Stepwise evolution of stable sociality in primates.
ABSTRACT Although much attention has been focused on explaining and describing the diversity of social grouping patterns among primates, less effort has been devoted to understanding the evolutionary history of social living. This is partly because social behaviours do not fossilize, making it difficult to infer changes over evolutionary time. However, primate social behaviour shows strong evidence for phylogenetic inertia, permitting the use of Bayesian comparative methods to infer changes in social behaviour through time, thereby allowing us to evaluate alternative models of social evolution. Here we present a model of primate social evolution, whereby sociality progresses from solitary foraging individuals directly to large multi-male/multi-female aggregations (approximately 52 million years (Myr) ago), with pair-living (approximately 16 Myr ago) or single-male harem systems (approximately 16 Myr ago) derivative from this second stage. This model fits the data significantly better than the two widely accepted alternatives (an unstructured model implied by the socioecological hypothesis or a model that allows linear stepwise changes in social complexity through time). We also find strong support for the co-evolution of social living with a change from nocturnal to diurnal activity patterns, but not with sex-biased dispersal. This supports suggestions that social living may arise because of increased predation risk associated with diurnal activity. Sociality based on loose aggregation is followed by a second shift to stable or bonded groups. This structuring facilitates the evolution of cooperative behaviours and may provide the scaffold for other distinctive anthropoid traits including coalition formation, cooperative resource defence and large brains.
- SourceAvailable from: Alexandra E Müller[show abstract] [hide abstract]
ABSTRACT: The evolution and origin of primate social organisation has attracted the attention of many researchers, and a solitary pattern, believed to be present in most nocturnal prosimians, has been generally considered as the most primitive system. Nocturnal prosimians are in fact mostly seen alone during their nightly activities and therefore termed 'solitary foragers', but that does not mean that they are not social. Moreover, designating their social organisation as 'solitary', implies that their way of life is uniform in all species. It has, however, emerged over the last decades that all of them exhibit not only some kind of social network but also that those networks differ among species. There is a need to classify these social networks in the same manner as with group-living (gregarious) animals if we wish to link up the different forms of primate social organisation with ecological, morphological or phylogenetic variables. In this review, we establish a basic classification based on spatial relations and sociality in order to describe and cope properly with the social organisation patterns of the different species of nocturnal prosimians and other mammals that do not forage in cohesive groups. In attempting to trace the ancestral pattern of primate social organisation, the Malagasy mouse and dwarf lemurs and the Afro-Asian bushbabies and lorises are of special interest because they are thought to approach the ancestral conditions most closely. These species have generally been believed to exhibit a dispersed harem system as their pattern of social organisation ('dispersed' means that individuals forage solitarily but exhibit a social network). Therefore, the ancestral pattern of primate social organisation was inferred to be a dispersed harem. In fact, new field data on cheirogaleids combined with a review of patterns of social organisation in strepsirhines (lemurs, bushbabies and lorises) revealed that they exhibit either dispersed multi-male systems or dispersed monogamy rather than a dispersed harem system. Therefore, the concept of a dispersed harem system as the ancestral condition of primate social organisation can no longer be supported. In combination with data on social organisation patterns in 'primitive' placentals and marsupials, and in monotremes, it is in fact most probable that promiscuity is the ancestral pattern for mammalian social organisation. Subsequently, a dispersed multi-male system derived from promiscuity should be regarded as the ancestral condition for primates. We further suggest that the gregarious patterns of social organisation in Aotus and Avahi, and the dispersed form in Tarsius evolved from the gregarious patterns of diurnal primates rather than from the dispersed nocturnal type. It is consequently proposed that, in addition to Aotus and Tarsius, Avahi is also secondarily nocturnal.Biological Reviews 09/2000; 75(3):405-35. · 10.26 Impact Factor
Article: The Socioecology of Primate Groups11/2003; 17:111-136.
