Press.
27 Barton RA (1995) Neocortex size an d beh aviou ral ecology in prim ates. Proc R Soc Lon don B,
263:173–177.
28 Joffe TH , Du n bar RIM (1997) Visu al an d
socio-cogn itive in form ation processin g in prim ate brain evolu tion . Proc R Soc Lon don B,
264:1303–1307.
29 Steph an H , Frah m H , Baron G (1981) New
an d revised data on volu m es of brain stru ctu res
in in sectivores an d pr im ates. Folia Pr im a tol
35:1–29.
30 Sawagu ch i T, Ku do H (1990) Neocortical developm en t an d social stru ctu re in prim ates. Prim ates 31:283–290.
31 Barton RA, Pu rvis A (1994) Prim ate brain s
an d ecology: Lookin g ben eath th e su rface. In
An derson JR, Th ierry B, H erren sch m idt N (eds),
Cu rren t Prim atology, pp 1–12. Strasbou rg: Un iversity of Strasbou rg Press.
32. O’Keefe J, Nadel L (1978) The Hippocam pu s
as a Cogn itive Map. Oxford: Oxford Un iversity
Press.
33. Krebs JH , Sh erry DF, H ealy SD, Perry VH ,
Vaccarin o AL (1989) H ippocam pal specialisation
of food-storin g birds. Proc Natl Acad Sci USA
86:1488–1492.
34. Aiello LC, Du n bar RIM (1993) Neocortex
size, grou p size an d th e evolu tion of lan gu age.
Cu rr An th ropol 34:184–193.
35 Du n bar RIM (1995) Neocortex size an d grou p
size in prim ates: A test of th e h ypoth esis. J H u m
E vol 28:287–296.
36. Bearder SK (1987) Lorises, bu sh babies an d
tarsiers: Diverse societies in solitary foragers. In
Sm u ts B, Ch en ey D, Seyfarth R, Wran gh am R,
Stru h saker T (eds), Prim ate S ocieties, pp 12–24.
Ch icago: Ch icago Un iversity Press.
37 Du n bar RIM, Bever J (n .d.) Neocortex size
determ in es grou p size in carn ivores an d in sectivores. E th ology, in press.
38 Gittlem an JH (1986) Carn ivore brain size,
beh aviou ral ecology an d ph ylogen y. J Mam m al
67:23–36.
39 Marin o L (1996) Wh at can dolph in s tell u s
abou t prim ate evolu tion ? E vol An th ropol 5:81–86.
40 Tsch u din A (1996) Th e Use of Neu roim agin g
in th e Assessm en t of Brain Size an d Stru ctu re in
Odon tocetes. MSc th esis, Un iversity of Natal
(Du rban ).
41 Barton RA, Pu rvis A, H arvey PH (1995) E volu tion ary radiation of visu al an d olfactory brain
system s in prim ates, bats an d in sectivores. Ph ilos
Tran s R Soc B 348:381–392.
42 Kolb B, Wish aw LO (1996) Fu n dam en tals of
Hu m an N eu ropsychology. San Fran cisco: Freem an .
43 Fletch er P, H appé F, Frith U, Baker SC, Dolan
RJ, Frakowiak RSJ, Frith CD (1996) Oth er m in ds
in th e brain : A fu n ction al im agin g stu dy of ‘‘th eory
of m in d’’ in story com preh en sion . Cogn ition 00:
000–000.
44 Frith C (1996) Brain m ech an ism s for ‘‘h avin g
a th eory of m in d.’’ J Psych oph arm acol 10:9–16.
45 Arm stron g E , Clarke MR, H ill E M (1987)
Relative size of th e an terior th alam ic n u clei differen tiates an th ropoids by social system . Brain Beh av E vol 30:263–271.
46 Kevern e E B, Martel FL, Nevison CM (1996)
Prim ate brain evolu tion : Gen etic an d fu n ction al
con sideration s. Proc R Soc Lon d B 262:689–696.
47 Pawlowski B, Du n bar R, Lowen C (n .d.) Neocortex size, social skills an d m atin g su ccess in
prim ates. Beh aviou r, in press.
48 Joffe TH (1997) Social pressu res h ave selected
for an exten ded ju ven ile period in prim ates. J
H u m E vol, in press.
49 Ben n ett PM, H arvey PH (1985) Brain size,
developm en t an d m etabolism in birds an d m am m als. J Zool Lon don 207:491–509.