Article: Evolution of Primate Social Systems[show abstract] [hide abstract]
ABSTRACT: We review evolutionary processes and mechanisms that gave rise to the diversity of primate social systems. We define social organization, social structure and mating system as distinct components of a social system. For each component, we summarize levels and patterns of variation among primates and discuss evolutionary determinants of this variation. We conclude that conclusive explanations for a solitary life and pair-living are still lacking. We then focus on interactions among the 3 components in order to identify main targets of selection and potential constraints for social evolution. Social organization and mating system are more closely linked to each other than either one is to social structure. Further, we conclude that it is important to seek a priori measures for the effects of presumed selective factors and that the genetic contribution to social systems is still poorly examined. Finally, we examine the role of primate socio-ecology in current evolutionary biology and conclude that primates are not prominently represented because the main questions asked in behavioral ecology are often irrelevant for primate behavior. For the future, we see a rapprochement of these areas as the role of disease and life-history theory are integrated more fully into primate socio-ecology.International Journal of Primatology 07/2002; 23(4):707-740. · 1.79 Impact Factor
Stepwise evolution of stable sociality in primates
Susanne Shultz1, Christopher Opie1& Quentin D. Atkinson1,2
Although much attention has been focused on explaining and
describing the diversity of social grouping patterns among
primates1–3, less effort has been devoted to understanding the
evolutionary history of social living4. This is partly because social
behaviours do not fossilize, making it difficult to infer changes
over evolutionary time. However, primate social behaviour shows
strong evidence for phylogenetic inertia, permitting the use of
social evolution. Here we present a model of primate social evolu-
directly to large multi-male/multi-female aggregations (approxi-
mately 52 million years (Myr) ago), with pair-living (approximately
16Myr ago) or single-male harem systems (approximately 16Myr
ago) derivative from this second stage. This model fits the data
significantly better than the two widely accepted alternatives (an
unstructured model implied by the socioecological hypothesis or a
model that allows linear stepwise changes in social complexity
through time). We also find strong support for the co-evolution
of social living with a change from nocturnal to diurnal activity
patterns, but not with sex-biased dispersal. This supports sugges-
tionsthatsocial living mayarisebecauseofincreasedpredation risk
is followed by a second shift to stable or bonded groups. This struc-
turing facilitates the evolution of cooperative behaviours5and may
coalition formation, cooperative resource defence and large brains.
stable groups and bonded relationships between individuals6.
Explaining how primate social systems evolved is central to under-
standing the evolution of our closest relatives and the emergence of
early human social behaviour7. Conventional explanations have
appealed more to adaptive reasoning than phylogenetic history to
account for patterns of sociality4. Adaptive arguments often invoke
patterns of aggregation in response to ecological conditions3,9. This
focus has resulted in less emphasis on the historical processes and
phylogenetic constraints that have informed other areas of evolution-
highly inflexible social structures, and social traits cluster according to
taxonomic grouping across the order4. Strong historical constraints
make it crucial to incorporate phylogeny when testing adaptive expla-
tionary pathways leading to extant primate grouping patterns.
multi-male; see Supplementary Information for further discussion of
alternative classification schemes) for 217 species onto a primate
consensus tree (Fig. 1 and Supplementary Information) derived from
genetic data12. We then evaluated the strength of phylogenetic inertia
in the data (historical non-independence) using Pagel’s lambda (l)13.
A l value of 0 implies evolution independent of the phylogenetic tree,
between species is proportional to their relatedness. Social grouping
patterns showed a strong phylogenetic signal (lmax50.983, max-
imum likelihood (LLmax)52150.038) (significantly different from a
from a l value of 1 (LL152141.12, P50.189)). Flexible social struc-
ture is characteristic of only two groups, the Callitrichidae and
Lemuridae (Fig. 1).
tionary pathways leading to extant primate grouping patterns.
Theoretical models suggest two possibilities. First, the socioecological
patterns are facultative, transitions between all possible social states
(and polymorphic states within species) should be equally likely.
Second, primate social complexity has been proposed to increase in
a stepwise fashion from solitary individuals, through small groups to
large, socially complex groups14–16. From this ‘increasing complexity’
model we would predict that pair-living was the earliest form of social
a model of social evolution through pair-bonds has been found in
We used a Bayesian framework, implemented in BayesTraits20, to
evaluate four alternative models of social evolution (Fig. 2), including
the two described above, on a posterior distribution of primate trees.
The simplest model estimates a single rate of transition between all
in which all state changes occur at the same underlying rate. We
contrast this with a second, parameter-rich model in which rates are
allowed to vary across all transitions. This model implies that some
transitions are more likely than others, for example the rate from
solitary to pair-living may be different from the rate from pair-living
to solitary or to some other state—but does not make assumptions
complexity by restricting possible transitions to stepwise changes up
is derived from the data and identifies likely transitions using the
(setting to zero) transition rate parameters.
analysis (Supplementary Table 1) suggests that social evolution pro-
male groups, whereas transitions between pair-living and harems do
not occur. Transitions from solitary to social are not reversed; such
that once a lineage becomes social it remains so. We used Bayes
Factors20–22to test whether there is sufficient signal in the primate
sociality data to support decisively any of the four alternative models.
Table 1 shows that the reversible-jump-derived model is not only the
1Institute ofCognitive and EvolutionaryAnthropology,64 BanburyRoad,Universityof Oxford,Oxford OX26PN, UK.2DepartmentofPsychology,Universityof Auckland,Private Bag 92019,Auckland,New
1 0 N O V E M B E R 2 0 1 1 | V O L 4 7 9 | N A T U R E | 2 1 9
Macmillan Publishers Limited. All rights reserved
best fit to the data, but is also decisively better at explaining the data
than the equal rates, the fully parameterized or the increasing com-
We used the reversible-jump-derived model of social evolution to
reconstruct the evolutionary history of social organization across the
primate tree (Fig. 1). Ancestral node reconstructions reveal that the
transition from solitary foraging at the primate root (74Myr ago) to
social aggregations was established at the anthropoid root (52Myr
ago) and the root of the Indriidae and Lemuridae (32Myr ago) in
prosimians. Other forms of social grouping evolved later in primates;
harems appeared at the root of the Colobinae (16Myr ago), followed
soon after in the Cercopithecini (14Myr ago). Pair-living arose at the
root of the Callitrichidae (16Myr ago), Hylobatidae (8.6Myr ago),
Avahi (6.4Myr ago), hapalemurs (6.3Myr ago), Aotus (4.8Myr ago)
and Callicebus (4.5Myr ago). Thus, the fundamental shift to sociality
occurred with the appearance of aggregations, followed later by
derived grouping structures, including pair-living.
We next examined two possible catalysts of primate social evolu-
tion. First, the switch to social living is presumed to occur under
to diurnal activity. We used a test of co-evolution in BayesTraits20to
to social living. There was decisive support22for the dependent model
(that is, co-evolution between activity and sociality, Fig. 3a) over the
pared with LLI(independent)5241.7160.04 s.e.m.; Bayes Factor 3.39;
Supplementary Table 2), supporting the proposed link between the
evolution of activity patterns and social living. Additionally, both
80 60 40 0 20
0 0.5 1
0 0.5 1
0 0.5 1
0 0.5 1
0 0.5 1
0 0.5 1
Figure 1 | Primate phylogeny showing ancestral state reconstructions for
sociality under the reversible-jump Markov chain Monte Carlo-derived
model of evolution. The tree topology is the maximum clade credibility tree
from the 10kTrees Project12posterior distribution with branch lengths drawn
proportional to time. Branches and tips are coloured for solitary (purple), uni-
male (orange), multi-male (red), pair-living (pink) where the combined
combined probability is less than 0.7, the branch is grey. Histograms represent
theposteriorprobability distribution of each social state at thenodes indicated
Homo split; f, Old World monkey root).