50 Marin o L (1997) Th e relation sh ip between
gestation len gth , en ceph alisation an d body weigh t
in odon tocetes. Marin e Mam m al Sci 14:143–148.
51 Ku do H , Lowen S, Du n bar RIM (n .d.) Neocortex size an d social n etwork size in prim ates.
Beh aviou r, su bm itted for pu blication .
52 Su gawara K (1984) Spatial proxim ity an d
bodily con tact am on g th e cen tral Kalah ari San .
Afr Stu dies Mon ogr 3.
53 Du n bar RIM, Spoor M (1995) Social n etworks, su pport cliqu es an d kin sh ip. H u m Natu re
6:273–290.
54 Du n bar RIM (1988) Prim ate S ocial S ystem s.
Lon don : Ch apm an & H all.
55 Du n bar RIM (1989) Social system s as optim al
strategy sets. In Stan den V, Foley R (eds), Com parative S ocioecology, pp 141–149. Oxford: Blackwell Scien tific.
ARTI CLES
190 Evolutionary Anthropology
56 Du n bar RIM (1993) Coevolu tion of n eocortical size, grou p size an d lan gu age in h u m an s.
Beh av Brain Sci 11:681–735.
57 Ch en ey DL, Seyfarth RM (1990) How Mon keys
S ee the World. Ch icago: Ch icago Un iversity Press.
58 Povin elli DJ, Nelson KE (1990) In feren ces
abou t gu essin g an d kn owin g in ch im pan zees. J
Com p Psych ol 104:203–210.
59 Povin elli DJ (1994) Wh at ch im pan zees (m igh t)
kn ow abou t th e m in d. In Wran gh am R, McGrew
W, de Waal F, H eltn e P (eds), Chim pan zee Cu ltu res, pp 285–300. Cam bridge: H arvard Un iversity Press.
60 O’Con n ell SM (1996) Th eory of Min d in Ch im pan zees. Ph .D. th esis, Un iversity of Liverpool.
61 Pern er J, Wim m er D (1985) ‘‘Joh n th in ks th at
Mary th in ks th at. . . .’’ Attribu tion of secon dorder beliefs by 5 an d 10 year-old ch ildren . J E xp
Ch ild Psych ol 39:437–471.
62 Kin derm an P, Du n bar RIM, Ben tall RP (n .d.)
Th eory of m in d deficits an d cau sal attribu tion s.
Br J Psych ol, in press.
63 Desalles J-L (n .d.) Altru ism , statu s an d th e
origin of relevan ce. In H u rford JR, Stu ddertKen n edy M, Kn igh t C (eds), Evolu tion of Lan gu age. Cam bridge: Cam bridge Un iversity Press.
64 Du n bar RIM (1997) Groom in g, Gossip an d the
Evolu tion of Lan gu age. Cam bridge: H arvard Un iversity Press.
65 H arvey PH , Krebs JR (1990) Com parin g brain s.
Scien ce 249:150–156.
66 Steph an H (1972) E volu tion of prim ate brain s:
A com parative an atom ical in vestigation . In Tu ttle
R (ed), Fu n ction al an d Evolu tion ary Biology of
Prim ates, pp 165–174. Ch icago: Aldin e-Ath erton .
67 Killworth PD, Bern ard H R, McCarty C (1984)
Measu rin g pattern s of acqu ain tan cesh ip. Cu rr
An th ropol 25:391–397.
68 Man ge A, Man ge E (1980) Gen etics: Hu m an
Aspects. W.B. Sau n ders.
69 Bryan t FC (1981) We’re All Kin : A Cu ltu ral
S tu dy of a Mou n tain N eighbou rhood. Kn oxville:
Un iversity of Ten n essee Press.
70 Gou ld SJ (1975) Allom etry in prim ates, with
em ph asis on scalin g an d th e evolu tion of th e
brain . Con trib Prim atol 5:244–292.
71 Jolly A (1969) Lem u r social beh aviou r an d
prim ate in telligen ce. Scien ce 163:501–506.
72 H u m ph rey NK (1976) Th e social fu n ction of
in tellect. In Bateson PPG, H in de RA (eds) Grow in g Poin ts in Ethology, pp 303–317. Cam bridge:
Cam bridge Un iversity Press.
73 Arm stron g E (1985) Relative brain size in
m on keys an d prosim ian s. Am J Ph ys An th ropol
66:263–273.
r 1998 Wiley-Liss, Inc.