2 2 0 | N A T U R E | V O L 4 7 9 | 1 0 N O V E M B E R 2 0 1 1
Macmillan Publishers Limited. All rights reserved
as the transition rate from these states to social/diurnal is an order of
magnitude higher than any other transition. This suggests that the
long been argued to provide anti-predator benefits1, and the shift to
diurnal social living in primates would have opened up a vast new
adaptive space in a highly visual world23.
The second possible catalyst is the switch to sex-biased dispersal,
characteristic24. Changes in dispersal behaviour may be important in
the evolution of sociality because in its extreme form, philopatry, one
sex foregoes dispersal and remains in the natal range resulting in kin
structured groups. A switch to sex-biased dispersal could therefore
facilitate kin selection and the emergence of cooperative social
groups25–27. The extension of the mother–daughter bond to groups of
underpinning mammalian sociality28. We used Discrete to evaluate
whether sex-biaseddispersal precedesthe shift tosociality inprimates.
Although we find support for co-evolution between social grouping
and dispersal patterns (mean LLD5273.2760.03 s.e.m. versus
is much weaker than between sociality and activity patterns and inde-
pendent models are sampled above chance (Supplementary Table 3).
primate (and mammalian) default24, the ancestral state for primates is
bi-sexual dispersal (posterior probability of 0.93) and the estimated
to sociality rather than precedes it (Fig. 3b).
Dispersal changes, therefore, do not trigger socialliving,butas they
follow the emergence of social living they could be associated with a
secondary transition to stable groups. A similar suggestion was put
forward in a controversial model for the evolution of cooperative
sociality in eusocial insects5. The model argues that aggregating indi-
viduals first create population structure. Stable groups then emerge
secondarily through increased persistence resulting from silenced dis-
persal in at least one sex. To test whether this model explains the
evolution of stable primate groups, we classified species as solitary,
unstable social or stable social (the later defined as species with natal
philopatry coupled with no/limited secondary dispersal or those with
stable, long-term pair bonds). We then used the reversible jump pro-
cedure to identify the most likely model for the evolution of group
stability; our model suggests that solitary living is the ancestral state,
groups (marginal LL5265.260.019, Supplementary Table 4). This
model is a better fit to the data than either an equal rates
(LL5271.5960.021, Bayes Factor52.77) or a parameter-rich
model, where transitionsare
(LL5269.0860.052; Bayes Factor51.66; Fig. 4). It thus appears
that although the evolution of social groups does not occur through
for a model of stepwise transitions leading from solitary living to
based on either kinship or reproductive ties. Although transitions to
social grouping are not uncommon in vertebrates, this secondary
transition to stable grouping is, and may hold the key to the evolu-
tion of cooperative sociality characteristic of anthropoid primates,
Our analyses demonstrate a model of primate social evolution,
which highlights the initial switch from solitary foraging to multi-
male/multi-female aggregations. Although we cannot directly test
adaptive explanations, our findings show this switch co-evolved with
a change from a nocturnal to a diurnal lifestyle, supporting the role of
predation in driving social evolution. Although group size has often
been used as a proxy of social complexity in primates, relationship
or group stability represents a more important indication of social
Equal rates (null) and parameter rich
Figure 2 | Alternative evolutionary models of primate social evolution.
Arrows represent allowable transitions between modes of social living under
each model. Under the complexity and parameter-rich model, transition rates
represented by each arrow can vary. Under the equal rates model, all rates are
fixed to a singleoptimized rate parameter. The reversible-jump-derived model
is a significantly better fit to the data than the alternative models.
Table 1 | Comparison of alternative model performance
Model RankParameters Mean likelihoodLog10[Bayes Factor]
Reversible-jump Markov chain Monte Carlo-derived model
Parameter-rich (unconstrained) model
Equal rate ‘null’ model
Table shows number of model parameters, model rank, likelihood and log10[Bayes Factors] (see Supplementary Information). The Bayes Factor indicates relative support for the reversible-jump-derived model
over alternatives (0–0.5 minimal; 0.5–1.0 substantial; 1.0–2.0 strong; .2.0 decisive)21.
Figure 3 | Estimated transition rates for co-evolution of social living.
Estimated transition rates with (a) activity and (b) dispersal patterns. Thin
lines, an estimated median transition rate .0 but ,0.01; heavy lines, a rate
.0.01; dashed lines, a median estimated zero transition rate. Full estimated
rates are reported in Supplementary Information.
1 0 N O V E M B E R 2 0 1 1 | V O L 4 7 9 | N A T U R E | 2 2 1
Macmillan Publishers Limited. All rights reserved
complexity6. Our models suggest that the initial switch to sociality
involved loose or unstable multi-male/multi-female aggregations
(as exemplified in diurnal lemurs) followed by secondary transitions
to bonded social relationships between mothers and daughters28
aggregations, with bonding associated with pair-living (for example,
birds, ungulates, carnivores), and kin-based groups limited to a few
phyla would reveal whether the pathways suggested for primate evolu-
tion are more widely characteristic of cooperative sociality.
To account for uncertainty in the underlying phylogeny, model testing was
undertaken across a Bayesian posteriordistribution of10,000ultrametric primate
trees derived from genetic data as part of version 2 of the 10kTrees Project12. The
maximum clade credibility tree we present was inferred from the complete
10kTrees sample using TreeAnnotator30. Pagel’s lambda was estimated using
the Ape and Geiger (see Supplementary Information) packages in R.
BayesTraits20uses an Markov chain Monte Carlo method to derive posterior
distributions of log-likelihoods, the rate parameters of models of evolution, and
parameter model). Rates were allowed to vary freely to parameterize the flexible
model. All rates in the increasing complexity model except the forward and
backward transitions between solitary/pair-living, pair-living/uni-male and uni-
male/multi-male groups were restricted to zero. Model transition rates were also
determined using the reversible-jump procedure in BayesTraits. Reversible-jump
models were ranked in order of their posterior probability to identify the top
ranked model. Model performance was compared using a log10[Bayes Factor]21.
Co-evolution between behavioural traits was assessed using the Discrete package
inBayesTraits. Socialorganizationwasclassedas solitary(0)orsocial(1)(includ-
bi-sexual (0) or sex-biased (1) (see Supplementary Materials for further details).
Full Methods and any associated references are available in the online version of
the paper at www.nature.com/nature.
Received 12 May; accepted 30 September 2011.
Terborgh, J. & Janson, C. H. The socioecology of primate groups. Annu. Rev. Ecol.
Syst. 17, 111–135 (1986).
Sterck, H. M., Watts, D. P. & van Schaik, C. P. The evolution of female social
relationships in nonhuman primates. Behav. Ecol. Sociobiol. 41, 291–309 (1997).
byusingphylogenetic methods. Proc.Natl Acad. Sci. USA 91, 9941–9945 (1994).
Nowak,M. A., Tarnita, C.E.& Wilson, E.O.Theevolution ofeusociality.Nature466,
Shultz, S.& Dunbar, R.I.M.The evolution ofthesocial brain: anthropoid primates
contrast with other vertebrates. Proc. R. Soc. Lond. B 274, 2429–2436 (2007).
Elton, S. Forty years on and still going strong: the use of hominin-cercopithecid
comparisons in palaeoanthropology. J. R. Anthropol. Inst. 12, 19–38 (2006).
Jarman, P. J. The social organisation of antelope in relation to their ecology.
Behaviour 48, 215–268 (1974).
Kappeler, P. M. & van Schaik, C. P. Evolution of primate social systems. Int. J.
Primatol. 23, 707–740 (2002).
10. Gould, S. J. & Lewontin, R. C. The spandrels of San Marco and the Panglossian
paradigm: a critique of the adaptationist programme. Proc. R. Soc. Lond. B 205,
11. Stirling, D. G., Re ´ale, D. & Roff, D. A. Selection, structure and the heritability of
behaviour. J. Evol. Biol. 15, 277–289 (2002).
12. Arnold, C., Matthews, L. J. & Nunn, C. L. The 10kTrees website: a new online
resource for primate phylogeny, Evol. Anthropol. 19, 114–118 (2010).
13. Pagel, M. Inferring the historical patterns of biological evolution. Nature 401,
14. Dunbar, R. I. M. in Primate Males: Causes and Consequences of Variation in Group
Composition (ed. Kappeler, P. M.) 259–268 (Cambridge Univ. Press, 2000).
16. Mu ¨ller, A. E. & Thalmann, U. Origin and evolution of primate social organisation: a
reconstruction. Biol. Rev. Camb. Philos. Soc. 75, 405–435 (2000).
17. Cornwallis, C. K., West, S. A., Davis, K. E. & Griffin, A. S. Promiscuity and the
evolutionary transition to complex societies. Nature 466, 969–972 (2010).
avian brain evolution. Biol. J. Linn. Soc. 100, 111–123 (2010).
19. Hughes, W. O. H., Oldroyd, B. P., Beekman, M. & Ratnieks, F. L. W. Ancestral
monogamy shows kin selection is key to the evolution of eusociality. Science 320,
20. Pagel, M. & Meade, A. Bayesian analysis of correlated evolution of discrete
charactersbyreversible-jump Markovchain MonteCarlo.Am. Nat.167,808–825
21. Rambaut, A. & Drummond, A. J. Tracer v1.4. Tracer Æhttp://beast.bio.ed.ac.uk/
22. Kass, R. E. & Raftery, A. E. Bayes Factors. J. Am. Stat. Assoc. 90, 773–795 (1995).
23. Barton, R. A. Visual specialization and brain evolution in primates. Proc. R. Soc.
Lond. B 265, 1933–1937 (1998).
Anim. Behav. 28, 1140–1162 (1980).
25. Frank, S. A. Foundations of Social Evolution (Princeton Univ. Press, 1998).
26. Hamilton, W. D. The genetical evolution of social behaviour. II. J. Theor. Biol. 7,
27. Gompper, M. E., Gittleman, J.L. & Wayne, R. K. Genetic relatedness, coalitions and
social behaviour of white-nosed coatis, Nasua narica. Anim. Behav. 53, 781–797
of mammaliansocial relationships. Phil. Trans. R. Soc. B 361, 2199–2214 (2006).
29. Shultz, S. & Dunbar, R. I. M. Encephalization is not a universal macroevolutionary
107, 21582–21586 (2010).
trees. BMC Evol. Biol. 7, 214 (2007).
Supplementary Information is linked to the online version of the paper at
We thank R.I.M. Dunbar for comments on the manuscript.
Author Contributions S.S. designed the study, compiled the data and executed
analyses. C.O. executed analyses. Q.A. was involved in study design and advised on
statistical analyses. All authors contributed to the manuscript.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of this article at
www.nature.com/nature. Correspondence and requests for materials should be
addressed to S.S. (email@example.com).
Figure 4 | Alternative evolutionary models for the evolution of stable
grouping patterns. a, The model with the highest posterior support for the
groups evolve from sociality through unstable social groups. b, Alternative
same rate (equal rates model) or allowed to vary (parameter-rich model).
2 2 2 | N A T U R E | V O L 4 7 9 | 1 0 N O V E M B E R 2 0 1 1
Macmillan Publishers Limited. All rights reserved
Primate data. Primates were classified as solitary, pair-living and group-living9;
group-living were further split into single and multi-male groups (Supplementary
were coded as follows: solitary (n540), pair-living (n553), single-male, multi-
female(n567)ormulti-male, multi-female (n5121). Dispersalwasclassifiedas
male-biased (n586), female-biased (n514) or bi-sexual dispersal (n5105).
Recent papers have argued that dispersal is more flexible than classification
schemes acknowledge36. However, here we attempt to capture the characteristic
cathermeral species were classed as polymorphic for activity. Species were also
classified in multiple states when variation between or within populations was
reported. One classification decision we faced was how to categorize species that
and stable social groups or had stable sleeping associations (for example, Loris,
Microcebus, Galago, Pongo). Although no primate is truly solitary, these species
are particularly problematical as a few well-documented studies suggest stable
community structures in nocturnal species, yet they do not form stable foraging
these species: (1) solitary foragers, (2) polymorphic (solitary foraging plus social
category), and (3) solitary foraging except for Pongo. This way we were able to
evaluate theimpact their classificationhadonmodel performance. The classifica-
tion of Pongo, had the highest mean likelihood (LL5264.80), followed by the
polymorphic classification scheme (LL5266.14), and finally the scheme that
classified Pongo as solitary (LL5271.71). Pongo classification affects model fit
versus non-zero) for each transition across all three classification schemes
(Supplementary Fig. 2). Finally, we classified stability based on reported adult
dispersal or migration events for both males (typically secondary dispersal after
joining new groups) and females (classified as post-partum dispersal). This
classification is more subjective than the previous traits as the data are limited
and often descriptive. We classified pair-living species as stable if group turnover
events were typically associated with death or severe injury to one of the adults
(rather than regular emigration by resident adults). For group-living species, we
defined stability as at least one sex typically remaining in the group throughout
adulthood (resulting in kin-based groups). The primary references that the
classifications were based on are found in the Supplementary Table 6.
Tree. The primate phylogeny was based on a sample of 10,000 ultrametric trees
from version 2 of the 10kTrees Project12. This provides a posterior distribution of
phylogenies using Bayesian inference from six mitochondrial (CYTB, COX1,
COX2, 12S rRNA, 16S rRNA and a gene cluster) and three autosomal genes
(MC1R, CCR5, SRY) for 230 primate species. The nodes of the consensus tree
known fossil calibration points12. The consensus tree is a maximum credibility
missing data, we included all species from the tree block rather than pruning the
tree to fit the data.
Phylogenetic signal. Phylogenetic signal in data indicates that related species are
more similar in a particular trait than would be expected by chance (that is, the
trait of a daughter species is not independent of that of the parent). To quantify
phylogenetic signal in our primate sociality data, we used the fitDiscrete function
in the Geiger38package in R to calculate the maximum likelihood value of Pagel’s
lambda13,39on the maximum credibility tree. A l value of 1 is consistent with a
of the phylogenetic tree40. A likelihood ratio test was used to compare the
fitted maximum likelihood value of l with a model implying no phylogenetic
signal (l50) to a model of evolution along the tree (l51). The likelihood ratio
test follows a x2distribution, with one degree of freedom. Polymorphisms were
collapsed such that flexible species were assigned an additional flexible social
Model settings and performance. To identify the model best supported by the
data for each analysis (social evolution, stability, social-activity and social-
dispersal models, plus the social-stability data sets), we used the Discrete and
Multistate optionin BayesTraits20. Webeganwith thereversible-jump procedure,
ranging between 0 and 2 (ref. 41). We initially explored using a uniform hyper-
prior to seed exponential rate priors with mean and variance ranging between 0
and 2. Model performance was robust to choice of hyper prior. ‘Rate dev’ settings
were set to achieve acceptance values within 20–40% (for most models this was
posterior distribution and trace of harmonic mean log-likelihoods; we assumed
convergence when this distribution was approximately normal, the likelihood
traces did not show large jumps across runs. Models visited by the Markov chain
were ranked in order of their posterior probability (Supplementary Tables 1–4).
The posteriorsample oftransitionratesforthesocialevolutionmodelisshownin
Supplementary Fig. 3.
Each Markov chain Monte Carlo simulation was run five times for 30million
iterations sampled every 100, with the first 25million iterations discarded as the
burn-in period. Examination of the post-burn-in log-likelihood and rate para-
meters across the Markov chain plotted in Tracer21indicated that runs had
for the parameters of interest were all above 2,000. We report the posterior dis-
tribution for rate parameters, marginal log-likelihoods21and states at ancestral
nodes from the run with the median likelihood.
Model comparison: social evolution. We constructed four different models of
a ‘flexible’ model. Third, we ran a ‘complexity’ model where transitions were
restricted so that movements were only allowed between solitary and pair-living,
pair-living and uni-male harems, and uni-male harems and multi-male social
organization.Finally, themodelstructurewith the highestposteriorsupportfrom
the reversible-jump analysis was run, allowing transitions from solitary to multi-
were set to zero. Final models were run using uniform rate priors (0–0.3) across a
range informed by either the reversible-jump analyses for the data driven models
ormaximumlikelihood analyses fortheoretical models.Examination ofposterior
the models was checked by evaluating variance in the mean log-likelihood values
over five iterations of the final analyses.
To compare alternative models of social evolution, we calculated both the
marginal likelihood and Bayes Factor(the ratio ofthe marginallikelihoods)using
Tracer21. The Bayes Factor (BF) shows the weight of evidence to support one
model over another, from 0 to 0.5 (minimal), to 0.5–1.0 (substantial), to 1.0–2.0
(strong), to greater than 2.0 (decisive)22.
Ancestral states. We used BayesTraits to infer the posterior probability of social
behaviours at each ancestral node in the primate tree under the model with the
highest posterior probability from the reversible-jump analysis. Although the
results presented in Supplementary Fig. 1 are drawn on the maximum clade
credibility tree, the analysis was performed across the posterior distribution of
10,000 primate trees. The ancestral state probabilities for each branch of the tree
are the combined posterior probability of each state on that branch with the
posterior probability that the branch itself exists.
test the hypotheses that either dispersal or activity patterns determine social
organization in primates by investigating the correlation and relative timing of
changes in social organization with those in dispersal and activity. We ran the
Discrete analysis with social organization as solitary (0) or social (1) (including
pair-living), dispersal as bi-sexual (0) or either female or male (1) and activity as
nocturnal (0) or diurnal (1) with cathermeral as (01). Model parameters and
Exponential rate priors were seeded from a uniform hyper prior with mean and
variance ranging between 0 and 2 (ref. 41). The posterior sample of reversible-
jump Markov chain Monte Carlo models for social-activity analyses is shown in
Supplementary Table 2, and for social-dispersal analyses in Supplementary Table
3. A Bayes Factor21comparison was made between the independent and the
dependent reversible-jump hyperprior model runs such that independent evolu-
entmodelsin thedependentrun to assess whetherthis wasabove chance20. Mean
and median transition rates for the two dependent analyses are reported in
Supplementary Table 5.
Perspective (Oxford Univ. Press, 2007).
32. Dixson, A. Primate Sexuality: Comparative Studies of the Prosimians, Monkeys, Apes,
and Human Beings (Oxford Univ. Press, 1998).
33. Dunbar, R. I. M. Primate Social Systems (Chapman & Hall, 1988).
(Univ. Chicago Press, 1987).
35. Isbell, L. A. & van Vuren, D. Differential costs of locationalandsocial dispersal and
Macmillan Publishers Limited. All rights reserved
36. Jack, K. M. & Isbell, L. A. Dispersal in primates: advancing an individualized
approach. Behaviour 146, 429–436 (2009).
37. Schulke, O. & Ostner, J. Big times for dwarfs: social organization, sexual
selection, and cooperation in the Cheirogaleidae. Evol. Anthropol. 14, 170–185
38. Harmon, L. J., Weir, J. T., Brock, C. D., Glor, R. E. & Challenger, W. GEIGER:
investigating evolutionary radiations. Bioinformatics 24, 129–131 (2008).
39. Pagel, M. D. Detecting correlated evolution on phylogenies: a general method for
the comparative analysis of discrete characters. Proc. R. Soc. Lond. B 255, 37–45
40. Freckleton, R. P., Harvey, P. H. & Pagel, M. D. Phylogenetic analysis and
comparative data: a test and review of evidence. Am. Nat. 160, 712–726 (2002).
41. Pagel, M. D., Meade, A. & Barker, D. Bayesian estimation of ancestral character
states on phylogenies. Syst. Biol. 53, 673–684 (2004).
Macmillan Publishers Limited. All rights reserved